Psychonomic Bulletin amp Review2002 9 (2) 185-211

In the past 30 years or so research in motor learning hascome a long way in describing and explaining how perfor-mance and learning of motor skills is affected by differentvariables These include for example the distribution ofpractice (massed vs distributed see eg Lee amp Genovese1988 1989) the feedback provided to the learner (timingtype frequency see eg Salmoni Schmidt amp Walter1984 Schmidt 1991a) the organization of practice (prac-tice variability contextual interference see eg Magill ampHall 1990 Shapiro amp Schmidt 1982) types of practice(eg physical observational mental see eg Jeannerod1994 McCullagh Weiss amp Ross 1989) and various guid-ance procedures (eg verbal physical guidance see egHagman 1983 Winstein Pohl amp Lewthwaite 1994) Ingeneral however the tasks used to examine the effects ofthese variables on learning have been relatively simplemdashoften including only one degree of freedom requiringcomparatively small amounts of practice to reach perfor-mance asymptotes and placing relatively modest demandson attention memory andor processing capacity Obviouslythere are several advantages to using more basic tasks of

the type typically used in the study of motor learning (egease and objectivity in measuring performance savings interms of money and other resources) Yet if the goal is tounderstand motor skill learning in general and to providerecommendations for the training of motor skills in ap-plied settings (eg in sports music or industry) it seemsto be necessary to study the acquisition and learning ofmore ldquocomplexrdquo skills that at least initially pose greaterchallenges to the cognitive capacity of the learner

Defining movement or task complexity for a wide vari-ety of motor tasks on a single task characteristic or move-ment outcome continuum is a difficult if not impossiblechallenge for many practical and theoretical reasons Forexample movement complexity has been proposed to in-crease with increases in reaction time (RT eg Henry ampRogers 1964 Klapp 1995) movement time (MT egFitts 1954) response errorsvariability (eg Schmidt Ze-laznik Hawkins Frank amp Quinn 1979) or number of de-grees of freedom (eg Bernstein 1967) Although each ofthe continua is relatively effective in describing task com-plexity when multiple tasks are carefully constructed alongthe continuum of interest no one continuum is satisfactoryin quantifying the complexity of the wide variety of motortasks that have been investigated For example MT is effec-tive in describing task difficulty for certain classes of sim-ple aiming tasks in which for example movement ampli-tude and target width are varied but is not effective inquantifying differences in complexity between aimingmovements and more continuous ski simulator or sta-

185 Copyright 2002 Psychonomic Society Inc

The preparation of this paper was supported by Grant PR 11818-2from the Deutsche Forschungsgemeinschaft We thank Robert ProctorRichard A Schmidt and an anonymous reviewer for their constructivecomments on earlier versions of the manuscript Correspondence con-cerning this article should be addressed to GWulf Department of Ki-nesiology University of Nevada Las Vegas 4505 Maryland ParkwayLas Vegas NV 89154-3034 (e-mail gabrielewulfccmailnevada edu)

Principles derived from the study of simple skillsdo not generalize to complex skill learning

GABRIELE WULFUniversity of Nevada Las Vegas Nevada

and

CHARLES H SHEATexas AampM University College Station Texas

We review research related to the learning of complex motor skills with respect to principles developedon the basis of simple skill learning Although some factors seem to have opposite effects on the learn-ing of simple and of complex skills other factors appear to be relevant mainly for the learning of morecomplex skills We interpret these apparently contradictory findings as suggesting that situations withlow processing demands benefit from practice conditions that increase the load and challenge the per-former whereas practice conditions that result in extremely high load should benefit from conditionsthat reduce the load to more manageable levels The findings reviewed here call into question the gen-eralizability of results from studies using simple laboratory tasks to the learning of complex motor skillsThey also demonstrate the need to use more complex skills in motor-learning research in order to gainfurther insights into the learning process

THEORETICAL AND REVIEW ARTICLES

186 WULF AND SHEA

bilometer tasks Likewise movement complexity tends toincrease for reaching tasks as the number of degrees of free-dom increases but complexity may actually decrease fortwo-handed juggling as compared with one-handed (cas-cade) juggling even though the number of degrees offreedom is increased for the two-handed case Thus in thelarger context the determination of task complexity musttake into consideration where a task fits on a number ofcontinua how the continua interact to fully define com-plexity and ultimately the demands placed on the mem-ory and processing capacity of the learner

This is complicated further if one considers that func-tional task complexity changes as a result of expertisePresumably this change is preceded by a concomitant re-duction in memory and processing demands as performancebecomes more automated That is as an individual per-former develops expertise with a task the functional com-plexity of the task decreases This could be expressed infaster RTs (eg Adams 1976) and MTs smaller responseerrors reduced attention demands (eg Schneider ampShiffrin 1977) increased movement efficiency (eg Du-rand Geoffroi Varray amp Preacutefaut 1994 Kahnemann1973) and either a decrease (freezing) or an increase (un-freezing) in the number of degrees of freedom (egVereijken van Emmerik Whiting amp Newell 1992) uti-lized to produce the movement after experience with atask as compared with initial trials of practice The impor-tant point is that although task complexity is a multidi-mensional construct that is hard to quantify across a widerange of tasks this quantification problem should not pre-clude making more qualitative judgments of task com-plexity so that we could conclude with reasonable cer-tainty that a ski simulator task for example that requiresseveral days of practice to develop moderate levels of skillis initially more complex than typical barrier-knockdowntasks or sequential keypress tasks in which performanceasymptotes are sometimes achieved in as little as 54 trials(eg Limons amp Shea 1988 J B Shea amp Morgan 1979J B Shea amp Titzer 1993)

Thus although an exact definition of complexity is notpossible for present purposes we will judge tasks to becomplex if they generally cannot be mastered in a singlesession have several degrees of freedom and perhaps tendto be ecologically valid Tasks will be judged as simple ifthey have only one degree of freedom can be mastered ina single practice session and appear to be artificial Al-though this portioning might not be perfect it will beshown to be useful nonetheless

Cognitive Effort and Information- Processing Demands

Several principles have emerged from the study of rel-atively simple skills in the last few years On the basis offindings from a number of different paradigms it has beensuggested that practice conditions that promote additionalcognitive effort (Lee Swinnen amp Serrien 1994) or requirethe learner to engage in additional information-processing

activities that are critical for test performance (Schmidt ampBjork 1992) are most effective for learning For examplebecause a random practice order of different tasks exercisesretrieval processes that later facilitate test performancemdashand therefore requires more effort than does blocked practicemdashperformance at the time of the test is usuallyenhanced (for a review see Magill amp Hall 1990) Simi-larly making augmented feedback difficult to usemdashforexample by delaying it or by withholding it on some prac-tice trialsmdashhas been argued to be beneficial for test per-formance because it forces learners to develop their owninternal error-correction-and-detection mechanisms (fora review see Schmidt 1991a) Thus adding an additionaldegree of ldquodifficultyrdquo for the learner during practice andthereby promoting greater effort andor information-processing activities on his or her part is assumed to resultin payoffs under ldquotestrdquo conditions

Although this notion is intuitively appealing the ques-tion is whether these and other principles developed pri-marily through the study of simple laboratory skills gen-eralize to the learning of more complex skills such as manyskills required in real life Consider a student driver for ex-ample who has to learn to coordinate the actions of the leftfoot depressing the clutch with those of the right foot onthe gas pedal those of the right arm shifting into the ap-propriate gears and those of the left arm steering and ac-tivating the signal (Similar difficulties are encountered byexperienced drivers having to drive a British car in theUK) Would learning be enhanced if the student driverpracticed under conditions that required additional cogni-tive effort and additional information-processing activi-ties If so should beginning drivers practice during therush hour in a big city where in addition to coordinatinghis or her limb movements the driver is required to nego-tiate his or her way through traffic Or would not onlyperformance but also learning (as measured by later per-formance under test conditions) be facilitated if the novicedriver practiced under conditions with little or no trafficwhere the processing demands are more manageable Inother words is it possible that complex skill learningwould benefit from reduced task demands during practiceas opposed to the learning of simple skills which seemsto benefit from increased task demands

For relatively simple skills a rough movement represen-tation may be developed within a few practice trials andit is easy to see that a further refinement of the skill is de-pendent on the extent to which the learner is challenged bythe practice conditions On the other hand the develop-ment of a movement representation for a more complex skillthat requires the coordination of many degrees of freedomand may initially be stored as a series of relatively indepen-dent subcomponents typically takes considerably longerand inherently requires more effort and information-processing activities on the part of the learner It is con-ceivable that introducing additional demands for thelearner during this process is actually detrimental ratherthan beneficial because the additional demands compete

COMPLEX SKILL LEARNING 187

with the essential processing activities for a limited amountof processing capacity during the learning process Insteadcomplex skill learning might be enhanced by providingthe learner with practice conditions that facilitate perfor-mance (at least until a relatively stable movement presen-tation is acquired) In addition to possible differential ef-fects that some variables might have on simple versuscomplex skill learning it is also possible that complexskills are sensitive to variables that are not particularly rel-evant for simple skills

In this paper we will try to assess what if anything thestudy of simple skills can tell us about the learning ofcomplex skills Although research using complex motorskills is still comparatively sparse in the past few years anincreasing number of studies have utilized more complexskills In the first part of this paper we will review variablesthat have been examined in the context of both simple andcomplex skill learning These variables include the sched-ule of practice (contextual interference CI) the feedbackgiven to the learner and the use of physical guidanceThese factors indeed seem to have different effects de-pending on the complexity of the skill indicating that atleast some of the principles that have been found for theacquisition of simple skills do not generalize very well tothe learning of more complex skills In the second partwe will review factors that seem to be particularly rele-vant for the learning of complex skillsmdashthat is observa-tion of and interaction between learners and the learnerrsquosfocus of attention These findings also challenge the as-sumption that a complete understanding of motor-learningprocesses can be gained by using simple laboratory-typetasks (eg Adams 1971) Rather they suggest that re-search should systematically include more complex tasksto enhance our insights into these processes and to enableus to provide adequate recommendations for practical ap-plications Only when simple and complex skills are sys-tematically studied can boundary conditions for soundmotor-learning principles be established In the final sec-tion we attempt to provide a synopsis and to give an out-line for future research

PRACTICE VARIABLES HAVING DIFFERENTIAL EFFECTS ON

SIMPLE AND COMPLEX SKILLS

This section focuses on variables that have been exam-ined fairly extensively in the motor-learning research lit-erature However most of these studies used rather simpletasks Only recently have researchers begun to examine theeffectiveness of such variables as CI the frequency andorganization of feedback or physical guidance on the learn-ing of more complex tasks

Contextual InterferenceJ B Shea and Morgan (1979) were the first to demon-

strate the CI effect for motor skill learning In their studyparticipants had to learn three different versions of a barrier-

knockdown task with the order of barriers being differentfor each task Practicing the tasks in a blocked order inwhich all trials on one task were completed before the par-ticipant was switched to the next task (low CI) resulted inmore effective performance during practice than did prac-ticing the tasks in a random order (high CI) However whenlearning was assessed in retention and transfer tests therandom-practice group demonstrated clearly superior per-formance to the blocked group

Since the J B Shea and Morgan (1979) study the learn-ing advantages of random as compared with blockedpractice have been replicated in numerous laboratory ex-periments Several studies used multisegment tasks sim-ilar to those in J B Shea and Morgan and CI effects havebeen found independently of whether the task required theparticipants to produce the movement sequence as fast aspossible (eg Gabriele Hall amp Buckolz 1987 Lee ampMagill 1983 Experiments 1 and 2 Limons amp Shea 1988J B Shea amp Zimny 1983) to meet a certain overall move-ment time (eg Carnahan Van Eerd amp Allard 1990Gabriele Hall amp Lee 1989 Gabriele Lee amp Hall 1991)or to produce certain segment times (eg Lee amp Magill1983 Experiment 3 Lee Wulf amp Schmidt 1992 SekiyaMagill Sidaway amp Anderson 1994 Wulf amp Lee 1993)Furthermore learning advantages for random over blockedpractice have been shown for very simple aiming tasks(Young Cohen amp Husak 1993) anticipation-timingtasks (eg Del Rey 1982 Del Rey Whitehurst Wughal-ter amp Barnwell 1983 Del Rey Wughalter amp Carnes 1987Del Rey Wughalter amp Whitehurst 1982 Goode 1986)movement patterning tasks (Sekiya Magill amp Anderson1996 Wulf 1992) and tracking tasks (Jelsma amp Pieters1989 Jelsma amp Van Merrieumlnboer 1989) Thus the CI ef-fect has proven to be a fairly robust phenomenon at leastfor the learning of simple skills in laboratory situations

To explain the CI phenomenon several hypotheseshavebeen put forward The most prominent ones are the elab-oration hypothesis of J B Shea and colleagues (J B Sheaamp Morgan 1979 J B Shea amp Zimny 1983) and the for-getting or reconstruction hypothesis of Lee and Magill(1983 1985) According to the elaboration hypothesisrandom practice leads to more distinctive and elaboratememory representations than does blocked practice be-cause participants use multiple and variable information-processing strategies Since the different tasks to be learnedreside together in working memory they can be comparedduring practice (which is not possible under blocked con-ditions) increasing the level of distinctiveness Also theuse of different encoding strategies supposedly leads to amore elaborate memorial representation than does the im-poverished encoding under blocked conditions The moredistinctive and elaborate representation of the skill afterrandom practice is assumed to be responsible for the moreeffective retention and transfer performance

According to the reconstruction hypothesis (Lee ampMagill 1983 1985) on the other hand the CI created byrandom practice leads to forgetting of the action plan or

188 WULF AND SHEA

motor program (Magill amp Hall 1990) owing to the inter-ference of the interspersed tasks Random practice there-fore necessitates repeated reconstructions of the motorprogram that are not necessary under blocked practiceconditions since the motor program is already in workingmemory The repeated action plan reconstructions in ran-dom practice are supposed to be responsible for the learn-ing advantages as compared with blocked practice

Other hypotheses have been suggested as well (J B Sheaamp Titzer 1993 Wulf amp Schmidt 1994) Even thoughthere is some support for all of the proposed hypotheses(for a review see Magill amp Hall 1990) it is still unclearwhich explanation is the most viable Also none of the ac-counts makes differential predictions regarding the effec-tiveness of random versus blocked practice for simple ver-sus complex skills Yet whereas the CI effect has beendemonstrated fairly consistently for relatively simple lab-oratory tasks this does not seem to be the case for morecomplex skills

One study that found benefits of high-CI practice forthe learning of a sport skill is a study by Smith and Davies(1995) who used the Pawlata kayak roll Participantswithout prior experience in kayaking showed more effec-tive retention (full roll) and transfer (half roll) if they hadfrequently alternated the direction of the roll during prac-tice as compared with participants who first completedhalf of the practice trials with the roll in their preferred di-rection and then performed the other half to the other sideAlso Wrisberg and colleagues (Wrisberg 1991 Wrisbergamp Liu 1991) have found additional convincing evidencefor the greater effectiveness of practice conditions that in-cluded frequent changes between tasks In both studiesalternating different serves produced more effective learn-ing than did blocked practice

However other studies using relatively complex taskshave yielded mixed results For example Tsutsui Lee andHodges (1998) found beneficial effects of random prac-tice for the learning of three patterns of a bimanual coor-dination task only if the patterns were practiced on differ-ent days (Experiment 2) but not if they were practiced onone day (Experiment 1) under blocked conditions Goodeand Magill (1986) had participants practice three types ofserves in a blocked a serial or a random order In reten-tion and transfer tests the random group tended to be moreeffective than the blocked group However these advan-tages were seen only for one serve and were significant inan interaction only when the superior performance of theblocked group on the last acquisition block was includedBortoli Robazza Durigon and Carra (1992) who exam-ined the learning of volleyball skills (volley bump serve)under different CI conditions also found only very limitedsupport for enhanced learning through increased CIThere were no learning advantages (as measured by accu-racy in hitting a target) of random or serial practice ascompared with blocked practice in retention or transferfrom a shorter distance Only when transfer was requiredfrom a longer distance did random practice and serial prac-

tice with low interference produce higher accuracy scoresthan did blocked or serial practice with high interference

Although the above-mentioned studies provide at leastsome evidence for the advantages of random practice forthe learning of complex skills there are other studies thatcall into question the effectiveness of random practice es-pecially for novice performers attempting to learn a com-plex movement pattern In a study by Hebert Landin andSolmon (1996) undergraduate students with differentskill levels (low skilled high skilled) practiced forehandand backhand tennis groundstrokes under either a blockedor an alternating schedule Whereas there was no differ-ence in the posttest score for high-skilled learners theblocked schedule was more advantageous than the alter-nating schedule for low-skilled learners

Several studies with children add support for the viewthat random practice apparently can be too demandingand therefore less effective for the learning of complexskills if performers are relatively inexperienced For ex-ample for teaching ninth-grade students different volley-ball skills (overhead set forearm pass serve) French Rinkand Werner (1990) found no differential effectiveness ofblocked random or mixed randomndashblocked practice sched-ules Farrow and Maschette (1997) had 8ndash9 year olds and10ndash12 year olds practice the tennis forehand groundstrokewith the preferred and the nonpreferred hands In a post-test random practice led to better preferred groundstrokesthan did blocked practice in the older group howeverblocked practice was more effective than random practicefor the younger group In addition blocked practiceproved to be generally more effective for performing thenonpreferred forehand Similarly Pinto-Zipp and Gentile(1995) found that blocked practice of a Frisbee-throwingtask benefited young children (5ndash6 years) whereas ran-dom practice was better for adults In fact they found thatwhen only limited amounts of practice were providedboth adults and children learned more via blocked than viarandom practice Pigott and Shapiro (1984) had children7 years 6 months of age practice throwing bean bags ofdifferent weights at a target They found that practice con-ditions with a medium level of CI (randomndashblocked prac-tice) was most effective and in contrast to random prac-tice produced significantly better learning than did blockedpractice Finally Al-Mustafa (1989) had 1- 5- 7- 11- and19-year-old participants practice a throwing task underrandom or blocked conditions In his experiment randompractice facilitated learning for older children and adultshowever younger children benefited from blocked practiceIn a second experiment Al-Mustafa found that 7-year-oldchildren with related movement experience achieved bet-ter retention results after random practice

Overall these results seem to indicate that when the tasksare more difficult because of high attention memory andor motor demands (or when learners are relatively inex-perienced) random practice may overload the system andthus disrupt the potential benefits of random practice Re-cently Albaret and Thon (1999) directly tested the hypoth-

COMPLEX SKILL LEARNING 189

esis that the complexity of the task could modulate the CIeffect They reasoned that complex tasks are accompaniedby relatively high memory demands that obscure the nor-mally beneficial effects of intertask interference In orderto manipulate task difficulty a set of drawing tasks wasconstructed that differed in terms of the number of linesegments The results indicated a clear advantage of ran-dom practice for the simplest version of the tasks on de-layed retention and transfer tests However the effect wassystematically reduced as the number of segments was in-creased and was even reversed (blocked practice betterthan random) albeit not significantly for the most diffi-cult tasks

J B Shea and colleagues (eg J B Shea amp Morgan1979 J B Shea amp Zimny 1983) proposed that one of themajor benefits of random practice arises from intertaskelaboration resulting from multiple items simultaneouslyheld in working memory being compared and contrastedHowever complex tasks presumably requiring consider-ably more memory capacity per task may exceed the ca-pacity of the performer This may be especially true earlyin practice where a single complex task may be initiallystored as a series of subcomponents (eg Povel amp Collard1982) In addition J B Shea and Zimny concede that inmore complex tasks there may be sufficient intratask in-terference (eg Battig 1972) arising from the elaborationrelative to the movement subcomponents of a single taskto promote learning nullifying the potential advantage ofintertask interference Consistent with the notion of over-loading working memory Wright Li and Whitacre (1992also see Lee Wishart Cunningham amp Carnahan 1997)found that additional processing disrupted learning underrandom-practice conditions but had positive effects underblocked conditions

In addition a number of recent experiments (eg Lai ampShea 1998 1999 Lai Shea Wulf amp Wright 2000 C H Shea Lai Wright Immink amp Black 2001 Whit-acre amp Shea 2000) consistently found relative timing ad-vantages for constant and blocked practice relative to se-rial and random practice directly opposite the typical CIeffect These experiments have used what on the surfaceappear to be relatively simple tasks which in other exper-iments (minus the relative timing demands) have pro-duced typical CI effects The difference is that participantshave been asked in these experiments to attempt not onlya goal movement outcome (as in most CI experiments)but also to produce a specific relative force andor timingpattern The additional demands of balancing relative andabsolute demands apparently significantly increase thecomplexity of the task and also the effect of CI In factwhen the relative timing pattern required is greatly sim-plified by making the segment requirements equal acrossthe task (Wright amp Shea 2001) the advantages of blockedrelative to random practice on relative timing learning arelost

Overall the picture that emerges from the studies re-viewed above seems to show that random practice is usu-

ally more effective than blocked practice when it comes tothe learning of simple skills for which the memory de-mands are relatively low andor when participants are ex-perienced to the extent that they have functionally reducedthe demands on memory (Hall Domingues amp Cavazos1994) For the learning of more complex skills for whichmemory and processing demands are high blocked prac-tice might be more effective at least early in the learningprocess Carr and Shepherd (2000) also extended this no-tion to neurological and some orthopedic patients wholike children and novices often have difficulty controllingthe intersegmental dynamics inherent in complex whole-body tasks Support for this view also comes from a studyby C H Shea Kohl and Indermill (1990) in which blockedpractice was more effective for the learning of a rapidforce production task when practice was limited to 50 tri-als However with increasing amounts of practice trials(200 400) random practice produced more effective re-tention performance than did blocked practice suggest-ing that a certain level of experience is required for ran-dom practice to become more effective than blockedpractice

This position can be reconciled with both the elabora-tion and the reconstruction accounts for the CI effect Forexample when complex tasks are first practiced short-term memory may be overloaded preventing or at leastlimiting processing (elaboration) of the various tasks Al-ternatively full reconstruction may be difficult from trialto trial In either case the development of memory repre-sentations or motor programs for the different tasks wouldbe degraded since random practice adds to the alreadyhigh processing loads The intratask interference inherentin complex tasks may be sufficient for effective learningunder blocked conditions at least in the early phases ofpractice Once the movements become more automatedmdashand therefore less difficult for the performermdashintroducingfurther challenges by varying the practice context mightprovide the necessary stimuli for continued learning

FeedbackOne of the most important variables in the motor-learning

process is the feedback provided to the learner attemptingto acquire a new motor skill Consequently this variablehas received a great amount of interest in the research lit-erature In this section we will review findings related tothe frequency of feedback and the organization of feed-back about different task components

Feedback frequency A number of studies mainlyusing relatively simple tasks have demonstrated that re-ducing the proportion of trials for which augmented feed-back (knowledge of results knowledge of performance) isprovided can result in more effective learning than doesgiving feedback after every single trial (eg Nicholson ampSchmidt 1991 Weeks amp Kordus 1998 Winstein ampSchmidt 1990 Wulf Lee amp Schmidt 1994 Wulf ampSchmidt 1989 Wulf Schmidt amp Deubel 1993 WulfShea amp Rice 1996 however see Lai amp Shea 1998) For

190 WULF AND SHEA

example using a lever-patterning task Winstein andSchmidt found that reducing the feedback frequency to50 of the trials was more beneficial for the learning ofthis task than giving 100 feedback Furthermore delay-ing the feedback for a few seconds has been found to pro-duce more effective learning than does giving feedbackimmediately after or even concurrently with the move-ment (eg Schmidt amp Wulf 1997 Swinnen SchmidtNicholson amp Shapiro 1990 Vander Linden Cauraugh ampGreene 1993) In particular concurrent feedback typi-cally has strong performance-enhancing effects duringpractice but results in clear performance decrements rel-ative to postresponse feedback when it is withdrawn inretention or transfer tests (Schmidt amp Wulf 1997 VanderLinden et al 1993 Winstein et al 1996)

Some of these findings appear to have parallels in theliterature on animal conditioning In this work partial re-inforcement has been shown to slow acquisition but alsoto maintain learning for much longer than continuous re-inforcement does The reason for this is seen mainly to liein the fact that partial reinforcement tends to simulate ex-tinction conditions (eg Tarpy 1982 chap 7) Howeveras Winstein and Schmidt (1990 see also Wulf amp Schmidt1989) have pointed out one set of findings that arguesagainst such an explanation for human learning is thatbeneficial learning effects of reduced feedback are seennot only in no-feedback retention tests but also in 100feedback retention tests Such a finding would seem un-likely in animal learning Furthermore immediate rein-forcement is more effective than delayed reinforcement inanimal learning but this is not the case for skill learningin humans This suggests that although the principles un-derlying animal reinforcement and feedback in humanshave some similarities they show some important differ-ences as well

The degrading effects of frequent and immediate feed-back on delayed retention and transfer tests found inhuman learning are typically explained with the guidancehypothesis (Salmoni et al 1984) According to this hy-pothesis feedback has positive effects such as guiding thelearner to the correct response however frequent feedbackis also argued to have several side effects that degradelearning For example learners seem to become too depen-dent on the information provided by the augmented feed-back and neglect the processing of intrinsic (eg proprio-ceptive) feedback which they will have to rely on whenthe additional information is no longer available (ie inno-feedback retention or transfer) In addition to makinglearners dependent on it frequent feedback has beenshown to make movement production quite variable pre-sumably preventing the learner from developing a stablemovement representation (Lai amp Shea 1998 Wulf ampSchmidt 1994 see also Schmidt 1991a) Finally feed-back has been argued to work proactively by facilitatingnext-response planning and retrieval In this sense fre-quent feedback might provide too much facilitation in theplanning of the subsequent response thereby reducing the

participantrsquos need to perform memory retrieval operationsthought to be critical for learning (Wulf amp Schmidt 1994)

Although the learning of simple skills seems to benefitfrom reducing or delaying the augmented feedback (al-though this effect appears to be stronger for variable prac-tice conditions see Lai amp Shea 1998) there is evidenceto suggest that more frequent feedback might be requiredfor the learning of complex skills Some support for thisnotion comes from studies on summary feedback (or sum-mary knowledge of results) where feedback about eachtrial is given only after a certain number of trials have beencompleted Whereas in experiments using relatively sim-ple tasks the largest summary feedback length proved tobe the most effective for learning (eg a summary of 16 tri-als in Gable Shea amp Wright 1991 20 trials in Lavery1962 15 trials in Schmidt Young Swinnen amp Shapiro1989) Schmidt Lange and Young (1990) showed that theoptimal number of summary trials for a more complexsimulated batting task was lower than those found formore simple tasks In this case a summary of 5 trials wasmore beneficial than longer (15) or shorter (1) feedbacksummaries Also Yao Fischman and Wang (1994) foundthat both summary feedback and average feedback (wherean average error or performance score is provided after aset of trials) about 5-trial blocks was more advantageousfor learning than was feedback about 15-trial blocks orevery-trial feedback Finally Guadagnoli Dornier andTandy (1996) directly demonstrated that task complexityas well as task-related experience interacted with the op-timal number of trials summarized That is whereas rela-tively long feedback summaries benefited the learning ofa relatively simple striking task for both novice and expe-rienced participants as well as the learning of a more com-plex double-striking task for experienced participants single-trial feedback was more effective than longer feed-back summaries for novices trying to learn a complextask

Presumably the most complex task that has been usedto examine the effects of a reduced feedback frequency isthe ski simulator task (Wulf Shea amp Matschiner 1998Experiment 2) The ski simulator consists of two bowedrails and a platform on wheels that is attached to the endsof the apparatus by elastic rubber belts (see Figure 1) Ifthe platform is displaced the rubber belts pull the plat-form back to its center position The platform can be madeto move sideways on the rails by exerting force on it Thegoal of the performer standing on the platform is to makeoscillatory slalom-type movements with the goal beingto produce the largest possible amplitudes (and sometimesfrequency) Similar to many sport skills this task requiresextensive practice and learners usually continue to showimprovements in performance across several days (egden Brinker amp van Hekken 1982 Durand et al 1994Vereijken 1991 Wulf Shea amp Matschiner 1998)

Wulf Shea and Matschiner (1998 Experiment 2) pro-vided learners on the ski simulator with feedback about aperformance measure the so-called relative force onset

COMPLEX SKILL LEARNING 191

Figure 1 Platform position and force curves of the left and right feet for 1 participant onDays 1 (top) 3 (middle) and 6 (bottom) in the Wulf Shea and Matschiner (1998) studyForce onsets are indicated on the force curves as can be seen the force onsets occurred laterand later in the movement cycle (eg between two successive reversals of the platform) acrosspractice

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

186 WULF AND SHEA

bilometer tasks Likewise movement complexity tends toincrease for reaching tasks as the number of degrees of free-dom increases but complexity may actually decrease fortwo-handed juggling as compared with one-handed (cas-cade) juggling even though the number of degrees offreedom is increased for the two-handed case Thus in thelarger context the determination of task complexity musttake into consideration where a task fits on a number ofcontinua how the continua interact to fully define com-plexity and ultimately the demands placed on the mem-ory and processing capacity of the learner

This is complicated further if one considers that func-tional task complexity changes as a result of expertisePresumably this change is preceded by a concomitant re-duction in memory and processing demands as performancebecomes more automated That is as an individual per-former develops expertise with a task the functional com-plexity of the task decreases This could be expressed infaster RTs (eg Adams 1976) and MTs smaller responseerrors reduced attention demands (eg Schneider ampShiffrin 1977) increased movement efficiency (eg Du-rand Geoffroi Varray amp Preacutefaut 1994 Kahnemann1973) and either a decrease (freezing) or an increase (un-freezing) in the number of degrees of freedom (egVereijken van Emmerik Whiting amp Newell 1992) uti-lized to produce the movement after experience with atask as compared with initial trials of practice The impor-tant point is that although task complexity is a multidi-mensional construct that is hard to quantify across a widerange of tasks this quantification problem should not pre-clude making more qualitative judgments of task com-plexity so that we could conclude with reasonable cer-tainty that a ski simulator task for example that requiresseveral days of practice to develop moderate levels of skillis initially more complex than typical barrier-knockdowntasks or sequential keypress tasks in which performanceasymptotes are sometimes achieved in as little as 54 trials(eg Limons amp Shea 1988 J B Shea amp Morgan 1979J B Shea amp Titzer 1993)

Thus although an exact definition of complexity is notpossible for present purposes we will judge tasks to becomplex if they generally cannot be mastered in a singlesession have several degrees of freedom and perhaps tendto be ecologically valid Tasks will be judged as simple ifthey have only one degree of freedom can be mastered ina single practice session and appear to be artificial Al-though this portioning might not be perfect it will beshown to be useful nonetheless

Cognitive Effort and Information- Processing Demands

Several principles have emerged from the study of rel-atively simple skills in the last few years On the basis offindings from a number of different paradigms it has beensuggested that practice conditions that promote additionalcognitive effort (Lee Swinnen amp Serrien 1994) or requirethe learner to engage in additional information-processing

activities that are critical for test performance (Schmidt ampBjork 1992) are most effective for learning For examplebecause a random practice order of different tasks exercisesretrieval processes that later facilitate test performancemdashand therefore requires more effort than does blocked practicemdashperformance at the time of the test is usuallyenhanced (for a review see Magill amp Hall 1990) Simi-larly making augmented feedback difficult to usemdashforexample by delaying it or by withholding it on some prac-tice trialsmdashhas been argued to be beneficial for test per-formance because it forces learners to develop their owninternal error-correction-and-detection mechanisms (fora review see Schmidt 1991a) Thus adding an additionaldegree of ldquodifficultyrdquo for the learner during practice andthereby promoting greater effort andor information-processing activities on his or her part is assumed to resultin payoffs under ldquotestrdquo conditions

Although this notion is intuitively appealing the ques-tion is whether these and other principles developed pri-marily through the study of simple laboratory skills gen-eralize to the learning of more complex skills such as manyskills required in real life Consider a student driver for ex-ample who has to learn to coordinate the actions of the leftfoot depressing the clutch with those of the right foot onthe gas pedal those of the right arm shifting into the ap-propriate gears and those of the left arm steering and ac-tivating the signal (Similar difficulties are encountered byexperienced drivers having to drive a British car in theUK) Would learning be enhanced if the student driverpracticed under conditions that required additional cogni-tive effort and additional information-processing activi-ties If so should beginning drivers practice during therush hour in a big city where in addition to coordinatinghis or her limb movements the driver is required to nego-tiate his or her way through traffic Or would not onlyperformance but also learning (as measured by later per-formance under test conditions) be facilitated if the novicedriver practiced under conditions with little or no trafficwhere the processing demands are more manageable Inother words is it possible that complex skill learningwould benefit from reduced task demands during practiceas opposed to the learning of simple skills which seemsto benefit from increased task demands

For relatively simple skills a rough movement represen-tation may be developed within a few practice trials andit is easy to see that a further refinement of the skill is de-pendent on the extent to which the learner is challenged bythe practice conditions On the other hand the develop-ment of a movement representation for a more complex skillthat requires the coordination of many degrees of freedomand may initially be stored as a series of relatively indepen-dent subcomponents typically takes considerably longerand inherently requires more effort and information-processing activities on the part of the learner It is con-ceivable that introducing additional demands for thelearner during this process is actually detrimental ratherthan beneficial because the additional demands compete

COMPLEX SKILL LEARNING 187

with the essential processing activities for a limited amountof processing capacity during the learning process Insteadcomplex skill learning might be enhanced by providingthe learner with practice conditions that facilitate perfor-mance (at least until a relatively stable movement presen-tation is acquired) In addition to possible differential ef-fects that some variables might have on simple versuscomplex skill learning it is also possible that complexskills are sensitive to variables that are not particularly rel-evant for simple skills

In this paper we will try to assess what if anything thestudy of simple skills can tell us about the learning ofcomplex skills Although research using complex motorskills is still comparatively sparse in the past few years anincreasing number of studies have utilized more complexskills In the first part of this paper we will review variablesthat have been examined in the context of both simple andcomplex skill learning These variables include the sched-ule of practice (contextual interference CI) the feedbackgiven to the learner and the use of physical guidanceThese factors indeed seem to have different effects de-pending on the complexity of the skill indicating that atleast some of the principles that have been found for theacquisition of simple skills do not generalize very well tothe learning of more complex skills In the second partwe will review factors that seem to be particularly rele-vant for the learning of complex skillsmdashthat is observa-tion of and interaction between learners and the learnerrsquosfocus of attention These findings also challenge the as-sumption that a complete understanding of motor-learningprocesses can be gained by using simple laboratory-typetasks (eg Adams 1971) Rather they suggest that re-search should systematically include more complex tasksto enhance our insights into these processes and to enableus to provide adequate recommendations for practical ap-plications Only when simple and complex skills are sys-tematically studied can boundary conditions for soundmotor-learning principles be established In the final sec-tion we attempt to provide a synopsis and to give an out-line for future research

PRACTICE VARIABLES HAVING DIFFERENTIAL EFFECTS ON

SIMPLE AND COMPLEX SKILLS

This section focuses on variables that have been exam-ined fairly extensively in the motor-learning research lit-erature However most of these studies used rather simpletasks Only recently have researchers begun to examine theeffectiveness of such variables as CI the frequency andorganization of feedback or physical guidance on the learn-ing of more complex tasks

Contextual InterferenceJ B Shea and Morgan (1979) were the first to demon-

strate the CI effect for motor skill learning In their studyparticipants had to learn three different versions of a barrier-

knockdown task with the order of barriers being differentfor each task Practicing the tasks in a blocked order inwhich all trials on one task were completed before the par-ticipant was switched to the next task (low CI) resulted inmore effective performance during practice than did prac-ticing the tasks in a random order (high CI) However whenlearning was assessed in retention and transfer tests therandom-practice group demonstrated clearly superior per-formance to the blocked group

Since the J B Shea and Morgan (1979) study the learn-ing advantages of random as compared with blockedpractice have been replicated in numerous laboratory ex-periments Several studies used multisegment tasks sim-ilar to those in J B Shea and Morgan and CI effects havebeen found independently of whether the task required theparticipants to produce the movement sequence as fast aspossible (eg Gabriele Hall amp Buckolz 1987 Lee ampMagill 1983 Experiments 1 and 2 Limons amp Shea 1988J B Shea amp Zimny 1983) to meet a certain overall move-ment time (eg Carnahan Van Eerd amp Allard 1990Gabriele Hall amp Lee 1989 Gabriele Lee amp Hall 1991)or to produce certain segment times (eg Lee amp Magill1983 Experiment 3 Lee Wulf amp Schmidt 1992 SekiyaMagill Sidaway amp Anderson 1994 Wulf amp Lee 1993)Furthermore learning advantages for random over blockedpractice have been shown for very simple aiming tasks(Young Cohen amp Husak 1993) anticipation-timingtasks (eg Del Rey 1982 Del Rey Whitehurst Wughal-ter amp Barnwell 1983 Del Rey Wughalter amp Carnes 1987Del Rey Wughalter amp Whitehurst 1982 Goode 1986)movement patterning tasks (Sekiya Magill amp Anderson1996 Wulf 1992) and tracking tasks (Jelsma amp Pieters1989 Jelsma amp Van Merrieumlnboer 1989) Thus the CI ef-fect has proven to be a fairly robust phenomenon at leastfor the learning of simple skills in laboratory situations

To explain the CI phenomenon several hypotheseshavebeen put forward The most prominent ones are the elab-oration hypothesis of J B Shea and colleagues (J B Sheaamp Morgan 1979 J B Shea amp Zimny 1983) and the for-getting or reconstruction hypothesis of Lee and Magill(1983 1985) According to the elaboration hypothesisrandom practice leads to more distinctive and elaboratememory representations than does blocked practice be-cause participants use multiple and variable information-processing strategies Since the different tasks to be learnedreside together in working memory they can be comparedduring practice (which is not possible under blocked con-ditions) increasing the level of distinctiveness Also theuse of different encoding strategies supposedly leads to amore elaborate memorial representation than does the im-poverished encoding under blocked conditions The moredistinctive and elaborate representation of the skill afterrandom practice is assumed to be responsible for the moreeffective retention and transfer performance

According to the reconstruction hypothesis (Lee ampMagill 1983 1985) on the other hand the CI created byrandom practice leads to forgetting of the action plan or

188 WULF AND SHEA

motor program (Magill amp Hall 1990) owing to the inter-ference of the interspersed tasks Random practice there-fore necessitates repeated reconstructions of the motorprogram that are not necessary under blocked practiceconditions since the motor program is already in workingmemory The repeated action plan reconstructions in ran-dom practice are supposed to be responsible for the learn-ing advantages as compared with blocked practice

Other hypotheses have been suggested as well (J B Sheaamp Titzer 1993 Wulf amp Schmidt 1994) Even thoughthere is some support for all of the proposed hypotheses(for a review see Magill amp Hall 1990) it is still unclearwhich explanation is the most viable Also none of the ac-counts makes differential predictions regarding the effec-tiveness of random versus blocked practice for simple ver-sus complex skills Yet whereas the CI effect has beendemonstrated fairly consistently for relatively simple lab-oratory tasks this does not seem to be the case for morecomplex skills

One study that found benefits of high-CI practice forthe learning of a sport skill is a study by Smith and Davies(1995) who used the Pawlata kayak roll Participantswithout prior experience in kayaking showed more effec-tive retention (full roll) and transfer (half roll) if they hadfrequently alternated the direction of the roll during prac-tice as compared with participants who first completedhalf of the practice trials with the roll in their preferred di-rection and then performed the other half to the other sideAlso Wrisberg and colleagues (Wrisberg 1991 Wrisbergamp Liu 1991) have found additional convincing evidencefor the greater effectiveness of practice conditions that in-cluded frequent changes between tasks In both studiesalternating different serves produced more effective learn-ing than did blocked practice

However other studies using relatively complex taskshave yielded mixed results For example Tsutsui Lee andHodges (1998) found beneficial effects of random prac-tice for the learning of three patterns of a bimanual coor-dination task only if the patterns were practiced on differ-ent days (Experiment 2) but not if they were practiced onone day (Experiment 1) under blocked conditions Goodeand Magill (1986) had participants practice three types ofserves in a blocked a serial or a random order In reten-tion and transfer tests the random group tended to be moreeffective than the blocked group However these advan-tages were seen only for one serve and were significant inan interaction only when the superior performance of theblocked group on the last acquisition block was includedBortoli Robazza Durigon and Carra (1992) who exam-ined the learning of volleyball skills (volley bump serve)under different CI conditions also found only very limitedsupport for enhanced learning through increased CIThere were no learning advantages (as measured by accu-racy in hitting a target) of random or serial practice ascompared with blocked practice in retention or transferfrom a shorter distance Only when transfer was requiredfrom a longer distance did random practice and serial prac-

tice with low interference produce higher accuracy scoresthan did blocked or serial practice with high interference

Although the above-mentioned studies provide at leastsome evidence for the advantages of random practice forthe learning of complex skills there are other studies thatcall into question the effectiveness of random practice es-pecially for novice performers attempting to learn a com-plex movement pattern In a study by Hebert Landin andSolmon (1996) undergraduate students with differentskill levels (low skilled high skilled) practiced forehandand backhand tennis groundstrokes under either a blockedor an alternating schedule Whereas there was no differ-ence in the posttest score for high-skilled learners theblocked schedule was more advantageous than the alter-nating schedule for low-skilled learners

Several studies with children add support for the viewthat random practice apparently can be too demandingand therefore less effective for the learning of complexskills if performers are relatively inexperienced For ex-ample for teaching ninth-grade students different volley-ball skills (overhead set forearm pass serve) French Rinkand Werner (1990) found no differential effectiveness ofblocked random or mixed randomndashblocked practice sched-ules Farrow and Maschette (1997) had 8ndash9 year olds and10ndash12 year olds practice the tennis forehand groundstrokewith the preferred and the nonpreferred hands In a post-test random practice led to better preferred groundstrokesthan did blocked practice in the older group howeverblocked practice was more effective than random practicefor the younger group In addition blocked practiceproved to be generally more effective for performing thenonpreferred forehand Similarly Pinto-Zipp and Gentile(1995) found that blocked practice of a Frisbee-throwingtask benefited young children (5ndash6 years) whereas ran-dom practice was better for adults In fact they found thatwhen only limited amounts of practice were providedboth adults and children learned more via blocked than viarandom practice Pigott and Shapiro (1984) had children7 years 6 months of age practice throwing bean bags ofdifferent weights at a target They found that practice con-ditions with a medium level of CI (randomndashblocked prac-tice) was most effective and in contrast to random prac-tice produced significantly better learning than did blockedpractice Finally Al-Mustafa (1989) had 1- 5- 7- 11- and19-year-old participants practice a throwing task underrandom or blocked conditions In his experiment randompractice facilitated learning for older children and adultshowever younger children benefited from blocked practiceIn a second experiment Al-Mustafa found that 7-year-oldchildren with related movement experience achieved bet-ter retention results after random practice

Overall these results seem to indicate that when the tasksare more difficult because of high attention memory andor motor demands (or when learners are relatively inex-perienced) random practice may overload the system andthus disrupt the potential benefits of random practice Re-cently Albaret and Thon (1999) directly tested the hypoth-

COMPLEX SKILL LEARNING 189

esis that the complexity of the task could modulate the CIeffect They reasoned that complex tasks are accompaniedby relatively high memory demands that obscure the nor-mally beneficial effects of intertask interference In orderto manipulate task difficulty a set of drawing tasks wasconstructed that differed in terms of the number of linesegments The results indicated a clear advantage of ran-dom practice for the simplest version of the tasks on de-layed retention and transfer tests However the effect wassystematically reduced as the number of segments was in-creased and was even reversed (blocked practice betterthan random) albeit not significantly for the most diffi-cult tasks

J B Shea and colleagues (eg J B Shea amp Morgan1979 J B Shea amp Zimny 1983) proposed that one of themajor benefits of random practice arises from intertaskelaboration resulting from multiple items simultaneouslyheld in working memory being compared and contrastedHowever complex tasks presumably requiring consider-ably more memory capacity per task may exceed the ca-pacity of the performer This may be especially true earlyin practice where a single complex task may be initiallystored as a series of subcomponents (eg Povel amp Collard1982) In addition J B Shea and Zimny concede that inmore complex tasks there may be sufficient intratask in-terference (eg Battig 1972) arising from the elaborationrelative to the movement subcomponents of a single taskto promote learning nullifying the potential advantage ofintertask interference Consistent with the notion of over-loading working memory Wright Li and Whitacre (1992also see Lee Wishart Cunningham amp Carnahan 1997)found that additional processing disrupted learning underrandom-practice conditions but had positive effects underblocked conditions

In addition a number of recent experiments (eg Lai ampShea 1998 1999 Lai Shea Wulf amp Wright 2000 C H Shea Lai Wright Immink amp Black 2001 Whit-acre amp Shea 2000) consistently found relative timing ad-vantages for constant and blocked practice relative to se-rial and random practice directly opposite the typical CIeffect These experiments have used what on the surfaceappear to be relatively simple tasks which in other exper-iments (minus the relative timing demands) have pro-duced typical CI effects The difference is that participantshave been asked in these experiments to attempt not onlya goal movement outcome (as in most CI experiments)but also to produce a specific relative force andor timingpattern The additional demands of balancing relative andabsolute demands apparently significantly increase thecomplexity of the task and also the effect of CI In factwhen the relative timing pattern required is greatly sim-plified by making the segment requirements equal acrossthe task (Wright amp Shea 2001) the advantages of blockedrelative to random practice on relative timing learning arelost

Overall the picture that emerges from the studies re-viewed above seems to show that random practice is usu-

ally more effective than blocked practice when it comes tothe learning of simple skills for which the memory de-mands are relatively low andor when participants are ex-perienced to the extent that they have functionally reducedthe demands on memory (Hall Domingues amp Cavazos1994) For the learning of more complex skills for whichmemory and processing demands are high blocked prac-tice might be more effective at least early in the learningprocess Carr and Shepherd (2000) also extended this no-tion to neurological and some orthopedic patients wholike children and novices often have difficulty controllingthe intersegmental dynamics inherent in complex whole-body tasks Support for this view also comes from a studyby C H Shea Kohl and Indermill (1990) in which blockedpractice was more effective for the learning of a rapidforce production task when practice was limited to 50 tri-als However with increasing amounts of practice trials(200 400) random practice produced more effective re-tention performance than did blocked practice suggest-ing that a certain level of experience is required for ran-dom practice to become more effective than blockedpractice

This position can be reconciled with both the elabora-tion and the reconstruction accounts for the CI effect Forexample when complex tasks are first practiced short-term memory may be overloaded preventing or at leastlimiting processing (elaboration) of the various tasks Al-ternatively full reconstruction may be difficult from trialto trial In either case the development of memory repre-sentations or motor programs for the different tasks wouldbe degraded since random practice adds to the alreadyhigh processing loads The intratask interference inherentin complex tasks may be sufficient for effective learningunder blocked conditions at least in the early phases ofpractice Once the movements become more automatedmdashand therefore less difficult for the performermdashintroducingfurther challenges by varying the practice context mightprovide the necessary stimuli for continued learning

FeedbackOne of the most important variables in the motor-learning

process is the feedback provided to the learner attemptingto acquire a new motor skill Consequently this variablehas received a great amount of interest in the research lit-erature In this section we will review findings related tothe frequency of feedback and the organization of feed-back about different task components

Feedback frequency A number of studies mainlyusing relatively simple tasks have demonstrated that re-ducing the proportion of trials for which augmented feed-back (knowledge of results knowledge of performance) isprovided can result in more effective learning than doesgiving feedback after every single trial (eg Nicholson ampSchmidt 1991 Weeks amp Kordus 1998 Winstein ampSchmidt 1990 Wulf Lee amp Schmidt 1994 Wulf ampSchmidt 1989 Wulf Schmidt amp Deubel 1993 WulfShea amp Rice 1996 however see Lai amp Shea 1998) For

190 WULF AND SHEA

example using a lever-patterning task Winstein andSchmidt found that reducing the feedback frequency to50 of the trials was more beneficial for the learning ofthis task than giving 100 feedback Furthermore delay-ing the feedback for a few seconds has been found to pro-duce more effective learning than does giving feedbackimmediately after or even concurrently with the move-ment (eg Schmidt amp Wulf 1997 Swinnen SchmidtNicholson amp Shapiro 1990 Vander Linden Cauraugh ampGreene 1993) In particular concurrent feedback typi-cally has strong performance-enhancing effects duringpractice but results in clear performance decrements rel-ative to postresponse feedback when it is withdrawn inretention or transfer tests (Schmidt amp Wulf 1997 VanderLinden et al 1993 Winstein et al 1996)

Some of these findings appear to have parallels in theliterature on animal conditioning In this work partial re-inforcement has been shown to slow acquisition but alsoto maintain learning for much longer than continuous re-inforcement does The reason for this is seen mainly to liein the fact that partial reinforcement tends to simulate ex-tinction conditions (eg Tarpy 1982 chap 7) Howeveras Winstein and Schmidt (1990 see also Wulf amp Schmidt1989) have pointed out one set of findings that arguesagainst such an explanation for human learning is thatbeneficial learning effects of reduced feedback are seennot only in no-feedback retention tests but also in 100feedback retention tests Such a finding would seem un-likely in animal learning Furthermore immediate rein-forcement is more effective than delayed reinforcement inanimal learning but this is not the case for skill learningin humans This suggests that although the principles un-derlying animal reinforcement and feedback in humanshave some similarities they show some important differ-ences as well

The degrading effects of frequent and immediate feed-back on delayed retention and transfer tests found inhuman learning are typically explained with the guidancehypothesis (Salmoni et al 1984) According to this hy-pothesis feedback has positive effects such as guiding thelearner to the correct response however frequent feedbackis also argued to have several side effects that degradelearning For example learners seem to become too depen-dent on the information provided by the augmented feed-back and neglect the processing of intrinsic (eg proprio-ceptive) feedback which they will have to rely on whenthe additional information is no longer available (ie inno-feedback retention or transfer) In addition to makinglearners dependent on it frequent feedback has beenshown to make movement production quite variable pre-sumably preventing the learner from developing a stablemovement representation (Lai amp Shea 1998 Wulf ampSchmidt 1994 see also Schmidt 1991a) Finally feed-back has been argued to work proactively by facilitatingnext-response planning and retrieval In this sense fre-quent feedback might provide too much facilitation in theplanning of the subsequent response thereby reducing the

participantrsquos need to perform memory retrieval operationsthought to be critical for learning (Wulf amp Schmidt 1994)

Although the learning of simple skills seems to benefitfrom reducing or delaying the augmented feedback (al-though this effect appears to be stronger for variable prac-tice conditions see Lai amp Shea 1998) there is evidenceto suggest that more frequent feedback might be requiredfor the learning of complex skills Some support for thisnotion comes from studies on summary feedback (or sum-mary knowledge of results) where feedback about eachtrial is given only after a certain number of trials have beencompleted Whereas in experiments using relatively sim-ple tasks the largest summary feedback length proved tobe the most effective for learning (eg a summary of 16 tri-als in Gable Shea amp Wright 1991 20 trials in Lavery1962 15 trials in Schmidt Young Swinnen amp Shapiro1989) Schmidt Lange and Young (1990) showed that theoptimal number of summary trials for a more complexsimulated batting task was lower than those found formore simple tasks In this case a summary of 5 trials wasmore beneficial than longer (15) or shorter (1) feedbacksummaries Also Yao Fischman and Wang (1994) foundthat both summary feedback and average feedback (wherean average error or performance score is provided after aset of trials) about 5-trial blocks was more advantageousfor learning than was feedback about 15-trial blocks orevery-trial feedback Finally Guadagnoli Dornier andTandy (1996) directly demonstrated that task complexityas well as task-related experience interacted with the op-timal number of trials summarized That is whereas rela-tively long feedback summaries benefited the learning ofa relatively simple striking task for both novice and expe-rienced participants as well as the learning of a more com-plex double-striking task for experienced participants single-trial feedback was more effective than longer feed-back summaries for novices trying to learn a complextask

Presumably the most complex task that has been usedto examine the effects of a reduced feedback frequency isthe ski simulator task (Wulf Shea amp Matschiner 1998Experiment 2) The ski simulator consists of two bowedrails and a platform on wheels that is attached to the endsof the apparatus by elastic rubber belts (see Figure 1) Ifthe platform is displaced the rubber belts pull the plat-form back to its center position The platform can be madeto move sideways on the rails by exerting force on it Thegoal of the performer standing on the platform is to makeoscillatory slalom-type movements with the goal beingto produce the largest possible amplitudes (and sometimesfrequency) Similar to many sport skills this task requiresextensive practice and learners usually continue to showimprovements in performance across several days (egden Brinker amp van Hekken 1982 Durand et al 1994Vereijken 1991 Wulf Shea amp Matschiner 1998)

Wulf Shea and Matschiner (1998 Experiment 2) pro-vided learners on the ski simulator with feedback about aperformance measure the so-called relative force onset

COMPLEX SKILL LEARNING 191

Figure 1 Platform position and force curves of the left and right feet for 1 participant onDays 1 (top) 3 (middle) and 6 (bottom) in the Wulf Shea and Matschiner (1998) studyForce onsets are indicated on the force curves as can be seen the force onsets occurred laterand later in the movement cycle (eg between two successive reversals of the platform) acrosspractice

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 187

with the essential processing activities for a limited amountof processing capacity during the learning process Insteadcomplex skill learning might be enhanced by providingthe learner with practice conditions that facilitate perfor-mance (at least until a relatively stable movement presen-tation is acquired) In addition to possible differential ef-fects that some variables might have on simple versuscomplex skill learning it is also possible that complexskills are sensitive to variables that are not particularly rel-evant for simple skills

In this paper we will try to assess what if anything thestudy of simple skills can tell us about the learning ofcomplex skills Although research using complex motorskills is still comparatively sparse in the past few years anincreasing number of studies have utilized more complexskills In the first part of this paper we will review variablesthat have been examined in the context of both simple andcomplex skill learning These variables include the sched-ule of practice (contextual interference CI) the feedbackgiven to the learner and the use of physical guidanceThese factors indeed seem to have different effects de-pending on the complexity of the skill indicating that atleast some of the principles that have been found for theacquisition of simple skills do not generalize very well tothe learning of more complex skills In the second partwe will review factors that seem to be particularly rele-vant for the learning of complex skillsmdashthat is observa-tion of and interaction between learners and the learnerrsquosfocus of attention These findings also challenge the as-sumption that a complete understanding of motor-learningprocesses can be gained by using simple laboratory-typetasks (eg Adams 1971) Rather they suggest that re-search should systematically include more complex tasksto enhance our insights into these processes and to enableus to provide adequate recommendations for practical ap-plications Only when simple and complex skills are sys-tematically studied can boundary conditions for soundmotor-learning principles be established In the final sec-tion we attempt to provide a synopsis and to give an out-line for future research

PRACTICE VARIABLES HAVING DIFFERENTIAL EFFECTS ON

SIMPLE AND COMPLEX SKILLS

This section focuses on variables that have been exam-ined fairly extensively in the motor-learning research lit-erature However most of these studies used rather simpletasks Only recently have researchers begun to examine theeffectiveness of such variables as CI the frequency andorganization of feedback or physical guidance on the learn-ing of more complex tasks

Contextual InterferenceJ B Shea and Morgan (1979) were the first to demon-

strate the CI effect for motor skill learning In their studyparticipants had to learn three different versions of a barrier-

knockdown task with the order of barriers being differentfor each task Practicing the tasks in a blocked order inwhich all trials on one task were completed before the par-ticipant was switched to the next task (low CI) resulted inmore effective performance during practice than did prac-ticing the tasks in a random order (high CI) However whenlearning was assessed in retention and transfer tests therandom-practice group demonstrated clearly superior per-formance to the blocked group

Since the J B Shea and Morgan (1979) study the learn-ing advantages of random as compared with blockedpractice have been replicated in numerous laboratory ex-periments Several studies used multisegment tasks sim-ilar to those in J B Shea and Morgan and CI effects havebeen found independently of whether the task required theparticipants to produce the movement sequence as fast aspossible (eg Gabriele Hall amp Buckolz 1987 Lee ampMagill 1983 Experiments 1 and 2 Limons amp Shea 1988J B Shea amp Zimny 1983) to meet a certain overall move-ment time (eg Carnahan Van Eerd amp Allard 1990Gabriele Hall amp Lee 1989 Gabriele Lee amp Hall 1991)or to produce certain segment times (eg Lee amp Magill1983 Experiment 3 Lee Wulf amp Schmidt 1992 SekiyaMagill Sidaway amp Anderson 1994 Wulf amp Lee 1993)Furthermore learning advantages for random over blockedpractice have been shown for very simple aiming tasks(Young Cohen amp Husak 1993) anticipation-timingtasks (eg Del Rey 1982 Del Rey Whitehurst Wughal-ter amp Barnwell 1983 Del Rey Wughalter amp Carnes 1987Del Rey Wughalter amp Whitehurst 1982 Goode 1986)movement patterning tasks (Sekiya Magill amp Anderson1996 Wulf 1992) and tracking tasks (Jelsma amp Pieters1989 Jelsma amp Van Merrieumlnboer 1989) Thus the CI ef-fect has proven to be a fairly robust phenomenon at leastfor the learning of simple skills in laboratory situations

To explain the CI phenomenon several hypotheseshavebeen put forward The most prominent ones are the elab-oration hypothesis of J B Shea and colleagues (J B Sheaamp Morgan 1979 J B Shea amp Zimny 1983) and the for-getting or reconstruction hypothesis of Lee and Magill(1983 1985) According to the elaboration hypothesisrandom practice leads to more distinctive and elaboratememory representations than does blocked practice be-cause participants use multiple and variable information-processing strategies Since the different tasks to be learnedreside together in working memory they can be comparedduring practice (which is not possible under blocked con-ditions) increasing the level of distinctiveness Also theuse of different encoding strategies supposedly leads to amore elaborate memorial representation than does the im-poverished encoding under blocked conditions The moredistinctive and elaborate representation of the skill afterrandom practice is assumed to be responsible for the moreeffective retention and transfer performance

According to the reconstruction hypothesis (Lee ampMagill 1983 1985) on the other hand the CI created byrandom practice leads to forgetting of the action plan or

188 WULF AND SHEA

motor program (Magill amp Hall 1990) owing to the inter-ference of the interspersed tasks Random practice there-fore necessitates repeated reconstructions of the motorprogram that are not necessary under blocked practiceconditions since the motor program is already in workingmemory The repeated action plan reconstructions in ran-dom practice are supposed to be responsible for the learn-ing advantages as compared with blocked practice

Other hypotheses have been suggested as well (J B Sheaamp Titzer 1993 Wulf amp Schmidt 1994) Even thoughthere is some support for all of the proposed hypotheses(for a review see Magill amp Hall 1990) it is still unclearwhich explanation is the most viable Also none of the ac-counts makes differential predictions regarding the effec-tiveness of random versus blocked practice for simple ver-sus complex skills Yet whereas the CI effect has beendemonstrated fairly consistently for relatively simple lab-oratory tasks this does not seem to be the case for morecomplex skills

One study that found benefits of high-CI practice forthe learning of a sport skill is a study by Smith and Davies(1995) who used the Pawlata kayak roll Participantswithout prior experience in kayaking showed more effec-tive retention (full roll) and transfer (half roll) if they hadfrequently alternated the direction of the roll during prac-tice as compared with participants who first completedhalf of the practice trials with the roll in their preferred di-rection and then performed the other half to the other sideAlso Wrisberg and colleagues (Wrisberg 1991 Wrisbergamp Liu 1991) have found additional convincing evidencefor the greater effectiveness of practice conditions that in-cluded frequent changes between tasks In both studiesalternating different serves produced more effective learn-ing than did blocked practice

However other studies using relatively complex taskshave yielded mixed results For example Tsutsui Lee andHodges (1998) found beneficial effects of random prac-tice for the learning of three patterns of a bimanual coor-dination task only if the patterns were practiced on differ-ent days (Experiment 2) but not if they were practiced onone day (Experiment 1) under blocked conditions Goodeand Magill (1986) had participants practice three types ofserves in a blocked a serial or a random order In reten-tion and transfer tests the random group tended to be moreeffective than the blocked group However these advan-tages were seen only for one serve and were significant inan interaction only when the superior performance of theblocked group on the last acquisition block was includedBortoli Robazza Durigon and Carra (1992) who exam-ined the learning of volleyball skills (volley bump serve)under different CI conditions also found only very limitedsupport for enhanced learning through increased CIThere were no learning advantages (as measured by accu-racy in hitting a target) of random or serial practice ascompared with blocked practice in retention or transferfrom a shorter distance Only when transfer was requiredfrom a longer distance did random practice and serial prac-

tice with low interference produce higher accuracy scoresthan did blocked or serial practice with high interference

Although the above-mentioned studies provide at leastsome evidence for the advantages of random practice forthe learning of complex skills there are other studies thatcall into question the effectiveness of random practice es-pecially for novice performers attempting to learn a com-plex movement pattern In a study by Hebert Landin andSolmon (1996) undergraduate students with differentskill levels (low skilled high skilled) practiced forehandand backhand tennis groundstrokes under either a blockedor an alternating schedule Whereas there was no differ-ence in the posttest score for high-skilled learners theblocked schedule was more advantageous than the alter-nating schedule for low-skilled learners

Several studies with children add support for the viewthat random practice apparently can be too demandingand therefore less effective for the learning of complexskills if performers are relatively inexperienced For ex-ample for teaching ninth-grade students different volley-ball skills (overhead set forearm pass serve) French Rinkand Werner (1990) found no differential effectiveness ofblocked random or mixed randomndashblocked practice sched-ules Farrow and Maschette (1997) had 8ndash9 year olds and10ndash12 year olds practice the tennis forehand groundstrokewith the preferred and the nonpreferred hands In a post-test random practice led to better preferred groundstrokesthan did blocked practice in the older group howeverblocked practice was more effective than random practicefor the younger group In addition blocked practiceproved to be generally more effective for performing thenonpreferred forehand Similarly Pinto-Zipp and Gentile(1995) found that blocked practice of a Frisbee-throwingtask benefited young children (5ndash6 years) whereas ran-dom practice was better for adults In fact they found thatwhen only limited amounts of practice were providedboth adults and children learned more via blocked than viarandom practice Pigott and Shapiro (1984) had children7 years 6 months of age practice throwing bean bags ofdifferent weights at a target They found that practice con-ditions with a medium level of CI (randomndashblocked prac-tice) was most effective and in contrast to random prac-tice produced significantly better learning than did blockedpractice Finally Al-Mustafa (1989) had 1- 5- 7- 11- and19-year-old participants practice a throwing task underrandom or blocked conditions In his experiment randompractice facilitated learning for older children and adultshowever younger children benefited from blocked practiceIn a second experiment Al-Mustafa found that 7-year-oldchildren with related movement experience achieved bet-ter retention results after random practice

Overall these results seem to indicate that when the tasksare more difficult because of high attention memory andor motor demands (or when learners are relatively inex-perienced) random practice may overload the system andthus disrupt the potential benefits of random practice Re-cently Albaret and Thon (1999) directly tested the hypoth-

COMPLEX SKILL LEARNING 189

esis that the complexity of the task could modulate the CIeffect They reasoned that complex tasks are accompaniedby relatively high memory demands that obscure the nor-mally beneficial effects of intertask interference In orderto manipulate task difficulty a set of drawing tasks wasconstructed that differed in terms of the number of linesegments The results indicated a clear advantage of ran-dom practice for the simplest version of the tasks on de-layed retention and transfer tests However the effect wassystematically reduced as the number of segments was in-creased and was even reversed (blocked practice betterthan random) albeit not significantly for the most diffi-cult tasks

J B Shea and colleagues (eg J B Shea amp Morgan1979 J B Shea amp Zimny 1983) proposed that one of themajor benefits of random practice arises from intertaskelaboration resulting from multiple items simultaneouslyheld in working memory being compared and contrastedHowever complex tasks presumably requiring consider-ably more memory capacity per task may exceed the ca-pacity of the performer This may be especially true earlyin practice where a single complex task may be initiallystored as a series of subcomponents (eg Povel amp Collard1982) In addition J B Shea and Zimny concede that inmore complex tasks there may be sufficient intratask in-terference (eg Battig 1972) arising from the elaborationrelative to the movement subcomponents of a single taskto promote learning nullifying the potential advantage ofintertask interference Consistent with the notion of over-loading working memory Wright Li and Whitacre (1992also see Lee Wishart Cunningham amp Carnahan 1997)found that additional processing disrupted learning underrandom-practice conditions but had positive effects underblocked conditions

In addition a number of recent experiments (eg Lai ampShea 1998 1999 Lai Shea Wulf amp Wright 2000 C H Shea Lai Wright Immink amp Black 2001 Whit-acre amp Shea 2000) consistently found relative timing ad-vantages for constant and blocked practice relative to se-rial and random practice directly opposite the typical CIeffect These experiments have used what on the surfaceappear to be relatively simple tasks which in other exper-iments (minus the relative timing demands) have pro-duced typical CI effects The difference is that participantshave been asked in these experiments to attempt not onlya goal movement outcome (as in most CI experiments)but also to produce a specific relative force andor timingpattern The additional demands of balancing relative andabsolute demands apparently significantly increase thecomplexity of the task and also the effect of CI In factwhen the relative timing pattern required is greatly sim-plified by making the segment requirements equal acrossthe task (Wright amp Shea 2001) the advantages of blockedrelative to random practice on relative timing learning arelost

Overall the picture that emerges from the studies re-viewed above seems to show that random practice is usu-

ally more effective than blocked practice when it comes tothe learning of simple skills for which the memory de-mands are relatively low andor when participants are ex-perienced to the extent that they have functionally reducedthe demands on memory (Hall Domingues amp Cavazos1994) For the learning of more complex skills for whichmemory and processing demands are high blocked prac-tice might be more effective at least early in the learningprocess Carr and Shepherd (2000) also extended this no-tion to neurological and some orthopedic patients wholike children and novices often have difficulty controllingthe intersegmental dynamics inherent in complex whole-body tasks Support for this view also comes from a studyby C H Shea Kohl and Indermill (1990) in which blockedpractice was more effective for the learning of a rapidforce production task when practice was limited to 50 tri-als However with increasing amounts of practice trials(200 400) random practice produced more effective re-tention performance than did blocked practice suggest-ing that a certain level of experience is required for ran-dom practice to become more effective than blockedpractice

This position can be reconciled with both the elabora-tion and the reconstruction accounts for the CI effect Forexample when complex tasks are first practiced short-term memory may be overloaded preventing or at leastlimiting processing (elaboration) of the various tasks Al-ternatively full reconstruction may be difficult from trialto trial In either case the development of memory repre-sentations or motor programs for the different tasks wouldbe degraded since random practice adds to the alreadyhigh processing loads The intratask interference inherentin complex tasks may be sufficient for effective learningunder blocked conditions at least in the early phases ofpractice Once the movements become more automatedmdashand therefore less difficult for the performermdashintroducingfurther challenges by varying the practice context mightprovide the necessary stimuli for continued learning

FeedbackOne of the most important variables in the motor-learning

process is the feedback provided to the learner attemptingto acquire a new motor skill Consequently this variablehas received a great amount of interest in the research lit-erature In this section we will review findings related tothe frequency of feedback and the organization of feed-back about different task components

Feedback frequency A number of studies mainlyusing relatively simple tasks have demonstrated that re-ducing the proportion of trials for which augmented feed-back (knowledge of results knowledge of performance) isprovided can result in more effective learning than doesgiving feedback after every single trial (eg Nicholson ampSchmidt 1991 Weeks amp Kordus 1998 Winstein ampSchmidt 1990 Wulf Lee amp Schmidt 1994 Wulf ampSchmidt 1989 Wulf Schmidt amp Deubel 1993 WulfShea amp Rice 1996 however see Lai amp Shea 1998) For

190 WULF AND SHEA

example using a lever-patterning task Winstein andSchmidt found that reducing the feedback frequency to50 of the trials was more beneficial for the learning ofthis task than giving 100 feedback Furthermore delay-ing the feedback for a few seconds has been found to pro-duce more effective learning than does giving feedbackimmediately after or even concurrently with the move-ment (eg Schmidt amp Wulf 1997 Swinnen SchmidtNicholson amp Shapiro 1990 Vander Linden Cauraugh ampGreene 1993) In particular concurrent feedback typi-cally has strong performance-enhancing effects duringpractice but results in clear performance decrements rel-ative to postresponse feedback when it is withdrawn inretention or transfer tests (Schmidt amp Wulf 1997 VanderLinden et al 1993 Winstein et al 1996)

Some of these findings appear to have parallels in theliterature on animal conditioning In this work partial re-inforcement has been shown to slow acquisition but alsoto maintain learning for much longer than continuous re-inforcement does The reason for this is seen mainly to liein the fact that partial reinforcement tends to simulate ex-tinction conditions (eg Tarpy 1982 chap 7) Howeveras Winstein and Schmidt (1990 see also Wulf amp Schmidt1989) have pointed out one set of findings that arguesagainst such an explanation for human learning is thatbeneficial learning effects of reduced feedback are seennot only in no-feedback retention tests but also in 100feedback retention tests Such a finding would seem un-likely in animal learning Furthermore immediate rein-forcement is more effective than delayed reinforcement inanimal learning but this is not the case for skill learningin humans This suggests that although the principles un-derlying animal reinforcement and feedback in humanshave some similarities they show some important differ-ences as well

The degrading effects of frequent and immediate feed-back on delayed retention and transfer tests found inhuman learning are typically explained with the guidancehypothesis (Salmoni et al 1984) According to this hy-pothesis feedback has positive effects such as guiding thelearner to the correct response however frequent feedbackis also argued to have several side effects that degradelearning For example learners seem to become too depen-dent on the information provided by the augmented feed-back and neglect the processing of intrinsic (eg proprio-ceptive) feedback which they will have to rely on whenthe additional information is no longer available (ie inno-feedback retention or transfer) In addition to makinglearners dependent on it frequent feedback has beenshown to make movement production quite variable pre-sumably preventing the learner from developing a stablemovement representation (Lai amp Shea 1998 Wulf ampSchmidt 1994 see also Schmidt 1991a) Finally feed-back has been argued to work proactively by facilitatingnext-response planning and retrieval In this sense fre-quent feedback might provide too much facilitation in theplanning of the subsequent response thereby reducing the

participantrsquos need to perform memory retrieval operationsthought to be critical for learning (Wulf amp Schmidt 1994)

Although the learning of simple skills seems to benefitfrom reducing or delaying the augmented feedback (al-though this effect appears to be stronger for variable prac-tice conditions see Lai amp Shea 1998) there is evidenceto suggest that more frequent feedback might be requiredfor the learning of complex skills Some support for thisnotion comes from studies on summary feedback (or sum-mary knowledge of results) where feedback about eachtrial is given only after a certain number of trials have beencompleted Whereas in experiments using relatively sim-ple tasks the largest summary feedback length proved tobe the most effective for learning (eg a summary of 16 tri-als in Gable Shea amp Wright 1991 20 trials in Lavery1962 15 trials in Schmidt Young Swinnen amp Shapiro1989) Schmidt Lange and Young (1990) showed that theoptimal number of summary trials for a more complexsimulated batting task was lower than those found formore simple tasks In this case a summary of 5 trials wasmore beneficial than longer (15) or shorter (1) feedbacksummaries Also Yao Fischman and Wang (1994) foundthat both summary feedback and average feedback (wherean average error or performance score is provided after aset of trials) about 5-trial blocks was more advantageousfor learning than was feedback about 15-trial blocks orevery-trial feedback Finally Guadagnoli Dornier andTandy (1996) directly demonstrated that task complexityas well as task-related experience interacted with the op-timal number of trials summarized That is whereas rela-tively long feedback summaries benefited the learning ofa relatively simple striking task for both novice and expe-rienced participants as well as the learning of a more com-plex double-striking task for experienced participants single-trial feedback was more effective than longer feed-back summaries for novices trying to learn a complextask

Presumably the most complex task that has been usedto examine the effects of a reduced feedback frequency isthe ski simulator task (Wulf Shea amp Matschiner 1998Experiment 2) The ski simulator consists of two bowedrails and a platform on wheels that is attached to the endsof the apparatus by elastic rubber belts (see Figure 1) Ifthe platform is displaced the rubber belts pull the plat-form back to its center position The platform can be madeto move sideways on the rails by exerting force on it Thegoal of the performer standing on the platform is to makeoscillatory slalom-type movements with the goal beingto produce the largest possible amplitudes (and sometimesfrequency) Similar to many sport skills this task requiresextensive practice and learners usually continue to showimprovements in performance across several days (egden Brinker amp van Hekken 1982 Durand et al 1994Vereijken 1991 Wulf Shea amp Matschiner 1998)

Wulf Shea and Matschiner (1998 Experiment 2) pro-vided learners on the ski simulator with feedback about aperformance measure the so-called relative force onset

COMPLEX SKILL LEARNING 191

Figure 1 Platform position and force curves of the left and right feet for 1 participant onDays 1 (top) 3 (middle) and 6 (bottom) in the Wulf Shea and Matschiner (1998) studyForce onsets are indicated on the force curves as can be seen the force onsets occurred laterand later in the movement cycle (eg between two successive reversals of the platform) acrosspractice

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

188 WULF AND SHEA

motor program (Magill amp Hall 1990) owing to the inter-ference of the interspersed tasks Random practice there-fore necessitates repeated reconstructions of the motorprogram that are not necessary under blocked practiceconditions since the motor program is already in workingmemory The repeated action plan reconstructions in ran-dom practice are supposed to be responsible for the learn-ing advantages as compared with blocked practice

Other hypotheses have been suggested as well (J B Sheaamp Titzer 1993 Wulf amp Schmidt 1994) Even thoughthere is some support for all of the proposed hypotheses(for a review see Magill amp Hall 1990) it is still unclearwhich explanation is the most viable Also none of the ac-counts makes differential predictions regarding the effec-tiveness of random versus blocked practice for simple ver-sus complex skills Yet whereas the CI effect has beendemonstrated fairly consistently for relatively simple lab-oratory tasks this does not seem to be the case for morecomplex skills

One study that found benefits of high-CI practice forthe learning of a sport skill is a study by Smith and Davies(1995) who used the Pawlata kayak roll Participantswithout prior experience in kayaking showed more effec-tive retention (full roll) and transfer (half roll) if they hadfrequently alternated the direction of the roll during prac-tice as compared with participants who first completedhalf of the practice trials with the roll in their preferred di-rection and then performed the other half to the other sideAlso Wrisberg and colleagues (Wrisberg 1991 Wrisbergamp Liu 1991) have found additional convincing evidencefor the greater effectiveness of practice conditions that in-cluded frequent changes between tasks In both studiesalternating different serves produced more effective learn-ing than did blocked practice

However other studies using relatively complex taskshave yielded mixed results For example Tsutsui Lee andHodges (1998) found beneficial effects of random prac-tice for the learning of three patterns of a bimanual coor-dination task only if the patterns were practiced on differ-ent days (Experiment 2) but not if they were practiced onone day (Experiment 1) under blocked conditions Goodeand Magill (1986) had participants practice three types ofserves in a blocked a serial or a random order In reten-tion and transfer tests the random group tended to be moreeffective than the blocked group However these advan-tages were seen only for one serve and were significant inan interaction only when the superior performance of theblocked group on the last acquisition block was includedBortoli Robazza Durigon and Carra (1992) who exam-ined the learning of volleyball skills (volley bump serve)under different CI conditions also found only very limitedsupport for enhanced learning through increased CIThere were no learning advantages (as measured by accu-racy in hitting a target) of random or serial practice ascompared with blocked practice in retention or transferfrom a shorter distance Only when transfer was requiredfrom a longer distance did random practice and serial prac-

tice with low interference produce higher accuracy scoresthan did blocked or serial practice with high interference

Although the above-mentioned studies provide at leastsome evidence for the advantages of random practice forthe learning of complex skills there are other studies thatcall into question the effectiveness of random practice es-pecially for novice performers attempting to learn a com-plex movement pattern In a study by Hebert Landin andSolmon (1996) undergraduate students with differentskill levels (low skilled high skilled) practiced forehandand backhand tennis groundstrokes under either a blockedor an alternating schedule Whereas there was no differ-ence in the posttest score for high-skilled learners theblocked schedule was more advantageous than the alter-nating schedule for low-skilled learners

Several studies with children add support for the viewthat random practice apparently can be too demandingand therefore less effective for the learning of complexskills if performers are relatively inexperienced For ex-ample for teaching ninth-grade students different volley-ball skills (overhead set forearm pass serve) French Rinkand Werner (1990) found no differential effectiveness ofblocked random or mixed randomndashblocked practice sched-ules Farrow and Maschette (1997) had 8ndash9 year olds and10ndash12 year olds practice the tennis forehand groundstrokewith the preferred and the nonpreferred hands In a post-test random practice led to better preferred groundstrokesthan did blocked practice in the older group howeverblocked practice was more effective than random practicefor the younger group In addition blocked practiceproved to be generally more effective for performing thenonpreferred forehand Similarly Pinto-Zipp and Gentile(1995) found that blocked practice of a Frisbee-throwingtask benefited young children (5ndash6 years) whereas ran-dom practice was better for adults In fact they found thatwhen only limited amounts of practice were providedboth adults and children learned more via blocked than viarandom practice Pigott and Shapiro (1984) had children7 years 6 months of age practice throwing bean bags ofdifferent weights at a target They found that practice con-ditions with a medium level of CI (randomndashblocked prac-tice) was most effective and in contrast to random prac-tice produced significantly better learning than did blockedpractice Finally Al-Mustafa (1989) had 1- 5- 7- 11- and19-year-old participants practice a throwing task underrandom or blocked conditions In his experiment randompractice facilitated learning for older children and adultshowever younger children benefited from blocked practiceIn a second experiment Al-Mustafa found that 7-year-oldchildren with related movement experience achieved bet-ter retention results after random practice

Overall these results seem to indicate that when the tasksare more difficult because of high attention memory andor motor demands (or when learners are relatively inex-perienced) random practice may overload the system andthus disrupt the potential benefits of random practice Re-cently Albaret and Thon (1999) directly tested the hypoth-

COMPLEX SKILL LEARNING 189

esis that the complexity of the task could modulate the CIeffect They reasoned that complex tasks are accompaniedby relatively high memory demands that obscure the nor-mally beneficial effects of intertask interference In orderto manipulate task difficulty a set of drawing tasks wasconstructed that differed in terms of the number of linesegments The results indicated a clear advantage of ran-dom practice for the simplest version of the tasks on de-layed retention and transfer tests However the effect wassystematically reduced as the number of segments was in-creased and was even reversed (blocked practice betterthan random) albeit not significantly for the most diffi-cult tasks

J B Shea and colleagues (eg J B Shea amp Morgan1979 J B Shea amp Zimny 1983) proposed that one of themajor benefits of random practice arises from intertaskelaboration resulting from multiple items simultaneouslyheld in working memory being compared and contrastedHowever complex tasks presumably requiring consider-ably more memory capacity per task may exceed the ca-pacity of the performer This may be especially true earlyin practice where a single complex task may be initiallystored as a series of subcomponents (eg Povel amp Collard1982) In addition J B Shea and Zimny concede that inmore complex tasks there may be sufficient intratask in-terference (eg Battig 1972) arising from the elaborationrelative to the movement subcomponents of a single taskto promote learning nullifying the potential advantage ofintertask interference Consistent with the notion of over-loading working memory Wright Li and Whitacre (1992also see Lee Wishart Cunningham amp Carnahan 1997)found that additional processing disrupted learning underrandom-practice conditions but had positive effects underblocked conditions

In addition a number of recent experiments (eg Lai ampShea 1998 1999 Lai Shea Wulf amp Wright 2000 C H Shea Lai Wright Immink amp Black 2001 Whit-acre amp Shea 2000) consistently found relative timing ad-vantages for constant and blocked practice relative to se-rial and random practice directly opposite the typical CIeffect These experiments have used what on the surfaceappear to be relatively simple tasks which in other exper-iments (minus the relative timing demands) have pro-duced typical CI effects The difference is that participantshave been asked in these experiments to attempt not onlya goal movement outcome (as in most CI experiments)but also to produce a specific relative force andor timingpattern The additional demands of balancing relative andabsolute demands apparently significantly increase thecomplexity of the task and also the effect of CI In factwhen the relative timing pattern required is greatly sim-plified by making the segment requirements equal acrossthe task (Wright amp Shea 2001) the advantages of blockedrelative to random practice on relative timing learning arelost

Overall the picture that emerges from the studies re-viewed above seems to show that random practice is usu-

ally more effective than blocked practice when it comes tothe learning of simple skills for which the memory de-mands are relatively low andor when participants are ex-perienced to the extent that they have functionally reducedthe demands on memory (Hall Domingues amp Cavazos1994) For the learning of more complex skills for whichmemory and processing demands are high blocked prac-tice might be more effective at least early in the learningprocess Carr and Shepherd (2000) also extended this no-tion to neurological and some orthopedic patients wholike children and novices often have difficulty controllingthe intersegmental dynamics inherent in complex whole-body tasks Support for this view also comes from a studyby C H Shea Kohl and Indermill (1990) in which blockedpractice was more effective for the learning of a rapidforce production task when practice was limited to 50 tri-als However with increasing amounts of practice trials(200 400) random practice produced more effective re-tention performance than did blocked practice suggest-ing that a certain level of experience is required for ran-dom practice to become more effective than blockedpractice

This position can be reconciled with both the elabora-tion and the reconstruction accounts for the CI effect Forexample when complex tasks are first practiced short-term memory may be overloaded preventing or at leastlimiting processing (elaboration) of the various tasks Al-ternatively full reconstruction may be difficult from trialto trial In either case the development of memory repre-sentations or motor programs for the different tasks wouldbe degraded since random practice adds to the alreadyhigh processing loads The intratask interference inherentin complex tasks may be sufficient for effective learningunder blocked conditions at least in the early phases ofpractice Once the movements become more automatedmdashand therefore less difficult for the performermdashintroducingfurther challenges by varying the practice context mightprovide the necessary stimuli for continued learning

FeedbackOne of the most important variables in the motor-learning

process is the feedback provided to the learner attemptingto acquire a new motor skill Consequently this variablehas received a great amount of interest in the research lit-erature In this section we will review findings related tothe frequency of feedback and the organization of feed-back about different task components

Feedback frequency A number of studies mainlyusing relatively simple tasks have demonstrated that re-ducing the proportion of trials for which augmented feed-back (knowledge of results knowledge of performance) isprovided can result in more effective learning than doesgiving feedback after every single trial (eg Nicholson ampSchmidt 1991 Weeks amp Kordus 1998 Winstein ampSchmidt 1990 Wulf Lee amp Schmidt 1994 Wulf ampSchmidt 1989 Wulf Schmidt amp Deubel 1993 WulfShea amp Rice 1996 however see Lai amp Shea 1998) For

190 WULF AND SHEA

example using a lever-patterning task Winstein andSchmidt found that reducing the feedback frequency to50 of the trials was more beneficial for the learning ofthis task than giving 100 feedback Furthermore delay-ing the feedback for a few seconds has been found to pro-duce more effective learning than does giving feedbackimmediately after or even concurrently with the move-ment (eg Schmidt amp Wulf 1997 Swinnen SchmidtNicholson amp Shapiro 1990 Vander Linden Cauraugh ampGreene 1993) In particular concurrent feedback typi-cally has strong performance-enhancing effects duringpractice but results in clear performance decrements rel-ative to postresponse feedback when it is withdrawn inretention or transfer tests (Schmidt amp Wulf 1997 VanderLinden et al 1993 Winstein et al 1996)

Some of these findings appear to have parallels in theliterature on animal conditioning In this work partial re-inforcement has been shown to slow acquisition but alsoto maintain learning for much longer than continuous re-inforcement does The reason for this is seen mainly to liein the fact that partial reinforcement tends to simulate ex-tinction conditions (eg Tarpy 1982 chap 7) Howeveras Winstein and Schmidt (1990 see also Wulf amp Schmidt1989) have pointed out one set of findings that arguesagainst such an explanation for human learning is thatbeneficial learning effects of reduced feedback are seennot only in no-feedback retention tests but also in 100feedback retention tests Such a finding would seem un-likely in animal learning Furthermore immediate rein-forcement is more effective than delayed reinforcement inanimal learning but this is not the case for skill learningin humans This suggests that although the principles un-derlying animal reinforcement and feedback in humanshave some similarities they show some important differ-ences as well

The degrading effects of frequent and immediate feed-back on delayed retention and transfer tests found inhuman learning are typically explained with the guidancehypothesis (Salmoni et al 1984) According to this hy-pothesis feedback has positive effects such as guiding thelearner to the correct response however frequent feedbackis also argued to have several side effects that degradelearning For example learners seem to become too depen-dent on the information provided by the augmented feed-back and neglect the processing of intrinsic (eg proprio-ceptive) feedback which they will have to rely on whenthe additional information is no longer available (ie inno-feedback retention or transfer) In addition to makinglearners dependent on it frequent feedback has beenshown to make movement production quite variable pre-sumably preventing the learner from developing a stablemovement representation (Lai amp Shea 1998 Wulf ampSchmidt 1994 see also Schmidt 1991a) Finally feed-back has been argued to work proactively by facilitatingnext-response planning and retrieval In this sense fre-quent feedback might provide too much facilitation in theplanning of the subsequent response thereby reducing the

participantrsquos need to perform memory retrieval operationsthought to be critical for learning (Wulf amp Schmidt 1994)

Although the learning of simple skills seems to benefitfrom reducing or delaying the augmented feedback (al-though this effect appears to be stronger for variable prac-tice conditions see Lai amp Shea 1998) there is evidenceto suggest that more frequent feedback might be requiredfor the learning of complex skills Some support for thisnotion comes from studies on summary feedback (or sum-mary knowledge of results) where feedback about eachtrial is given only after a certain number of trials have beencompleted Whereas in experiments using relatively sim-ple tasks the largest summary feedback length proved tobe the most effective for learning (eg a summary of 16 tri-als in Gable Shea amp Wright 1991 20 trials in Lavery1962 15 trials in Schmidt Young Swinnen amp Shapiro1989) Schmidt Lange and Young (1990) showed that theoptimal number of summary trials for a more complexsimulated batting task was lower than those found formore simple tasks In this case a summary of 5 trials wasmore beneficial than longer (15) or shorter (1) feedbacksummaries Also Yao Fischman and Wang (1994) foundthat both summary feedback and average feedback (wherean average error or performance score is provided after aset of trials) about 5-trial blocks was more advantageousfor learning than was feedback about 15-trial blocks orevery-trial feedback Finally Guadagnoli Dornier andTandy (1996) directly demonstrated that task complexityas well as task-related experience interacted with the op-timal number of trials summarized That is whereas rela-tively long feedback summaries benefited the learning ofa relatively simple striking task for both novice and expe-rienced participants as well as the learning of a more com-plex double-striking task for experienced participants single-trial feedback was more effective than longer feed-back summaries for novices trying to learn a complextask

Presumably the most complex task that has been usedto examine the effects of a reduced feedback frequency isthe ski simulator task (Wulf Shea amp Matschiner 1998Experiment 2) The ski simulator consists of two bowedrails and a platform on wheels that is attached to the endsof the apparatus by elastic rubber belts (see Figure 1) Ifthe platform is displaced the rubber belts pull the plat-form back to its center position The platform can be madeto move sideways on the rails by exerting force on it Thegoal of the performer standing on the platform is to makeoscillatory slalom-type movements with the goal beingto produce the largest possible amplitudes (and sometimesfrequency) Similar to many sport skills this task requiresextensive practice and learners usually continue to showimprovements in performance across several days (egden Brinker amp van Hekken 1982 Durand et al 1994Vereijken 1991 Wulf Shea amp Matschiner 1998)

Wulf Shea and Matschiner (1998 Experiment 2) pro-vided learners on the ski simulator with feedback about aperformance measure the so-called relative force onset

COMPLEX SKILL LEARNING 191

Figure 1 Platform position and force curves of the left and right feet for 1 participant onDays 1 (top) 3 (middle) and 6 (bottom) in the Wulf Shea and Matschiner (1998) studyForce onsets are indicated on the force curves as can be seen the force onsets occurred laterand later in the movement cycle (eg between two successive reversals of the platform) acrosspractice

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 189

esis that the complexity of the task could modulate the CIeffect They reasoned that complex tasks are accompaniedby relatively high memory demands that obscure the nor-mally beneficial effects of intertask interference In orderto manipulate task difficulty a set of drawing tasks wasconstructed that differed in terms of the number of linesegments The results indicated a clear advantage of ran-dom practice for the simplest version of the tasks on de-layed retention and transfer tests However the effect wassystematically reduced as the number of segments was in-creased and was even reversed (blocked practice betterthan random) albeit not significantly for the most diffi-cult tasks

J B Shea and colleagues (eg J B Shea amp Morgan1979 J B Shea amp Zimny 1983) proposed that one of themajor benefits of random practice arises from intertaskelaboration resulting from multiple items simultaneouslyheld in working memory being compared and contrastedHowever complex tasks presumably requiring consider-ably more memory capacity per task may exceed the ca-pacity of the performer This may be especially true earlyin practice where a single complex task may be initiallystored as a series of subcomponents (eg Povel amp Collard1982) In addition J B Shea and Zimny concede that inmore complex tasks there may be sufficient intratask in-terference (eg Battig 1972) arising from the elaborationrelative to the movement subcomponents of a single taskto promote learning nullifying the potential advantage ofintertask interference Consistent with the notion of over-loading working memory Wright Li and Whitacre (1992also see Lee Wishart Cunningham amp Carnahan 1997)found that additional processing disrupted learning underrandom-practice conditions but had positive effects underblocked conditions

In addition a number of recent experiments (eg Lai ampShea 1998 1999 Lai Shea Wulf amp Wright 2000 C H Shea Lai Wright Immink amp Black 2001 Whit-acre amp Shea 2000) consistently found relative timing ad-vantages for constant and blocked practice relative to se-rial and random practice directly opposite the typical CIeffect These experiments have used what on the surfaceappear to be relatively simple tasks which in other exper-iments (minus the relative timing demands) have pro-duced typical CI effects The difference is that participantshave been asked in these experiments to attempt not onlya goal movement outcome (as in most CI experiments)but also to produce a specific relative force andor timingpattern The additional demands of balancing relative andabsolute demands apparently significantly increase thecomplexity of the task and also the effect of CI In factwhen the relative timing pattern required is greatly sim-plified by making the segment requirements equal acrossthe task (Wright amp Shea 2001) the advantages of blockedrelative to random practice on relative timing learning arelost

Overall the picture that emerges from the studies re-viewed above seems to show that random practice is usu-

ally more effective than blocked practice when it comes tothe learning of simple skills for which the memory de-mands are relatively low andor when participants are ex-perienced to the extent that they have functionally reducedthe demands on memory (Hall Domingues amp Cavazos1994) For the learning of more complex skills for whichmemory and processing demands are high blocked prac-tice might be more effective at least early in the learningprocess Carr and Shepherd (2000) also extended this no-tion to neurological and some orthopedic patients wholike children and novices often have difficulty controllingthe intersegmental dynamics inherent in complex whole-body tasks Support for this view also comes from a studyby C H Shea Kohl and Indermill (1990) in which blockedpractice was more effective for the learning of a rapidforce production task when practice was limited to 50 tri-als However with increasing amounts of practice trials(200 400) random practice produced more effective re-tention performance than did blocked practice suggest-ing that a certain level of experience is required for ran-dom practice to become more effective than blockedpractice

This position can be reconciled with both the elabora-tion and the reconstruction accounts for the CI effect Forexample when complex tasks are first practiced short-term memory may be overloaded preventing or at leastlimiting processing (elaboration) of the various tasks Al-ternatively full reconstruction may be difficult from trialto trial In either case the development of memory repre-sentations or motor programs for the different tasks wouldbe degraded since random practice adds to the alreadyhigh processing loads The intratask interference inherentin complex tasks may be sufficient for effective learningunder blocked conditions at least in the early phases ofpractice Once the movements become more automatedmdashand therefore less difficult for the performermdashintroducingfurther challenges by varying the practice context mightprovide the necessary stimuli for continued learning

FeedbackOne of the most important variables in the motor-learning

process is the feedback provided to the learner attemptingto acquire a new motor skill Consequently this variablehas received a great amount of interest in the research lit-erature In this section we will review findings related tothe frequency of feedback and the organization of feed-back about different task components

Feedback frequency A number of studies mainlyusing relatively simple tasks have demonstrated that re-ducing the proportion of trials for which augmented feed-back (knowledge of results knowledge of performance) isprovided can result in more effective learning than doesgiving feedback after every single trial (eg Nicholson ampSchmidt 1991 Weeks amp Kordus 1998 Winstein ampSchmidt 1990 Wulf Lee amp Schmidt 1994 Wulf ampSchmidt 1989 Wulf Schmidt amp Deubel 1993 WulfShea amp Rice 1996 however see Lai amp Shea 1998) For

190 WULF AND SHEA

example using a lever-patterning task Winstein andSchmidt found that reducing the feedback frequency to50 of the trials was more beneficial for the learning ofthis task than giving 100 feedback Furthermore delay-ing the feedback for a few seconds has been found to pro-duce more effective learning than does giving feedbackimmediately after or even concurrently with the move-ment (eg Schmidt amp Wulf 1997 Swinnen SchmidtNicholson amp Shapiro 1990 Vander Linden Cauraugh ampGreene 1993) In particular concurrent feedback typi-cally has strong performance-enhancing effects duringpractice but results in clear performance decrements rel-ative to postresponse feedback when it is withdrawn inretention or transfer tests (Schmidt amp Wulf 1997 VanderLinden et al 1993 Winstein et al 1996)

Some of these findings appear to have parallels in theliterature on animal conditioning In this work partial re-inforcement has been shown to slow acquisition but alsoto maintain learning for much longer than continuous re-inforcement does The reason for this is seen mainly to liein the fact that partial reinforcement tends to simulate ex-tinction conditions (eg Tarpy 1982 chap 7) Howeveras Winstein and Schmidt (1990 see also Wulf amp Schmidt1989) have pointed out one set of findings that arguesagainst such an explanation for human learning is thatbeneficial learning effects of reduced feedback are seennot only in no-feedback retention tests but also in 100feedback retention tests Such a finding would seem un-likely in animal learning Furthermore immediate rein-forcement is more effective than delayed reinforcement inanimal learning but this is not the case for skill learningin humans This suggests that although the principles un-derlying animal reinforcement and feedback in humanshave some similarities they show some important differ-ences as well

The degrading effects of frequent and immediate feed-back on delayed retention and transfer tests found inhuman learning are typically explained with the guidancehypothesis (Salmoni et al 1984) According to this hy-pothesis feedback has positive effects such as guiding thelearner to the correct response however frequent feedbackis also argued to have several side effects that degradelearning For example learners seem to become too depen-dent on the information provided by the augmented feed-back and neglect the processing of intrinsic (eg proprio-ceptive) feedback which they will have to rely on whenthe additional information is no longer available (ie inno-feedback retention or transfer) In addition to makinglearners dependent on it frequent feedback has beenshown to make movement production quite variable pre-sumably preventing the learner from developing a stablemovement representation (Lai amp Shea 1998 Wulf ampSchmidt 1994 see also Schmidt 1991a) Finally feed-back has been argued to work proactively by facilitatingnext-response planning and retrieval In this sense fre-quent feedback might provide too much facilitation in theplanning of the subsequent response thereby reducing the

participantrsquos need to perform memory retrieval operationsthought to be critical for learning (Wulf amp Schmidt 1994)

Although the learning of simple skills seems to benefitfrom reducing or delaying the augmented feedback (al-though this effect appears to be stronger for variable prac-tice conditions see Lai amp Shea 1998) there is evidenceto suggest that more frequent feedback might be requiredfor the learning of complex skills Some support for thisnotion comes from studies on summary feedback (or sum-mary knowledge of results) where feedback about eachtrial is given only after a certain number of trials have beencompleted Whereas in experiments using relatively sim-ple tasks the largest summary feedback length proved tobe the most effective for learning (eg a summary of 16 tri-als in Gable Shea amp Wright 1991 20 trials in Lavery1962 15 trials in Schmidt Young Swinnen amp Shapiro1989) Schmidt Lange and Young (1990) showed that theoptimal number of summary trials for a more complexsimulated batting task was lower than those found formore simple tasks In this case a summary of 5 trials wasmore beneficial than longer (15) or shorter (1) feedbacksummaries Also Yao Fischman and Wang (1994) foundthat both summary feedback and average feedback (wherean average error or performance score is provided after aset of trials) about 5-trial blocks was more advantageousfor learning than was feedback about 15-trial blocks orevery-trial feedback Finally Guadagnoli Dornier andTandy (1996) directly demonstrated that task complexityas well as task-related experience interacted with the op-timal number of trials summarized That is whereas rela-tively long feedback summaries benefited the learning ofa relatively simple striking task for both novice and expe-rienced participants as well as the learning of a more com-plex double-striking task for experienced participants single-trial feedback was more effective than longer feed-back summaries for novices trying to learn a complextask

Presumably the most complex task that has been usedto examine the effects of a reduced feedback frequency isthe ski simulator task (Wulf Shea amp Matschiner 1998Experiment 2) The ski simulator consists of two bowedrails and a platform on wheels that is attached to the endsof the apparatus by elastic rubber belts (see Figure 1) Ifthe platform is displaced the rubber belts pull the plat-form back to its center position The platform can be madeto move sideways on the rails by exerting force on it Thegoal of the performer standing on the platform is to makeoscillatory slalom-type movements with the goal beingto produce the largest possible amplitudes (and sometimesfrequency) Similar to many sport skills this task requiresextensive practice and learners usually continue to showimprovements in performance across several days (egden Brinker amp van Hekken 1982 Durand et al 1994Vereijken 1991 Wulf Shea amp Matschiner 1998)

Wulf Shea and Matschiner (1998 Experiment 2) pro-vided learners on the ski simulator with feedback about aperformance measure the so-called relative force onset

COMPLEX SKILL LEARNING 191

Figure 1 Platform position and force curves of the left and right feet for 1 participant onDays 1 (top) 3 (middle) and 6 (bottom) in the Wulf Shea and Matschiner (1998) studyForce onsets are indicated on the force curves as can be seen the force onsets occurred laterand later in the movement cycle (eg between two successive reversals of the platform) acrosspractice

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

190 WULF AND SHEA

example using a lever-patterning task Winstein andSchmidt found that reducing the feedback frequency to50 of the trials was more beneficial for the learning ofthis task than giving 100 feedback Furthermore delay-ing the feedback for a few seconds has been found to pro-duce more effective learning than does giving feedbackimmediately after or even concurrently with the move-ment (eg Schmidt amp Wulf 1997 Swinnen SchmidtNicholson amp Shapiro 1990 Vander Linden Cauraugh ampGreene 1993) In particular concurrent feedback typi-cally has strong performance-enhancing effects duringpractice but results in clear performance decrements rel-ative to postresponse feedback when it is withdrawn inretention or transfer tests (Schmidt amp Wulf 1997 VanderLinden et al 1993 Winstein et al 1996)

Some of these findings appear to have parallels in theliterature on animal conditioning In this work partial re-inforcement has been shown to slow acquisition but alsoto maintain learning for much longer than continuous re-inforcement does The reason for this is seen mainly to liein the fact that partial reinforcement tends to simulate ex-tinction conditions (eg Tarpy 1982 chap 7) Howeveras Winstein and Schmidt (1990 see also Wulf amp Schmidt1989) have pointed out one set of findings that arguesagainst such an explanation for human learning is thatbeneficial learning effects of reduced feedback are seennot only in no-feedback retention tests but also in 100feedback retention tests Such a finding would seem un-likely in animal learning Furthermore immediate rein-forcement is more effective than delayed reinforcement inanimal learning but this is not the case for skill learningin humans This suggests that although the principles un-derlying animal reinforcement and feedback in humanshave some similarities they show some important differ-ences as well

The degrading effects of frequent and immediate feed-back on delayed retention and transfer tests found inhuman learning are typically explained with the guidancehypothesis (Salmoni et al 1984) According to this hy-pothesis feedback has positive effects such as guiding thelearner to the correct response however frequent feedbackis also argued to have several side effects that degradelearning For example learners seem to become too depen-dent on the information provided by the augmented feed-back and neglect the processing of intrinsic (eg proprio-ceptive) feedback which they will have to rely on whenthe additional information is no longer available (ie inno-feedback retention or transfer) In addition to makinglearners dependent on it frequent feedback has beenshown to make movement production quite variable pre-sumably preventing the learner from developing a stablemovement representation (Lai amp Shea 1998 Wulf ampSchmidt 1994 see also Schmidt 1991a) Finally feed-back has been argued to work proactively by facilitatingnext-response planning and retrieval In this sense fre-quent feedback might provide too much facilitation in theplanning of the subsequent response thereby reducing the

participantrsquos need to perform memory retrieval operationsthought to be critical for learning (Wulf amp Schmidt 1994)

Although the learning of simple skills seems to benefitfrom reducing or delaying the augmented feedback (al-though this effect appears to be stronger for variable prac-tice conditions see Lai amp Shea 1998) there is evidenceto suggest that more frequent feedback might be requiredfor the learning of complex skills Some support for thisnotion comes from studies on summary feedback (or sum-mary knowledge of results) where feedback about eachtrial is given only after a certain number of trials have beencompleted Whereas in experiments using relatively sim-ple tasks the largest summary feedback length proved tobe the most effective for learning (eg a summary of 16 tri-als in Gable Shea amp Wright 1991 20 trials in Lavery1962 15 trials in Schmidt Young Swinnen amp Shapiro1989) Schmidt Lange and Young (1990) showed that theoptimal number of summary trials for a more complexsimulated batting task was lower than those found formore simple tasks In this case a summary of 5 trials wasmore beneficial than longer (15) or shorter (1) feedbacksummaries Also Yao Fischman and Wang (1994) foundthat both summary feedback and average feedback (wherean average error or performance score is provided after aset of trials) about 5-trial blocks was more advantageousfor learning than was feedback about 15-trial blocks orevery-trial feedback Finally Guadagnoli Dornier andTandy (1996) directly demonstrated that task complexityas well as task-related experience interacted with the op-timal number of trials summarized That is whereas rela-tively long feedback summaries benefited the learning ofa relatively simple striking task for both novice and expe-rienced participants as well as the learning of a more com-plex double-striking task for experienced participants single-trial feedback was more effective than longer feed-back summaries for novices trying to learn a complextask

Presumably the most complex task that has been usedto examine the effects of a reduced feedback frequency isthe ski simulator task (Wulf Shea amp Matschiner 1998Experiment 2) The ski simulator consists of two bowedrails and a platform on wheels that is attached to the endsof the apparatus by elastic rubber belts (see Figure 1) Ifthe platform is displaced the rubber belts pull the plat-form back to its center position The platform can be madeto move sideways on the rails by exerting force on it Thegoal of the performer standing on the platform is to makeoscillatory slalom-type movements with the goal beingto produce the largest possible amplitudes (and sometimesfrequency) Similar to many sport skills this task requiresextensive practice and learners usually continue to showimprovements in performance across several days (egden Brinker amp van Hekken 1982 Durand et al 1994Vereijken 1991 Wulf Shea amp Matschiner 1998)

Wulf Shea and Matschiner (1998 Experiment 2) pro-vided learners on the ski simulator with feedback about aperformance measure the so-called relative force onset

COMPLEX SKILL LEARNING 191

Figure 1 Platform position and force curves of the left and right feet for 1 participant onDays 1 (top) 3 (middle) and 6 (bottom) in the Wulf Shea and Matschiner (1998) studyForce onsets are indicated on the force curves as can be seen the force onsets occurred laterand later in the movement cycle (eg between two successive reversals of the platform) acrosspractice

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 191

Figure 1 Platform position and force curves of the left and right feet for 1 participant onDays 1 (top) 3 (middle) and 6 (bottom) in the Wulf Shea and Matschiner (1998) studyForce onsets are indicated on the force curves as can be seen the force onsets occurred laterand later in the movement cycle (eg between two successive reversals of the platform) acrosspractice

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

192 WULF AND SHEA

that was found to correlate with performance in terms ofmovement amplitude in their Experiment 1 That is expertson this task not only produced larger amplitudes but alsoshowed later relative force onsets than did beginners Therelative force onset is viewed as a measure of movementefficiency Basically the force onset indicates when theperformer shifts his or her weight from one foot to theother More specifically it is the point in time at which theperformer shifts his or her weight to the left foot whilemoving to the right and vice versa Optimally when mov-ing say to the right side the body weight should only beon the right (ldquoouterrdquo) leg until the reversal point is reachedBeginners however tend to shift their weight relativelyearlymdashfor example when the platform passes the centerof the apparatus With an increased amount of practicethe force onset occurs later and later in the movementcycle and for some performers eventually coincides withthe platform reversal (The relative force onset is the forceonset in relation to the overall duration of the movementcycle)

In the Wulf Shea and Matschiner (1998 Experiment 2)study beginners were provided with feedback about forceonset and were instructed to try to delay the force onset untilthey reached the reversal point The feedback was pre-sented on an oscilloscope concurrently with and for thewhole duration of a 90-sec trial Whereas one group ofparticipants received feedback on all practice trials thatwere performed on 2 consecutive days (100 feedback)for another group the feedback was faded with the aver-age feedback frequency being 50 In addition there wasa control group without feedback In delayed retention allthe groups showed very similar performances in terms ofrelative force onset on the first of 10 no-feedback reten-tion trials However the group that had received the mostfeedback (100) during practice showed a clear perfor-mance improvement across trials whereas the 50 feed-back group showed no such performance gains and thecontrol group even demonstrated a performance decre-ment By the end of the retention test the participants inthe 100 group produced even later force onsets than theydid toward the end of the practice phase with feedbackpresent Furthermore only the 100 group was signifi-cantly more effective than the control group whereas the50 group did not differ significantly from either group

Thus there was no indication that the 100 feedbackparticipants developed a dependency on the feedback thatreduced the learning effectiveness of this condition Ratherit seems that the participants in the 100 feedback groupdeveloped a more effective error-detection-and-correctionmechanism (eg Salmoni et al 1984 Schmidt 1991a) ascompared with the 50 feedback and no-feedback (con-trol) conditions which enabled them to demonstrate fur-ther performance improvements even in the absence offeedback Thus contrary to studies that used more simpletasks in which reducing the relative feedback frequencywas more beneficial for learning than was providing learn-ers with feedback after or during every practice trial (egSchmidt amp Wulf 1997 Winstein amp Schmidt 1990 Wulfamp Schmidt 1989 Wulf et al 1993) the learning of this

more complex task was enhanced by 100 feedback In-terestingly this advantage occurred even though the feed-back given in the Wulf Shea and Matschiner (1998) studywas continuous and concurrent At least for the learningof simple skills concurrent feedback seems to produce aneven stronger dependency on it than postresponse feed-back after each trial does (Park Shea amp Wright 2000Schmidt amp Wulf 1997 Vander Linden et al 1993 Win-stein et al 1996) Since these dependency effects shouldbe attenuated by a reduced feedback frequency one mighthave expected a learning advantage for the 50 feedbackcondition This was not the case for this more complex taskhowever

In a recent study that required the learning of a complexbimanual coordination task Swinnen Lee VerschuerenSerrien and Bogaerds (1997) also found that continuousaugmented visual feedback enhanced performance andlearning Participants learning to produce cyclical armflexion and extension movements with a phase offset of90ordm demonstrated more effective performance in acquisi-tion and transfer if they received augmented visual infor-mation during practice as compared with reduced feed-back and normal vision practice conditions That isproviding learners with frequent augmented feedback notonly facilitated acquisition performance but also en-hanced learning as measured by various transfer testsmdashirrespective of the feedback conditions experienced inthese tests This study provides another example of thebeneficial effects of frequent feedback for complex motorskill learning

Thus it seems that in contrast to simple skill learningthe learning of more complex motor skills benefits fromrelatively high feedback frequencies (at least until a cer-tain level of experience is reached) This interpretation isin line with summary knowledge of results studies show-ing that the optimal number of trials summarized de-creases with more complex tasks (eg Schmidt et al1990 Schmidt et al 1989 Yao et al 1994) and is espe-cially in line with the findings and reasoning of Guada-gnoli et al (1996) who demonstrated an interaction be-tween task complexity or task-related experience andoptimal summary-KR length Another question related tothe effectiveness of feedback which seems to be of par-ticular importance for the learning of complex skills ishow feedback about different components of the taskshould be organized This issue will be addressed in thefollowing section

Feedback organizationComplex motor skills usuallyrequire the spatial and temporal coordination of varioussubmovements This raises the question of how feedbackshould be organized in order to enhance learning Shouldone concentrate on one aspect at a time or should themovement feature that feedback is given about be changedmore frequently Perhaps owing to the predominant use ofrelatively simple skills which often have only one overallgoal (eg a temporal or spatial target) and which thereforedo not lend themselves to examinations of feedback orga-nization effects relatively little research has been done toexamine this question Lee et al (Lee amp Carnahan 1990

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 193

Swanson amp Lee 1992) however used a task that requiredparticipants to knock down a series of barriers with cer-tain goal MTs being prescribed for the different segmentsAfter each trial feedback was given about one of themovement segments However the order of segments thatreceived feedback was different for different groupsWhereas for one group of participants this order was ran-domized another group received feedback about the dif-ferent segments in blocks of trials In (immediate) no-feedback retention tests the random feedback organizationturned out to be more effective than the blocked organiza-tion Lee and Carnahan argued that a blocked feedbackschedule might result in less effective learning than a ran-dom schedule because it focuses the learnersrsquo attention onone part of the task and causes them to neglect the otherparts or its integration into the whole action

Wulf et al (Wulf amp Buumlhner 1996 Wulf Houmlrger ampShea 1999) examined the generalizability of these find-ings to the learning of the more complex ski simulatortask Similar to the study by Wulf Shea and Matschiner(1998 Experiment 2) feedback was provided about theforces exerted by each foot on the platform and learnerswere instructed to try to delay the force onset of the innerfoot until the reversal of the platform In the Wulf Shea andMatschiner experiment feedback was given for one footat a time but the foot that the feedback referred to wasswitched on consecutive trials that is feedback was pro-vided in a serial order In a first study to examine whetherconstantly changing the component receiving feedback isindeed optimal for the learning of this more complex taskor whether a blocked schedule would be more effective inthis case Wulf and Buumlhner provided learners with 2 daysof practice on the ski simulator Feedback about the forcesproduced by each foot was presented in either a serial or ablocked order Whereas the foot receiving feedback wasswitched on each practice trial for the serial feedbackgroup under blocked feedback conditions feedback wasgiven for one foot per day After 2 days of practice theparticipants performed a no-feedback retention test onDay 3 In this retention test the blocked feedback groupdemonstrated relative force onsets that were significantlylater than those of the serial feedback group That is theblocked feedback order was more effective for the learn-ing of a delayed force onset than was serial feedback

These findings provided some preliminary evidencethat contrary to what has been found for the learning of arelatively simple skill (Lee amp Carnahan 1990 Swanson ampLee 1992) a frequent change in the aspect of the task thatreceives feedback might not be advantageous for the learn-ing of more complex skills It is conceivable that serial orrandom feedback is too demanding if the task to be learnedis relatively difficult leading to the learning decrementsseen in that study However it is also possible that withmore practicemdashand decreasing attentional demandsmdashserial (or random) feedback might eventually be benefi-cial for complex skill learning as well (see C H Shea et al1990) In a follow-up experiment Wulf Houmlrger and Shea(1999) therefore provided learners with twice as many

practice trials on the ski simulator as Wulf and Buumlhner(1996) had usedmdashthat is 4 days of practice with 10 trialsper day One question in this study was whether the effec-tiveness of serial versus blocked feedback would interactwith the amount of practice so that blocked feedbackwould be more beneficial early in practice but less bene-ficial later in practice as compared with serial feedbackFor this reason no-feedback retention trials were per-formed at the beginning of each practice day In additionthere was a retention test on Day 5 However even withrelatively extensive practice on this task no advantageswere found for serial feedback (see Figure 2) In fact theblocked feedback group (which was given feedback aboutone foot on Days 1 and 3 and about the other foot on Days2 and 4) produced consistently larger movement ampli-tudes than did the serial feedback group throughout the prac-tice phase as well as in all the retention tests In this exper-iment the beneficial effects of blocked practice were seenin amplitude whereas there were no significant group dif-ferences in force onset (even though feedback was givenabout force onset as in Wulf amp Buumlhner 1996) presum-ably because the instructions put more emphasis on theoverall goal of the taskmdashthat is the production of largeamplitudes At any rate the results replicated the findingsof Wulf and Buumlhner in demonstrating that blocked feed-back was clearly more effective for the learning of thiscomplex task

Thus even with considerable amounts of practice theserial feedback conditionmdashin which the participantrsquos at-tention was switched on every trial between focusing ondelaying the force onset of the left foot while moving tothe right and focusing on delaying the force onset of theright foot while moving to the leftmdashmight have been toodemanding and perhaps initially confusing and there-fore resulted in degraded learning relative to blockedfeedback Informal interviews at the end of the experimentconfirmed that the serial feedback participants found thistype of feedback to be very attention demanding The ad-ditional attention demands directed at translating the forceonset display information into action seemed to detractfrom the production of large movement amplitudeswhich was the ultimate goal of the task Under blockedfeedback conditions on the other hand the learners couldconcentrate on one sidefoot on each day making practiceless demanding and actually benefiting the learning of thiscomplex skill

Summary Overall the studies in which the effects offeedback frequency (Wulf Shea amp Matschiner 1998)and feedback organization (Wulf amp Buumlhner 1996 WulfHoumlrger amp Shea 1999) on the learning of a complex tasksuch as the ski simulator task were examined demonstratethat the findings derived from simple-task learning stud-ies do not generalize very well to complex skill learningIn contrast to the simple laboratory tasks used in manystudies reducing relative feedback frequency did not en-hance learning of the ski simulator task In fact providingconcurrent feedback on 100 of the practice trials turnedout to be most effective for learning This suggests that

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

194 WULF AND SHEA

more feedback might be required to optimize complex skilllearning Furthermore contrary to simple skill learningconcentrating the feedback on one task component at atime (blocked feedback) was found to be more beneficialfor the learning of the ski simulator task than were fre-quent changes in the component about which feedbackwas provided (serial feedback) Together these resultssuggest that whereas the learning of simple tasks might beenhanced by making practice more ldquodifficultrdquo or challeng-ing for the learner (eg by reducing feedback frequencyor providing serialrandom feedback) the learning ofcomplex skills might not benefit and might even be de-graded by increasing the demands imposed on the learnerThis is in line with the notion that in contrast to simpleskills the learning of complex skillsmdashwith inherentlyhigh attentional memory or control demandsmdashcan be fa-cilitated by providing the learner with relatively frequentfeedback It should also be noted however that feedbackin complex tasks is generally not as prescriptive as it oftenis in many of the simple tasks that have been used in thiscontext In complex tasks there are often different com-ponents that have to be coordinated to produce skilled per-formance making it much more diff icult for a singlefeedback measure to be truly prescriptive In attempts toimprove performance the learner has to rely on sources ofintrinsic feedback Thus the likelihood of the learnerrsquos be-coming dependent on extrinsic feedback and neglectingthe processing of intrinsic feedback is reduced as com-pared with the learning of simple skills That is the neg-ative effects thought to be associated with the guidanceprovided by frequent feedback seem to be reduced

Physical AssistanceA form of guidance that seems to be even stronger than

the guidance provided by feedback is physical guidanceIn the laboratory this is realized for example by moving

the performerrsquos limb to a target position (eg Holding ampMacrae 1964 Kelso 1977) or by using a mechanical stopto indicate the target position (eg Hagman 1983) Hag-man and Winstein et al (1994) examined the effects ofphysical guidance on the learning of positioning move-ments In both studies reducing the proportion of presen-tation trials in which performers were presented with themechanical stop produced more effective learning as as-sessed by retention or transfer tests without the stop thandid a relatively high proportion of presentation trials Sim-ilar to the effects of high feedback frequencies these re-sults have also been viewed as support for the guidancenotion according to which the learner becomes dependenton the guidance and therefore demonstrates less effectivelearning when the guidance is removed In fact Winsteinet al (1994) showed that physical guidance had effects onlearning similar to those of feedback about the movementoutcome

The use of physical guidance procedures can also beseen in many sport situationsmdashfor example when a gym-nast learning a new stunt is spotted by a coach or when achild learning to swim is provided with a flotation deviceThese procedures differ from those used in the above-mentioned studies however in that they are less prescrip-tive Although they provide support for the learner andthereby facilitate the achievement of the movement goalthe learner has more freedom for exploratory activitiesTherefore physical assistance might be a more appropri-ate term for such cases Even though guidance techniquessuch as these had for a long time been regarded as effec-tive means in teaching motor skills (eg Holding 1969Holding amp Macrae 1964 1966) the newer findings re-garding the effects of guidance on motor learning castdoubts on their effectiveness for learning (eg Salmoni et al 1984) As Schmidt (1988 1991b Schmidt amp Wris-berg 2000) points out the benefits of (physical) guidance

Figure 2 Movement amplitudes of the blocked and serial feedback (FB) groups on thepretest (Trial 1) during practice (Trials 2 9 11 18 20 27 29 36) and on the no-feedbackretention tests of Day 2 (Trial 10) Day 3 (Trial 19) Day 4 (Trial 28) and Day 5 (Trials 37ndash41)in the Wulf Houmlrger and Shea (1999) study Blocked feedback about force onset facilitatedthe production of large movement amplitudes throughout practice and on all retention tests

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 195

seem to be mainly temporary in nature Guidance oftenhas strong performance-enhancing effects during practicewhen it is present yet when it is withdrawn in retention ortransfer tests performance is often less effective than thatof learners who practiced the task without or with lessguidance Schmidt (1991b see also Schmidt amp Wrisberg2000) therefore recommended that physical guidance beused only sparely in the teaching of sport skills in orderto avoid the detrimental effects on learning that seem to beassociated with too much guidance

Very few studies have examined the effectiveness ofphysical guidance or assistance on the learning of morecomplex motor skills however In a series of experimentsWulf and colleagues (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) studied the effects of physical as-sistance devices on the learning of balancing tasks Inthese studies participants learning to maintain their bal-ance on the stabilometer (Wulf amp Shea 1997 1998 WulfShea amp Whitacre 1998) or to produce oscillatory move-ments on the ski simulator (Wulf Shea amp Whitacre1998) were or were not provided with a pair of (ski) polesThe use of poles which are placed on the floor in front ofthe apparatus facilitates performance by providing alarger base of support Thus the poles serve a functionsimilar to that of for example training wheels on chil-drenrsquos bicycles or the support provided by a parent in ice-skating or by a physical therapist to a patient who is learn-ing to walk again That is the physical assistance helps toguide the performer to the intended goal Even though theguidance provided by a mechanical stop in a positioningtask might provide more prescriptive information than theassistance provided by the ski poles (or similar devices inmore real-world settings) the ski poles also greatly facil-itate the achievement of the task goalmdashthat is the pro-duction of large movement amplitudes The physical as-sistance provided by the poles is therefore comparable toassisted devices that are often used in real-world settingswhere complex skills are being taught

On the basis of previous findings (Hagman 1983 Win-stein et al 1994) and reasoning (eg Schmidt 1991a1991b) one would expect learning advantages for unas-sisted practice conditions (without poles) particularly ascompared with conditions with 100 physical assistanceduring practice since these participants have more op-portunity to practice the relevant aspects of the task (egmaintaining balance or applying force to the ski simulatorplatform at the appropriate time) Even though physicalassistance should enhance performance during practiceparticipants practicing with poles might bypass some ofthe processing that would be required for effective perfor-mance when the poles are removed and as a result showless effective learning (Schmidt 1991a 1991b) Also froma specificity point of view (eg Henry 1968) accordingto which learning is specific to the conditions encounteredduring practice learners practicing without poles have aclear advantage over those practicing with poles sincetheir practice conditions are identical to the conditions en-countered in retention On the other hand it is also con-

ceivable that for more complex tasks in which the mem-ory and information-processing demands are high thephysical assistance provided by the poles could reduce thedemands to a more manageable level thereby helping learn-ers to get a feel for the goal movement or helping themfigure out how to produce an effective coordination pat-tern This may not be possible early in practice where theimmediate goal is simply to try not to fall off the device

Wulf Shea and Whitacre (1998 Experiment 1) exam-ined the effects of physical assistance devices on learningto perform slalom-type movements on the ski simulatorWhereas one group of learners practiced the task with skipoles another group practiced without poles Both groupsperformed immediate retention tests without poles at theend of each of the 2 days of practice as well as a delayedretention test on Day 3 As one would expect the use ofpoles clearly facilitated performance That is the polegroup produced larger amplitudes than did the group thatpracticed without poles throughout the acquisition phaseIn addition the pole group demonstrated later relativeforce onsets (see Figure 3) indicating that the poles alsofacilitated the production of a more efficient movementpattern (Wulf Shea amp Matschiner 1998) More impor-tant though when the poles were removed in retentionthe group that had practiced with the poles showed no per-formance decrements in amplitude or force onset Bothgroups had very similar amplitudes in the delayed reten-tion test and even more interesting the pole group demon-strated superior learning with regard to relative forceonset That is the participants who had practiced with thepoles showed force onsets under no-pole conditions thatwere similar to those produced during practice with thepoles Thus the benefits in force onset provided by thepoles during practice were not only temporary effectsRather these benefits transferred to a situation in whichno poles could be used and they were relatively perma-nent in nature that is the poles enhanced the learning ofa more efficient movement pattern relative to practicewithout poles

These results are not in line with guidance (eg Schmidt1991a 1991b Schmidt amp Wrisberg 2000) or specificity-of-learning notions (eg Henry 1968) according towhich unguided practice should have been more effectivefor learning than was practice with physical guidance orassistance On the ski simulator the poles might have en-abled learners to experience (consciously or unconsciously)that it is more effective to shift the weight from the outerto the inner foot relatively latemdashthat is to delay the forceonset of the inner foot This experience provided by thepoles served to shorten the learning process and enabledthe learners to reach a given level of performance in lesstime These results again suggest that there may be limita-tions to the generalizability of findings from studies usingtypical laboratory tasks with few degrees of freedom Forcomplex movements with many degrees of freedom suchas those encountered in many sports situations theremight actually be advantages to using physical assistanceprocedures corroborating the view that complex skill

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

196 WULF AND SHEA

learning benefits from reducing the demands imposed onthe learner

However there might be other factors (besides the com-plexity of the task) that determine to what extent physicalguidance or assistance is beneficial or detrimental to learn-ing For example using the stabilometer taskmdashthat is adynamic balance task that requires performers to maintaintheir balance on a wooden board pivoting over a fulcrumto the left and rightmdashWulf and Shea (1997 1998) foundthat providing learners with poles was not advantageousfor learning Independent of whether the task was to keepthe platform in a horizontal position (Wulf amp Shea 1997)or to continuously move the platform so that its positioncoincided with a template presented on a computer screen(Wulf amp Shea 1998) practice with poles did not enhancelearning Even though the poles again facilitated perfor-mance during practice participants who had practicedwithout the poles were more effective in delayed no-poleretention tests It is conceivable that physical assistance isconducive to learning to the extent that it allows thelearner to explore the ldquoperceptualndashmotor workspacerdquo(Newell 1991) On the ski simulator task the poles en-abled learners to produce a movement patternmdashthat islarge movement amplitudesmdashthat otherwise they wouldnot have experienced until much later in practice or per-haps not at all In addition through the use of the poleslearners presumably experienced that a late force onset ismore effective That is in this case the poles might havefacilitated the learnerrsquos search for the optimal solution ofthe motor problem On the stabilometer task on the otherhand the poles might have prevented the performer fromexploring the perceptualndashmotor workspace For examplebecause of the additional support provided by the poles

learners were able to keep the stabilometer platform ldquoontargetrdquo (eg in the horizontal) for most of the time Thisway however they did not experience larger deviationsfrom the goal and consequently did not learn how to re-duce the deviations without the help of the poles More re-search is needed however to determine under what con-ditions the learning of complex motor skills is degraded orenhanced by physical guidance

VARIABLES INFLUENCING MAINLY COMPLEX SKILL LEARNING

As the previous sections have shown learning princi-ples derived from the study of simple skills are not neces-sarily generalizable to the learning of more complex skillsTherefore in order to understand the processes underlyingthe learning of complex motor skills and to be able to giverecommendations for the teaching of these skills it seemsimportant to examine directly the learning of more com-plex skills In addition there might be other advantages tousing more complex skills in motor-learning research Forexample some principles or phenomena that could be rel-evant for the learning of complex skills might be less pow-erful or might not show up in tasks such as those typicallyused in the laboratory One characteristic of complexmotor skills that is difficult to simulate with simple tasksis that the performer can direct his or her attention to var-ious aspects of the task This poses a challenge not only forthe practitioner who wants to optimize training by direct-ing the performerrsquos attention through instructions or feed-back to the relevant aspects of the task but also for the re-searcher who wants to understand the role and function ofattentional processes in skill acquisition Another variable

Figure 3 Relative force onsets of the pole and no-pole groups during practice (Tri-als 1ndash 6 and 8ndash13) and in the immediate (Trials 7 and 14) and delayed retention tests(Trials 15ndash 21) without poles in Experiment 1 of Wulf Shea and Whitacre (1998) Theuse of the poles facilitated the production of late force onsets not only during practicebut also in delayed retention

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 197

that for various reasons seems to have great potential forenhancing complex skill learning is observational prac-tice Whereas having learners observe another performerhas yielded mixed results for the learning of simple skillsit seems to have clear benefits for more complex skills Inthe subsequent sections we will review the effectivenessof these factors in more detail

Attentional FocusA number of studies have demonstrated that the per-

formerrsquos focus of attention can have a decisive influenceon performance and learning Whereas most of these stud-ies manipulated the learnerrsquos attention via instructionsmore recent work has also examined the effectiveness ofdifferent attentional foci induced by the feedback given tothe learner

Instructions In a series of studies Wulf and her col-leagues (Wulf Houmlszlig amp Prinz 1998 Wulf Lauterbach ampToole 1999 Wulf Shea amp Park 2001) manipulated thelearnersrsquo focus of attention by giving them different in-structions These studies were stimulated by a previous studywhich showed that giving learners instructions that couldbe assumed to enhance the learning process is not neces-sarily beneficial for the performance and learning of acomplex skill (Wulf amp Weigelt 1997) In the Wulf andWeigelt experiment performers attempting to producelarge-amplitude movements on the ski simulator eitherwere instructed to delay the forcing of the platform untilafter it had passed the center of the apparatusmdasha perfor-mance characteristic that had been found to be associatedwith expert performance (eg Vereijken 1991)mdashor weregiven no further instructions Wulf and Weigelt found thatthe instructions did not enhance the learning of this taskrather they were detrimental to performance during prac-tice and transfer to a ldquostressrdquo situation relative to no in-structions These findings might appear surprising at firstYet they seem to be in line with the results of subsequentstudies (Shea amp Wulf 1999 Wulf Houmlszlig amp Prinz 1998Wulf Lauterbach amp Toole 1999 Wulf Shea amp Park2001) which showed that instructions that direct the per-formersrsquo attention to their body movementsmdashsimilar tothe instructions given in Wulf and Weigeltrsquos (1997) studymdashare not very effective

Specifically in these studies Wulf and colleagues com-pared the effects of instructions that direct the learnerrsquos at-tention to the body movements that are required to pro-duce the goal action (internal focus of attention) with theeffects of those that direct his or her attention to the ex-ternal effects that these movements have on the environ-ment (external focus of attention) For example using theski simulator task Wulf Houmlszlig and Prinz (1998 Experi-ment 1) instructed one group of learners to focus on theirfeet and to try to exert force with the right foot when theplatform moved to the right (ie to delay the force onsetof the left foot until the reversal point of the platform) andvice versa (internal focus group) Another group receivedbasically the same instructions the only difference wasthat these participants were instructed to focus on the

wheels that were located directly under their feet That isthe participants were instructed to exert force on the rightpair of wheels when the platform moved to the right andvice versa (external focus group) The results indicated thatthe latter condition was much more effective for the pro-duction of large movement amplitudes than was the inter-nal focus condition in both practice and retention whichwas not different from a control condition without addi-tional instructions (see Figure 4) Thus the instructionsthat directed the learnersrsquo attention to the effects of theirmovement were more beneficial for learning than the in-structions focusing the learnersrsquo attention on their ownmovements Wulf Houmlszlig and Prinz replicated these find-ings in a second experiment In that experiment the learn-ing of the stabilometer task was also enhanced by an ex-ternal focus of attention (markers attached to the board infront of each foot) relative to an internal attentional focus(feet)

In a subsequent study Wulf Lauterbach and Toole(1999) examined the generalizability of the learning ad-vantages of an external focus of attention to the acquisi-tion of sports skills While practicing pitch shots in golfthe participantsrsquo attention was directed either to the swingof their arms (internal focus) or to the motion of the clubhead (external focus) Here again the external-focus in-structions greatly enhanced the accuracy of the shots inpractice and in delayed retention as compared with the internal-focus instructions Recently external-focus ben-efits have also been found for tennis (Maddox Wulf ampWright 2000) The findings demonstrate that this effectseems to be generalizable to the acquisition of other real-world skills as well

The results of a study by Riley Stoffregen Grocki andTurvey (1999) are nicely in line with these findings Rileyet al measured postural sway when participants standingupright with their eyes closed touched a curtain verylightly with their fingertips (A curtain was used becauseit would not provide any mechanical support for posture)Interestingly touching the curtain significantly reducedpostural fluctuation as compared with not touching it butonly when the participants were asked to minimize move-ments of the curtain resulting from their touch (touch-relevant condition) When the participants were told thattouching the curtain was irrelevant for the experiment(touch-irrelevant condition) postural sway was basicallythe same as under no-touch conditions These findings cor-roborate the view that distracting performers from theirown movements by having them focus on an externalmovement effect can greatly enhance performance

Similar observations were made by Singer Lidor andCauraugh (1993) who had participants practice a ball-throwing task with different attentional strategies Theyfound that a nonawarenessstrategy and the so-called five-step approach in which performers were instructed tofocus on a situational cue (eg the center of the target) re-sulted in more effective practice performance and (imme-diate) dual-task transfer than did an awareness strategy inwhich learners were to focus on their own movements The

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

198 WULF AND SHEA

awareness group was no more effective than a controlgroup without strategy-related instructions Even thoughthe instructions to focus on a situational cue did not nec-essarily direct the performersrsquo attention to the effects oftheir movements these instructions might at least haveserved to direct their attention away from their own move-ments Also Baumeister (1984) demonstrated that self-focused attention on motor performancemdashfor examplefrom competition a cash incentive or audience-inducedpressuremdashcan disrupt performance (see also Baumeisteramp Steinhilber 1984) In addition there is plenty of anec-dotal evidence for the detrimental effects of paying atten-tion to the coordination of onersquos movements at least whenit comes to the performance of well-practiced skills (egGallwey 1982 Schmidt 1988 Schneider amp Fisk 1983)

One question related to the use of attentional strategiesis whether there might be individual differences in thepreference and perhaps in the effectiveness of the atten-tional focus or whether the advantage of concentrating onthe effects of onersquos movements rather than on the move-ments themselves is a general phenomenon To examinethis question Wulf Shea and Park (2001) gave partici-pants the option to adopt either an internal or an externalfocus of attention In their study participants learning tobalance on the stabilometer were asked to find out for them-selves which type of attentional focus would work betterfor them In a retention test the participants were asked toadopt either an internal or an external attentional focus

depending on which one they had found to be more ad-vantageous during the first 2 days of practice Of interestwas first the number of learners who would find an in-ternal or external focus to be more effective and secondthe performances of these subgroups (if any) especiallyon the retention test The results provided further supportfor the advantages of an external focus of attention A con-siderably greater number of participants chose to focus onthe markers that were located in front of their feet (n = 16)rather than on the feet themselves (n = 4) Moreover theparticipants who chose to focus on the markers were clearlymore effective in retention than those who focused on theirfeet Thus these results provide support for the view thatthe beneficial effect of an external relative to an internalattentional focus is a general and robust phenomenon

Feedback The advantages of focusing on the outcomeof onersquos movements as compared with focusing on themovements themselves not only might be relevant for theformulation of instructions but also could have implica-tions for the feedback that is given to the learner C H Sheaand Wulf (1999) therefore examined whether feedbackwould also be more effective if it directs the performerrsquosattention away from his or her own movements and to theeffects of these movementsmdashthat is if it induces an ex-ternal focus of attention C H Shea and Wulf used the sta-bilometer task and presented two groups of participantswith the same concurrent visual feedback which essen-tially consisted of the platform movementsrsquo being dis-

Figure 4 Average movement amplitudes of the internal-focus external-focus andcontrol groups during practice (Days 1 and 2) and retention (Day 3) on the ski simu-lator in Experiment 1 of Wulf Houmlszlig and Prinz (1998) An external focus of attentionenhanced the learning of large-amplitude movements relative to an internal focus andto no instructions regarding the attentional focus

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 199

played on a computer screen However whereas one groupof learners was informed that the feedback representedtheir own movements (internal focus) the other group wastold that the feedback represented lines that were markedon the platform in front of each of the performerrsquos foot(external focus) C H Shea and Wulf argued that if thelearning advantages of an external focus of attention aregeneralizable to the feedback that is given to the learnersimilar benefits should be found for the external-focusrelative to the internal-focus feedback condition Also todetermine whether augmented internal-focus or external-focus feedback would result in additional benefits rela-tive to only giving performers internal-focus and external-focus instructions (as Wulf Houmlszlig amp Prinz 1998 had donein their Experiment 2) two groups with either internal- orexternal-focus instructions were also included

The results demonstrated that learning was more effec-tive not only when performers were given external-focusinstructions relative to internal-focus instructions butalso when they were provided ldquoexternalrdquo relative to ldquointer-nalrdquo feedback (see Figure 5) That is even though the feed-back display was identical for the two feedback groups thefeedback group that had adopted an external focus of at-tention had lower errors than the feedback group with aninternal attentional focus This suggests that the feedbackgiven to performers during practice can be more effectiveif it directs their attention to the movement effects ratherthan to the movements themselves

Interestingly the feedback provided to learners in theC H Shea and Wulf (1999) study generally enhanced per-formersrsquo ability to maintain their balance on the stabilome-ter relative to no feedback even though the feedbackcould have been argued to be redundant with respect totheir intrinsic feedback (Magill Chamberlin amp Hall1991) That is the learners had visual and kinestheticfeedback available to inform them about the platformrsquos de-viation from the horizontal Yet the visual display of theplatform movements on the screen considerably benefitedtheir performance as compared with the no-feedbackconditions One possible reason for this added benefit ofthe feedback is that it might have incremented the degreeto which the learners were able to maintain an externalfocus of attention independent of the (internal- or external-focus) instructions given to the learners It is also inter-esting to note that the withdrawal of the feedback in re-tention had no detrimental effect on performance This isin contrast to other studies where the withdrawal of con-current feedback resulted in clear performance decre-ments (eg Schmidt amp Wulf 1997 Vander Linden et al1993 Winstein et al 1996) Both the fact that the feed-back provided in this experiment considerably enhancedperformance during practice even though it was redun-dant with respect to the performerrsquos intrinsic feedbackand the fact that there was no performance decrement forthe feedback groups when the augmented feedback waswithdrawn indicate that the function of augmented feed-back was not only informational or motivational in nature

(see Adams 1987) Rather these findings seem to sug-gest that feedback can have the capacity to induce an ex-ternal focus of attention independent of attentional focusinstructions that benefits performance and learning

In contrast to the experimental technique used byC H Shea and Wulf (1999) where the feedback given todifferent groups of participants was identical in practicalsettings coaches or instructors typically provide the learnerwith verbal feedback that refers to that aspect of perfor-mance that needs the most improvement That is on thebasis of what the coach considers to be the critical mistakeor flaw he or she gives feedback that will hopefully helpthe performer to make appropriate changes on subsequentattempts The goal of a study by Wulf McConnel Gaumlrtnerand Schwarz (in press) was therefore to examine the gen-eralizability of the external-focus feedback benefits to thelearning of sport skills under conditions that approximatethose of athletic training situations In their Experiment 1different feedback statements were selected that are oftenused in volleyball training and that refer to the performerrsquosbody movements (internal-focus feedback) These state-ments were ldquotranslatedrdquo into statements that basicallycontained the same information but directed the learnersrsquoattention more to the movement effects For example in-stead of instructing learners to shift their weight from theback leg to the front leg while hitting the ball (internalfocus) they were instructed to shift their weight towardthe target (external focus) After every fifth practice trialthe performer was provided with the feedback statementthat was deemed most appropriate on the basis of his orher performance on the previous trials

The results showed that the accuracy of the serves wasgreatly enhanced by the external-focus relative to the internal-focus feedback not only during practice but alsoafter a 1-week retention interval in a retention test withoutfeedback That is the feedback that avoided direct refer-ences to the performerrsquos body movements led to a greateraccuracy in hitting a target This was true not only fornovices but also for advanced player who already had ex-perience with the ldquotennisrdquo serve Furthermore this ad-vantage in the movement outcome was not accomplishedat the expense of movement form As was determinedthrough expert ratings both types of feedback led to sim-ilar improvements in form That is although the move-ment form in the novice group was degraded by internal-focus feedback during practice this group caught up withthe external-focus group in the retention test where thefeedback was withdrawn Apparently these participantsldquorecoveredrdquo from the detrimental effects of the feedbackthat directed their attention to their movement coordina-tion in the no-feedback retention test For the experts external-focus feedback tended to result in a better move-ment form than did internal-focus feedback during bothpractice and retention

The results of recent study by Todorov Shadmehr andBizzi (1997) also suggest that feedback about the move-ment effect appears to be more beneficial than feedback

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

200 WULF AND SHEA

related to the movements that produced it Todorov et alargued that the highest level of motor planning and con-trol seems to be in terms of the kinematics of the end-effector and that therefore the feedback given to thelearner should be most effective if it represents the move-ments of the end-effector rather than the body move-ments Even though they did not compare these two typesof feedback Todorov et al showed that the learning oftable tennis shots was enhanced by providing performerswith concurrent feedback about the trajectory of theirpaddle (in relation to the paddle trajectory of an expert)Those participants receiving this type of feedback weremore accurate in hitting the target than were the partici-pants who were provided verbal feedback (on gross er-rors) and who hit 50 more balls It is conceivable that atleast part of the reason for the effectiveness of feedbackabout the paddle motion was that it induced an externalfocus of attention whereas the control participants (with-out feedback) paid more attention to their own movementpattern

Theoretical considerations The findings reviewedabove suggest that giving learners instructions or feed-back that direct their attention to the coordination of theirbody movements (ie induce an internal focus of atten-

tion) which are often used in teaching motor skills mightnot be optimal for learning Rather they demonstrate thatmotor skill learning can be enhanced by focusing thelearnersrsquo attention on the effects of their movements (ieby inducing an external attentional focus)

The exact reasons for the beneficial effects of an exter-nal relative to an internal focus of attention are still rela-tively unclear However more than 100 years ago WilliamJames (1890) already had suggested that actions are con-trolled more effectively if attention is directed to the (in-tended) outcome of the action or its ldquoremote effectsrdquorather than to the ldquoclose effectsrdquo that are directly associ-ated with it such as the kinesthetic feedback Similar ideashave been proposed by Lotze (1852) According to Lotzemovements are represented by codes of their perceived ef-fects and the desired outcome has the power to guide theaction so that this outcome is achieved More recentlyPrinz and colleagues (Prinz 1992 1997 Prinz Aschers-leben Hommel amp Vogt 1995) and Hommel (1997) havetaken up Jamesrsquos and Lotzersquos ideas that actions are plannedand controlled by their intended effects (action effect hy-pothesis Prinz 1997) In his common coding theoryPrinz (1990 1997) provides a possible explanation for theadvantages of focusing on the effects of onersquos movements

Figure 5 Root-mean square errors (RMSEs) of the no-feedbackexternal-focus no-feedbackinternal-focus feedbackexternal-focus and feedbackinternal-focus groups during practice (Days 1 and 2) and retention (Day 3) in the C HShea and Wulf (1999) study Both feedback and an external focus of attentionenhanced learning of the stabilometer task

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 201

rather than on the movements themselves Contrary to tra-ditional views which assume that there are different andincommensurate coding systems for afferent and efferentinformation (eg Massaro 1990 Sanders 1980 Welford1968) Prinz argued that there is a common representa-tional medium for perception and action According tothis view efferent and afferent codes can be generated andmaintained in a commensurate way only at a distant levelof representation That is action planning and perceptiontypically involves distal events since this is the only for-mat that allows for commensurate coding and thus for ef-ficient planning of action (see Prinz 1992) Therefore ac-tions should be more effective if they are planned in termsof their intended outcome rather than in terms of the spe-cific movement patterns

Some support for this notion comes from the findingsof Proctor and Dutta (1993) In their study participantswere provided with extensive practice on a two-choice re-action task The relations between stimulus locations re-sponse key location and effectors (right vs left finger)were different for different groups of participants Intransfer the participants were switched to new conditionsin which the spatial mapping between stimulus locationand response keys as well as the hand placement (uncrossedvs crossed) was either the same as or different from thatin practice The important finding in the present contextwas that switching hand placements did not lead to in-creases in RT when the spatial stimulusndashresponse map-pings remained constant However switching hand place-ments did cause performance decrements when the stimulusndashresponse mapping changed even if the map-pings of stimuli to fingers remained the same

The idea that people ldquonaturallyrdquo tend to focus on the ef-fects of their actions instead of focusing on the move-ments that are involved in the action has also been put for-ward by Vallacher and Wegner (1985 1987) In theiraction identification theory Vallacher and Wegner pro-posed a hierarchical order of action identities that are usedin the control of action They argued that if lower andhigher levels of action identity are available higher levelidentities become prepotent Lower levels in an actionrsquosidentity structure refer to the details of the action such asthe specific movements (eg striking the correct keys ona piano) whereas higher levels are more related to the ef-fects of the action (eg playing the melody) If the per-former selects too low a level of action controlmdashfor ex-ample because of pressure to perform wellmdashperformanceis often disrupted

It is possible that attempts to consciously control move-ments interferes with automatic motor control processeswhereas focusing on the movement effects allows themotor system to more naturally self-organize (eg Kelso1995) unconstrained by conscious control Focusing onthe remote effects of the movement might serve to let un-conscious control processes take over ldquofreeing uprdquo con-scious aspects of attention to be directed to other aspectsof the task resulting in more effective performance andlearning This might be particularly advantageous for com-

plex skills where consciously trying to control the manydegrees of freedom is almost bound to result in failureparticularly early in practice (see Willingham 1998 for adifferent perspective)

Support for this assumption has recently been found instudies by McNevin Shea and Wulf (in press) and WulfShea and Park (2001) In these studies the frequencycharacteristics (fast Fourier transformation FFT) of thebalance records of participants balancing on the sta-bilometer were determined Participants who focused ontheir feet (internal focus) showed higher amplitude andlower frequency movement adjustments than did partici-pants focusing on markers in front of their feet (externalfocus Wulf Shea amp Park 2001) Furthermore in theMcNevin et al study two groups that focused on markersat a greater distance from the feet made even more andsmaller corrections in maintaining their balance than dida group that focused on the markers close to the feet High-frequency responding has been argued to be a character-istic of a biological system with more active degrees of free-dom whereas constrained or compromised perceptualndashmotor systems exhibit lower frequency components (seeNewell amp Slifkin 1996 for a discussion of this issue)McNevin et al proposed a constrained action hypothesisto account for the external focus benefits According tothis hypothesis performers focusing on their body move-ment or on an effect that occurs in close proximity to theirbody tend to actively intervene in the control of theirmovements Thus one advantage of focusing attention ona (distant) movement effect seems to be that it allows un-conscious processes to control the movements required toachieve this effect resulting in fast movement adjustmentsand generally enhanced performance and learning

In summary it is important to note that to date researchon focus of attention has been directed only to relativelycomplex skills Presumably complex skills have beenused for these experiments because it was felt that an in-ternal focus of attention might be detrimental to learningbecause it increases the already high cognitive loads as-sociated with the task Alternatively an external focusmight actually reduce cognitive load by taking advantageof self-organizing capabilities of the motor system Re-cent results by Wulf McNevin and Shea (2001) supportthis view In their study participants balanced on the sta-bilometer and probe RTs were taken as a measure of theattention demands required under external versus internalattentional focus conditions External-focus participantsdemonstrated lower probe RTs than did internal-focusparticipants indicating reduced attention demands asso-ciated with an external focus Thus these results are inline with the view that complex skill learning can be en-hanced by reducing the attentional andor motor controldemands associated with it

Observational LearningAnother factor that might be more effective for the learn-

ing of complex skills relative to more simple skills is ob-servational learning The study of observational learning

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

202 WULF AND SHEA

has been the focus of considerable research since the early1960s In fact observational learning is one area of motorbehavior research that has typically utilized tasks that arerelatively complex although there are a number of exam-ples of simple tasks being used as well

For the most part observational learning although gen-erally considered not as effective as physical practice hasbeen demonstrated to be a viable method of practicingcomplex motor skills (eg Bachman 1961 Landers ampLanders 1973 Martens Burwitz amp Zuckerman 1976Schoumlnfelder-Zohdi 1992 C H Shea Wulf amp Whitacre1999 Sidaway amp Hand 1993 Whiting 1988 see Mc-Cullach et al 1989 for a review) On the other hand ob-servational practice in learning simple tasks has led toequivocal findings In a number of experiments involvingsimple motor tasks (eg Blandin Proteau amp Alain 1994Burwitz 1975 Lee amp White 1990) observation has beenshown to enhance learning However other experimentshave shown little or no benefit of observation (eg Brownamp Messersmith 1948 Burwitz 1975 Wright Li ampCoady 1994)

There are at least three possible reasons for why obser-vational practice seems to be more effective for complexskills than for simple skills First there may be funda-mentally more to ldquoseerdquo and therefore more to be extractedas the result of observation of relatively complex tasks ascompared with simple tasks This issue is illustrated byBurwitz (1975) who found observation beneficial whenhe used a Bachman ladder task a relatively complex taskin one experiment and no differences when a pursuit rotortask a simple task was used in another experiment Henoted that the participants reported in postexperiment in-terviews that they had found the movement pattern lead-ing to success in the task ldquovisiblerdquo only in the Bachmanladder task This led Burwitz to conclude that the benefitsof observation were mediated by task difficulty In a sim-ilar vein Gould (1980) suggested that observation wasmore effective when the informational load of the task washigh which is often the case for complex tasks than whenit was low which is more typical of simple tasks Thesefindings along with psychophysical evidence on the per-ception of biological motion (eg Johansson 1973)where observers were effective in reducing complex arraysinto relatively simple patterns of relative motion promptedScully and Newell (1985) to propose that observers are ef-fective in reducing perceptual complexity by extractingpatterns of coordination (relative motion) rather than in-formation that leads to the specific scaling of the actionpattern These theorists have argued that because ob-servers do not actually execute the movement and do notprocess direct sensory information observers are not aseffective as the model who is physically practicing a skillin learning to scale the movement sequence However ob-servers appear to be able to extract information necessaryto construct appropriate coordination patterns (Schoumlnfelder-Zohdi 1992 Scully amp Newell 1985 Whiting 1988) anddetermine appropriate implementation strategies (Kohl ampShea 1992) In addition observation may provide the

learner with a better ldquopicturerdquo of how the various sub-components of a complex task fit together to form thewhole task Just as analogies have been shown to reducememory demands by providing a framework in which toorganize memory (eg Anderson amp Fincham 1994 Feryamp VomHofe 2000 Zamani amp Richard 2000) observa-tion may facilitate the structuring of the memories sup-porting the movements thus effectively reducing the totalmemory demands When the task demands are relativelysimple this may not be important but when the demandsare high and subcomponents must be properly integratedthe influence could be much larger Thus even though ob-servational practice might not be as effective as physicalpractice in training specific task characteristics becausesome types of important processing are unique to physi-cal practice and cannot be developed effectively duringobservation observational practice is viewed as a viabletraining method for teaching general characteristics ofcomplex tasks

Second it is possible that observational practice offersthe observer the opportunity to engage in processing thatwould not or could not be effectively carried out early inpracticemdashparticularly when learning a complex skillmdashwhen most if not all of the cognitive resources of the per-former are required to physically perform the new task(Kohl amp Fisicaro 1996 C H Shea Wright Wulf ampWhitacre 2000) That is the observer may be able to ex-tract important information from observation concerningappropriate coordination patterns or subtle requirementsof the task andor evaluate effective or ineffective strate-gies that would be difficult if not impossible to do whileattempting a new task because of the high demands oncognitive resource From this perspective observationalpractice offers processing opportunities in complex tasksthat would not generally be afforded under physical prac-tice conditions until the skill had become well learned (au-tomated) and attention demands had been greatly reducedHowever later in practice inappropriate or even incorrectperformance elements may have become well learned andtherefore more dominant and diff icult to change (seeWightman amp Lintern 1985) Thus since complex skillsare generally more attention demanding the beneficial ef-fects of observational practice should be seen more clearlyfor complex tasks Simple tasks on the other hand forwhich cognitive demands tend to be much lower may notbenefit as much from the reduced processing loads char-acteristic of observational practice

If physical and observational practice offer unique op-portunities for processing that may independently incre-ment learning acquisition protocols that combine physi-cal and observational practice should be effective trainingprocedures Some support for this notion was found in ex-periments by Shebilske and colleagues (eg ShebilskeRegian Arthur amp Jordan 1992) and C H Shea WrightWulf and Whitacre (2000) Shebilske et al for exampleused a special form of dyadic training the so-called activeinterlocked modeling (AIM) protocol Their participantspracticed a video game (ldquoSpace Fortressrdquo) where one

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 203

partner controlled half of the complex task (eg the key-board) whereas the other partner controlled the other half(eg the joystick) While controlling either the keyboardor the joystick the participants were able to observe theother member of the dyad perform his or her respectivetask Presumably this was possible because the attentionaland cognitive demands required to control half of the totaltask were reduced to the extent that they could allocate re-sources to observation Hands-on practice of one half andobservational practice of the other half of the control pro-cedures were switched from trial to trial Thus as com-pared with an individual training group that controlled thewhole task on all the practice trials for the participants inthe AIM dyad the informational load was reduced Ontest games that required control of the whole task how-ever there were no differences between groupsmdashthat isboth groups showed the same amount of learning (also seeJordan 1997 for a meta-analysis) Thus facilitating prac-tice by reducing the task demands had no detrimental ef-fect on learning as compared with practicing the whole task

Many tasks do not lend themselves to dual control (iesegmentation) as does ldquoSpace Fortressrdquo however Anotherpotentially efficient and effective training protocol whichinvolves observation and can be easily applied to manytraining environments and tasks can be constructed byhaving participants alternate between physical and obser-vational practice Such a protocol takes advantage of theintervals between trialssessions While one learner is en-gaged in physical practice the other is afforded the oppor-tunity to observe The participants change roles on subse-quent trials Using such a protocol C H Shea et al (2000)even found learning advantages of combined observa-tional and physical practice over physical practice aloneC H Shea et al (2000) had participants learn to alter-nately press two response keys with the first finger of theirright and left hands in an attempt to keep a ldquodotrdquo centeredon a target line (similar to the Pew 1966 task) The ldquodotrdquoaccelerated to the left when the right key was depressedand to the right when the left key was depressed A uniquefeature of these experiments was the inclusion of a trans-fer test The authors argued that the transfer test empha-sized the relative characteristics of the coordination pat-tern and appropriate movement strategies but deemphasizedthe specific characteristics of the acquisition task Thus ifobservers are effective in extracting from observation therelative characteristics of the task (eg Scully amp Newell1985) and determining effective strategies (Kohl amp Shea1992) as has been proposed the transfer task should besensitive to these aspects of the task In contrast perfor-mance on a retention test which has been utilized almostexclusively in studies of observation could be dominatedby specific rather than general characteristics of the task

When 50 of the physical practice trials were replacedwith observational practice no decrements in retentionperformance were noted as compared with a group thatpracticed physically on all practice trials (C H Shea et al2000 Experiment 2 also see Weeks amp Anderson 2000)More important the performance of the combined obser-

vational and physical practice group was superior to thatof the physical practice group on the transfer test Thuseven though the participants in the combined group per-formed only half the physical practice trials the physicalpractice participants did they performed equally well onthe retention test and more effectively than the physicalpractice group on the transfer test These findings demon-strate that alternating observational practice with physicalpractice trials can be quite effective presumably becausethe interspersed observational trials give learners the op-portunity to perform information-processing activitiesthat they would not be able to do while performing a com-plex highly attention-demanding skill

A third reason why observational practice might be par-ticularly useful for complex skill learning refers to trainingefficiency Owing to the relatively high physical demandspracticing complex motor tasks often requires the inser-tion of rest periods between trials Providing learners withobservational practice during rest intervalsmdashfor exampleby having them watch a partner performmdashwould enhancetraining efficiency since two (or more) learners could betrained in the same amount of time as one Learning effi-ciency is assessed when one considers the time money po-tential for injury energy andor other personal and exper-imental resources that are expended in order to conduct thetraining sessions Clearly from a personal standpoint ob-servational practice requires the expenditure of less phys-ical energy reduces the risk of injury and is not dependenton equipment or specific space requirements as in thecase of physical practice The issue of learning efficiencyalthough relevant to the study of both simple and complexskill learning is especially important for complex skillsbecause these skills generally take longer to learn oftenare more physically demanding in many cases introducethe potential for more risk and generally require morespecialized equipment or facilities than do simple skills

In the experiments by Shebilske et al (1992) and C HShea et al (2000) doubling the number of participantstrained without an increase in time and other resources ledto increases in training efficiency of at least 100 whilenot sacrificing and even enhancing learning effective-ness C H Shea et al (1999) examined the effectivenessof this type of combined physical and observational prac-tice protocol for the learning of a dynamic balance task(stabilometer) which like many other complex continu-ous motor tasks requires intervals between practice trialsto avoid fatigue and provide relief from the high attentionconcentration demands C H Shea et al (1999) utilized acombined practice protocol that had participants work indyads so that they alternated between physical and obser-vational practice They also allowed the participants to en-gage in undirected dialogue during the rest interval be-tween practice trials The effectiveness of this trainingprotocol was compared with that of individual trainingwith the same number of physical practice trials

The results showed that practice with a partner wasmore effective than individual practice Even though therewas an initial performance decrement under dyad practice

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

204 WULF AND SHEA

conditions in acquisition relative to individual practicethe dyad group participants quickly caught up with the in-dividual group participants and even tended to be more ef-fective at the end of the acquisition phase (see Figure 6)More important learners who had practiced with a part-ner showed more effective delayed retention performancethan did individual learners This is particularly remark-able since retention trials were performed under individual-performance conditions That is the benefits of dyadtraining transferred to a situation in which participantshad to perform individually

In summary it appears that observational practice ismore uniformly effective in complex than in simple tasksin which there is more for the observer to ldquoseerdquo and there-fore extract from observation (eg Burwitz 1975 Gould1980) In addition and perhaps most important observa-tional practice offers a form of practice in which the cog-nitive demands are sufficiently reduced because the ob-server is not required to physically perform the task andcan concentrate on the fundamental elements of the taskand the relationships between components and can deviseor evaluate strategies that result in effective task perfor-mance The latter types of processing afforded under ob-servation practice should be especially helpful when ob-servation and physical practice are combined It is alsoimportant to note that combined observation and physicalpractice protocols have the potential to greatly increaselearning efficiency without sacrificing learning effective-ness (C H Shea et al 2000 C H Shea et al 1999) This

latter consideration is especially important for complexskills that often require that more time and other resourcesbe devoted to training than for simple skills that can be ef-fectively learned with relatively little practice

SUMMARY FUTURE DIRECTIONS AND CONCLUSIONS

For a long time the use of relatively simple tasks hasbeen predominant in motor-learning research In state-ments that may have had a profound influence on manyresearchers in the motor domain Adams (1971) addressedthis issue by stating that

the villain that has robbed ldquoskillrdquo of its precision is appliedresearch that investigates an activity to solve a particularproblem like kicking a football flying an airplane or op-erating a lathe This approach is backwards for scien-tific productivity because it results in disconnected pocketsof data that lack the unifying ideas that are general scien-tific principles The task-centered approach is justifiedwhen practical reasons require us to know about task andefficiency in them but it is a limited way of achieving thelarger scientific goals of laws and theory (pp 112ndash113)

These statements by Adams were clearly aimed at thedifferences between applied and theoretical researchHowever researchers may interpret this as an indictmentof complex skills by association In contrast Hoffman(1990) and Locke (1990) have argued that researchershave ldquopreferred to look for answers in the brightness ofthe laboratory even though the most valuable findingsmay lie in the darkness of real-world settingsrdquo (Hoffman1990 p 150) In essence Hoffman and Locke argued thatthe study of simple skills in sterile laboratory settings hasnot been successful in determining underlying processesand principles or in providing meaningful information topractitioners To some degree we agree with their state-ments because there has been a rather strong tendency notto explore the boundary conditions of learning effectsparticularly in terms of the complexity of the task and thelevel of skill of the learners

Presumably simple skills are used so extensively inmotor learning and control because researchers and theo-rists believe that by reducing a task down to its funda-mental elements the processes (or mechanism) can be ex-posed more completely and manipulated more effectivelyto answer theoretical questions These notions appear tohave been borrowed in motor learning and control as wellas in other fields of study in the social sciences from thebiological and physical sciences For example in exercisephysiology today it is en vogue to study movement pro-duction systems not only at the mechanism level but alsoat the molecular level Although the reductionist approachhas been productive (although this has been debated) inother disciplines this pivotal question should be ques-tioned more openly by motor-learning and control re-searchers (see Locke 1990) The most obvious reason forutilizing simple skills although not the most lofty in-volves the efficiency with which experiments can be con-

Figure 6 Root-mean square errors (RMSEs) of the individualand dyad groups during acquisition and retention in the C HShea Wulf and Whitacre (1999) study Individual practice con-sisted of physical practice with rest periods whereas dyad prac-tice involved alternating physical practice observation and dis-cussion Dyad practice produced more effective learning than didindividual practice

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 205

ducted In many motor-learning and control experimentsincluding many of our own the tasks were (1) easily andinexpensively constructed (2) novel to participants (sothat previous practice would not confound the results)and (3) characterized by a single performance measureand of course (4) practice could be completed (often toapparent performance asymptote) in a single session Innumerous experiments as few as 54 trials were utilizedperhaps involving as little as 15 min to learn the taskClearly being able to complete experiments in 1 or 2 daysis efficient provided efficiency has not been traded off interms of the power of the experiment to tell us somethingabout skill learning in general Thus we agree with ourpedagogy colleagues Hoffman (1990) and Locke that thestudy of simple tasks has not been effective in determin-ing meaningful principles that have application to morecomplex skills

The importance of using ecologically valid tasks ratherthan artificial laboratory tasks has also been discussedextensively in the area of memory research (eg Neisseramp Winograd 1988) Whereas some argue that the princi-ples derived from laboratory research do not apply to theway memory functions in real-world situations (egNeisser 1991) others point out the value of basic researchconducted in controlled laboratory settings for theory de-velopment and its high generalizability to everyday prob-lems (eg Banaji amp Crowder 1989) This debate has beenconcerned mainly with the ecological validity of the re-search methods used Although we see a need to use morecomplex skills in motor learning research we do not ad-vocate the use of more naturalistic methods and settings(Neisser 1991) In fact we agree with Banaji and Crow-der that ldquothe more complex a phenomenon the greater theneed to study it under controlled conditions and the lessit ought to be studied in its natural complexityrdquo (p 1192)We would also like to point out that we do not want to dis-credit the value of fundamental research using simpleskills for discovering principles of learning This type ofresearch has contributed immensely to our understandingof the learning process Furthermore the validity of thisresearch does not depend on the immediate applicabilityof the identified principles to real-world problems We dobelieve however that the inclusion of more complex andecologically valid tasks in our research would further en-hance our understanding of learning

A Case Against the Utilization of Primarily Simple Tasks in the Study of Motor Skills

As we have shown in the first section of this reviewprinciples developed on the basis of simple skills are notalways generalizable to more complex skills Some fac-tors that enhance the learning of simple skills do not ap-pear to be beneficial for complex skill learning For ex-ample frequent feedback has been found to degrade thelearning of simple skills Yet frequent feedback can be ad-vantageous if new complex skills have to be learned (egGuadagnoli et al 1996 Wulf Shea amp Matschiner 1998)Similarly blocked feedback although detrimental for

simple skill learning relative to random feedback seemsto be beneficial for the learning of complex skills (egLee amp Carnahan 1990 Wulf Houmlrger amp Shea 1999) Fur-thermore random practice which has often been shown toenhance motor learning when simple tasks were useddoes not seem to be beneficial for the learning of morecomplex tasks (or for inexperienced learners) ratherblocked practice can result in more effective learning inthese cases (eg Albaret amp Thon 1999 Hebert et al1996 J B Shea amp Morgan 1979) Finally frequent phys-ical assistance can enhance the learning of complex motorskills at least if it facilitates the exploration of movementstrategies or techniques relative to practice without thehelp of physical guidance (Wulf Shea amp Whitacre1998) The obvious lack of generalizability from simple tocomplex tasks suggests that the utilization of complex tasksis essential in order to gain a more complete understand-ing of the processes underlying motor skill learning and inorder to be able to give valid recommendations for practi-cal settings Thus our recommendation is not to eliminateresearch conducted with simple tasks but to more broadlyinclude tasks at the other end of the difficulty spectrumOnly by including tasks of varying difficulty can the para-metrics of motor-learning principles be discerned

Moreover as was shown in the second section of our re-view some practice variables might be particularly effec-tive for the learning of complex skills relative to moresimple skills That is the importance of such variables asobservation of or interaction with a partner (eg C H Sheaet al 1999) or the attentional focus induced by the in-structions or feedback provided to the learner (eg C HShea amp Wulf 1999 Wulf Houmlb amp Prinz 1998) might notbe apparent when studying skills with few degrees of free-dom where the information-processing demands arecomparatively low Thus the relevance of such variablesfor motor learning might not be discovered by using sim-ple tasks

These findings call into question the adequacy of uti-lizing primarily simple tasks in the study of motor learn-ing and control (see Hoffman 1990 for an earlier discus-sion on this topic) The usefulness of simple laboratorytasks for understanding the processes underlying motorlearning seems to be limited and making generalizationsto the training of complex skills on the basis of such find-ings appears to be problematic As we have shown someof the theoretical concepts developed from the research onsimple skills are not applicable to complex skills It seemsthat with motor skill learning positive and negative ef-fects of practice variables change depending on the com-plexity of the task and the demands the task places on thelearner Therefore more research is needed that contrastssimple and complex skills and also takes into account theskill of the performer Until these boundary conditions aredetermined valid principles and thus comprehensivetheories cannot be developed

Therefore future research in motor learning should in-clude more than is common at present skills that requirethe control of several degrees of freedom load or overload

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

206 WULF AND SHEA

the perceptual cognitive andor attention systems havemultiple emphases that need to be integrated andor per-haps even involve whole-body movements or real-worldskills This not only may provide a more reliable basis forgiving recommendations for the teaching of motor skillsin applied settings and bring to the forefront new lines ofresearch (eg practice efficiency attentional focus) andprinciples that are not typically at issue with simple skillsbut also should further our understanding of and our insightsinto motor learning and control processes in general

Do Complex Skills Late in Practice Operate Similarly to Simple Skills Early in Practice

As we have shown several variables that are detrimen-tal to simple skill learning actually benefit the learning ofcomplex skills For example frequent feedback has beenfound to enhance the learning of the ski simulator task(Wulf Shea amp Matschiner 1998) as well as a complextwo-hand coordination task (Swinnen et al 1997) An im-portant question regarding the frequency of feedback iswhether the learning of complex skills can also be en-hanced by reducing feedback later in practice It is possi-ble that with considerably more practice (than the 2 daysused in the Wulf Shea amp Matschiner 1998 study) thelearning of a complex task like this is also affected nega-tively if feedback is provided too often Whether the ac-quisition of complex skills eventually follows the sameprinciples as the learning of simple tasks or whether thedanger of learningrsquos being degraded by too much feed-back is generally reduced or nonexistent for complexmotor skills needs to be determined in future studies

Similarly even though blocked feedback was beneficialfor the learning of the ski simulator task when provided on4 days of practice (Wulf Houmlrger amp Shea 1999) it is con-ceivable that more frequent changes in the task compo-nent that feedback is provided about would eventually alsobe advantageous for complex skills as has been shown forcomparatively simple tasks (Lee amp Carnahan 1990 Swan-son amp Lee 1992) With increasing amounts of practiceperformance becomes less attention demanding (eg Fittsamp Posner 1967 Shiffrin amp Schneider 1977 Weiss1939) providing the learner with more capacity to handlethe diversity of feedback information which might even-tually result in learning benefits Also in analogy to fad-ing procedures in feedback frequency manipulations itmight be predicted that a gradual shift from blocked torandom feedback schedules with the time frame of thisshift depending on the complexity of the task would op-timize skill learning Further research is also needed to ex-amine the effectiveness of various combinations of feed-back frequency and scheduling (blocked vs random)manipulations as well as effects of summary or averagefeedback on complex skill learning at various stages ofpractice

Similarly some of the inconsistencies that have beennoted in this review for CI manipulations in complexskills might be resolved if additional practice were to beprovided It is possible that with relatively little practice

(the exact amount may increase with increasing task dif-ficulty) blocked practice groups may outperform random-practice groups (see C H Shea et al 1990) With addi-tional practice the differences may diminish with randompractice eventually resulting in superior retention or trans-fer performance The absolute amount of practice thatmay be required for a particular effect to emerge may bequite small for very simple tasks and quite extensive formore complex tasks

Although some experiments have investigated how therelative difficulty of a task changes as a result of practiceexperience by either looking at the performance of learn-ers after little or more extensive practice (eg C H Sheaet al 1990) or contrasting inexperienced and experiencedperformers on a single task (eg Del Rey et al 1982Guadagnoli et al 1996) little if any research has lookedat this question systematically One experimental ap-proach might be to utilize a relatively complex skill thatcan be broken down into its fundamental components Theimportant question then would be the following Do theways in which the components have been learned reallyprovide insight into how the whole task is learned Tosome extent experiments looking at part practice in com-plex skills have addressed this issue from the other sideDoes practicing a simplified fractionated or subcompo-nent version of a more complex skill enhance the learningof the complex (whole) skill What we are suggesting is acloser look at the learning of the components looking atthe extent to which analyses of the learning of these taskcomponents provide insight into the learning of the wholetask

Another more longitudinal approach would be to deter-mine the influence of various practice manipulations earlyand late in practice in an attempt to determine which prac-tice manipulations enhance learning and when in prac-tice the manipulation is most robust Lai et al (2000) forexample have demonstrated that blocked practice early inpractice facilitates the learning of the relative timing pat-tern whereas random practice later in practice leads to en-hanced absolute timing performance Reversing the prac-tice schedule (random practice early and then blockedpractice later) resulted in poor relative and absolute tim-ing performance The important point is that factors thatexert influence in a learning situation may do so differen-tially across practice and task difficulty

The differential effectiveness of various practice vari-ables (depending on the task and experience of the learner)suggests that judging difficulty on the basis of externalcharacteristics may not be as productive as characterizingtask situations in terms of their demands (cognitive at-tentional motor) Under a scheme based on the demandsthe task places on the cognitive motor system difficultywould be considered a relative task characteristic that de-pends on the degree to which resources are loaded or over-loaded This perspective is consistent with the notion thatmotor skills with low demands benefit from practice con-ditions that increase the load and challenge the performer(eg Bjork 1994 Lee et al 1994 Schmidt amp Bjork 1992)

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

COMPLEX SKILL LEARNING 207

however the acquisition of skills that place extremely highloads on the performer should benefit from conditionsthat reduce the load to more manageable levels (physicalassistance increased feedback observation practice ex-ternal focus of attention)

Balancing Learning Effectiveness and LearningEfficiency in Complex Skills

Learning efficiency is a potentially important issue inthe study of complex motor skills which is clearly of lessinterest in the study of simple skills Especially interestingis the degree to which learning effectiveness and learningefficiency trade off under various experimental condi-tions Because of the potentially higher costs of complexskill training in general this is a critical question for ap-plied reasons but learning efficiency is also intriguing fortheoretical reasons Asking the questions ldquowhat factorsallow a complex skill to be learned to a specified level ofperformance in less time with less effort or at lesscostrdquo is potentially as interesting as asking the more gen-eral question ldquowhat factors influence the learning of motorskillsrdquo Yet the question of learning efficiency and its re-lationship to learning effectiveness has been largely ig-nored in motor behavior literature

One way to enhance training efficiency is to providelearners the opportunity to observe other learners per-forming C H Shea et al (2000 C H Shea et al 1999)recently demonstrated quite large increases in learning ef-ficiency for dyad training by allowing participants to al-ternate between physical and observational practice Thismanipulation also resulted in substantial increases inlearning effectiveness Practice methods developed in fu-ture research should be evaluated in terms of both learn-ing effectiveness and learning efficiency

ConclusionsAs we have shown in the first section of this review

principles developed through the study of contextual in-terference the manipulation of feedback and physicalguidance when simple skills are utilized are not general-izable to more complex skills In fact research on morecomplex skills shows that the manipulation of practicevariables that result in enhanced learning of simple skillsare actually detrimental to the learning of complex skillsIn the second section we reviewed two important practicevariables that have been studied primarily with complexskills and that may not be applicable to simple skillsThese findings call into question the adequacy of utilizingprimarily simple tasks in the study of motor learning andcontrol We conclude that more intensive research oncomplex skills is required to fully determine the general-izability of current findings to advance motor learningand control theory to include task difficulty as a mediat-ing factor in practice protocols and to provide salient ad-vice to practitioners Furthermore we question the rele-vance of research on simple skills to purposes other thanunderstanding the control and learning of simple skillssuch as deriving general motor learning principles We

propose that only when motor skills of varying complex-ity are included in the literature can fundamental theoret-ical principles be derived

REFERENCES

Adams J A (1971) A closed-loop theory of motor learning Journal ofMotor Behavior 3 111-149

Adams J A (1976) Issues for a closed-loop theory of motor learningIn G E Stemach (Ed) Information processing in motor control andlearning (pp 229-240) New York Academic Press

Adams J A (1987) Historical review and appraisal of research on thelearning retention and transfer of human motor skills PsychologicalBulletin 101 41-74

Albaret J-M amp Thon B (1999) Differential effects of task complex-ity on contextual interference in a drawing task Acta Psychologica100 9-24

Al-Mustafa A A (1989 June) Contextual interference Blockingwithin a random order Paper presented at the Annual Meeting of theNorth American Society for the Psychology of Sport and Physical Ac-tivity Asilomar CA

Anderson J R amp Fincham J M (1994) Acquisition of proceduralskills from examples Journal of Experimental Psychology LearningMemory amp Cognition 20 1322-1340

Bachman J C (1961) Specificity vs generality in learning and per-forming two large muscle motor tasks Research Quarterly 32 3-11

Banaji M R amp Crowder R G (1989) The bankruptcy of everydaymemory American Psychologist 44 1185-1193

Battig W F (1972) Intratask interference as a source of facilitation in transfer and retention In R F Thompson amp J F Voss (Eds) Top-ics in learning and performance (pp 131-159) New York AcademicPress

Baumeister R F (1984) Choking under pressure Self-consciousnessand paradoxical effects of incentives on skillful performance Journalof Personality amp Social Psychology 46 610-620

Baumeister R F amp Steinhilber A (1984) Paradoxical effects ofsupportive audiences on performance under pressure The home fielddisadvantage Journal of Personality amp Social Psychology 47 85-93

Bernstein N (1967) The co-ordination and regulation of movementsOxford Pergamon

Bjork R A (1994) Memory and metamemory considerations in thetraining of human beings In J Metcalfe amp A Shimamura (Eds)Metacognition Knowing about knowing (pp 185-207) CambridgeMA MIT Press

Blandin Y Proteau L amp Alain C (1994) On the cognitiveprocesses underlying contextual interference and observational learn-ing Journal of Motor Behavior 26 18-26

Bortoli L Robazza C Durigon V amp Carra C (1992) Effects ofcontextual interference on learning technical sports skills Perceptualamp Motor Skills 75 555-562

Brown H S amp Messersmith L (1948) An experiment in teachingtumbling with and without motion pictures Research Quarterly 18304-307

Burwitz L (1975) Observational learning and motor performancePaper presented at the FEPSAC Conference Edinburgh Scotland

Carnahan H Van Eerd D L amp Allard F (1990) A note on therelationship between task requirements and the contextual interfer-ence effect Journal of Motor Behavior 22 159-169

Carr J amp Shepherd R (2000) Movement science Foundations forphysical therapy in rehabilitation Gaithersburg MD Aspen

Del Rey P (1982) Effects of contextual interference on memory ofolder females differing in levels of physical activity Perceptual ampMotor Skills 55 171-180

Del Rey P Whitehurst M Wughalter E amp Barnwell B (1983)Contextual interference and experience in acquisition and transferPerceptual amp Motor Skills 57 241-242

Del Rey P Wughalter M E amp Carnes M (1987) Level of exper-tise interpolated activity and contextual interference effects on mem-ory and transfer Perceptual amp Motor Skills 64 275-284

Del Rey P Wughalter M E amp Whitehurst M (1982) The ef-

208 WULF AND SHEA

fects of contextual interference on females with varied experience inopen sport skills Research Quarterly for Exercise amp Sport 53 108-115

den Brinker B P L M amp van Hekken M F (1982) The analysisof slalom-type movements using a ski-simulator apparatus HumanMovement Science 1 91-108

Durand M Geoffroi V Varray A amp Preacutefaut C (1994) Studyof the energy correlates in the learning of a complex self-paced cycli-cal skill Human Movement Science 13 785-799

Farrow D amp Maschette W (1997) The effects of contextual inter-ference on children learning forehand tennis groundstrokes Journalof Human Movement Studies 33 47-67

Fery Y A amp VomHofe A (2000) When will the ball rebound Evi-dence for the usefulness of mental analogies in appraising the durationof motions British Journal of Psychology 91 259-273

Fitts P M (1954)The information capacity of the human motor sys-tem in controlling the amplitude of movement Journal of Experi-mental Psychology 47 381-391

Fitts P M amp Posner M I (1967) Human performance BelmontCA BrooksCole

French K E Rink J E amp Werner P F (1990) Effects of contex-tual interference on retention of three volleyball skills Perceptual ampMotor Skills 71 179-186

Gable C D Shea C H amp Wright D L (1991) Summary knowl-edge of results Research Quarterly for Exercise amp Sport 62 285-292

Gabriele T E Hall C R amp Buckolz E E (1987) Practice sched-ule effects on the acquisition and retention of a motor skill HumanMovement Science 6 1-16

Gabriele T E Hall C R amp Lee T D (1989) Cognition in motorlearning Imagery effects on contextual interference Human Move-ment Science 8 227-245

Gabriele T E Lee T D amp Hall C R (1991) Contextual interfer-ence in movement timing Specific effects in retention and transferJournal of Human Movement Studies 20 177-188

Gallwey W T (1982) The inner game of tennis New York BantamBooks

Goode S L (1986) The contextual interference effect in learning anopen motor skill Unpublished doctoral dissertation University ofLouisiana Baton Rouge

Goode S L amp Magill R A (1986) The contextual interference ef-fect in learning three badminton serves Research Quarterly for Exer-cise amp Sport 57 308-314

Gould D (1980) The influence of motor task types on model effec-tiveness Unpublished doctoral dissertation University of Illinois Ur-bana

Guadagnoli M A Dornier L A amp Tandy R D (1996) Optimallength for summary knowledge of results The influence of task-related experience and complexity Research Quarterly for Exercise ampSport 67 239-248

Hagman J D (1983) Presentation- and test-trial effects on acquisitionand retention of distance and location Journal of Experimental Psy-chology Learning Memory amp Cognition 9 334-345

Hall K G Domingues D A amp Cavazos R (1994) Contextual in-terference effects with skilled baseball players Perceptual amp MotorSkills 78 835-841

Hebert E P Landin D amp Solmon M A (1996) Practice scheduleeffects on the performance and learning of low- and high-skilled stu-dents An applied study Research Quarterly for Exercise amp Sport 6752-58

Henry F M (1968) Specificity vs generality in learning motor skillIn R C Brown Jr amp G S Kenyon (Eds) Classical studies in phys-ical activity (pp 328-331) Englewood Cliffs NJ Prentice-Hall

Henry F M amp Rogers D E (1964) Increased response latency forcomplicated movements and the ldquomemory drumrdquo theory of neuro-motor reaction Research Quarterly 31 448-458

Hoffman S J (1990) Relevance application and the development ofan unlikely theory Quest 42 143-160

Holding D H (1969) Learning without errors In L E Smith (Ed)Psychology of motor learning (pp 59-81) Chicago Athletic Institute

Holding D H amp Macrae A W (1964) Guidance restriction andknowledge of results Ergonomics 7 289-295

Holding D H amp Macrae A W (1966) Rate and force of guidancein perceptualndashmotor tasks with reversed or random spatial correspon-dence Ergonomics 9 289-296

Hommel B (1997) Toward an action-concept model of stimulusndashresponse compatibility In B Hommel amp W Prinz (Eds) Theoreticalissues in stimulusndashresponse compatibility (pp 281-320) AmsterdamElsevier

James W (1890) The principles of psychology (Vol 2) New YorkDover

Jeannerod M (1994) The timing of natural prehension Journal ofMotor Behavior 16 201-211

Jelsma O amp Pieters J M (1989) Practice schedule and cognitive styleinteraction in learning a maze task Applied Cognitive Psychology 373-83

Jelsma O amp Van Merrieumlnboer J J G (1989) Contextual interfer-ence Interactions with reflectionndashimpulsivity Perceptual amp MotorSkills 68 1055-1064

Johansson G (1973) Visual perception of biological motion and amodel for its analysis Perception amp Psychophysics 14 201-211

Jordan J A (1997) The effectiveness of individual and dyadic train-ing protocols for complex skill acquisition in space fortress A meta-analysis Unpublished doctoral dissertation Texas AampM UniversityCollege Station

Kahnemann D (1973) Attention and effort Englewood Cliffs NJPrentice-Hall

Kelso J A S (1977) Planning and efferent components in the codingof movement Journal of Motor Behavior 9 33-47

Kelso J A S (1995) Dynamic patterns The self-organization of brainand behavior Cambridge MA MIT Press

Klapp S (1995) Motor response programming during simple andchoice reaction time The role of practice Journal of ExperimentalPsychology Human Perception amp Performance 21 1015-1022

Kohl R M amp Fisicaro S A (1996) Response intention and imageryprocesses Locus interaction and contribution to motor learning Be-havioral amp Brain Sciences 19 760-762

Kohl R M amp Shea C H (1992) Pew (1966) revisited Acquisitionof hierarchical control as a function of observational practice Journalof Motor Behavior 24 247-260

Lai Q amp Shea C H (1998) Generalized motor program (GMP)learning Effects of reduced frequency of knowledge of results andpractice variability Journal of Motor Behavior 30 51-59

Lai Q amp Shea C H (1999) Bandwidth knowledge of results en-hances generalized motor program learning Research Quarterly forExercise amp Sport 70 79-83

Lai Q Shea C H Wulf G amp Wright D L (2000) Optimizinggeneralized motor programs and parameter learning Research Quar-terly for Exercise amp Sport 71 10-24

Landers D M amp Landers D M (1973) Teacher versus peer mod-els Effects of modelrsquos presence and performance level on motor be-havior Journal of Motor Behavior 5 129-139

Lavery J J (1962) Retention of simple motor skills as a function oftype of knowledge of results Canadian Journal of Psychology 16300-311

Lee T D amp Carnahan H (1990) When to provide knowledge of re-sults during motor learning Scheduling effects Human Performance3 87-105

Lee T D amp Genovese E D (1988) Distribution of practice in motorskill acquisition Learning and performance effects reconsidered Re-search Quarterly for Exercise amp Sport 59 277-287

Lee T D amp Genovese E D (1989) Distribution of practice in motorskill acquisition Different effects for discrete and continuous tasksResearch Quarterly for Exercise amp Sport 60 59-65

Lee T D amp Magill R A (1983) The locus of contextual interferencein motor-skill acquisition Journal of Experimental PsychologyLearning Memory amp Cognition 9 730-746

Lee T D amp Magill R A (1985) Can forgetting facilitate skill ac-quisition In D Goodman R B Wilberg amp I M Franks (Eds) Differ-ing perspectives on memory learning and control (pp 3-22) Amster-dam North-Holland

Lee T D Swinnen S P amp Serrien D J (1994) Cognitive effortand motor learning Quest 46 328-344

COMPLEX SKILL LEARNING 209

Lee T D amp White M A (1990) Influence of an unskilled modelrsquospractice schedule on observational motor learning Human MovementScience 9 349-367

Lee T D Wishart L R Cunningham S amp Carnahan H (1997)Modeled timing information during random practice eliminates the con-textual interference effect Research Quarterly for Exercise amp Sport68 100-105

Lee T D Wulf G amp Schmidt R A (1992) Contextual interferencein motor learning Dissociated effects due to the nature of task varia-tions Quarterly Journal of Experimental Psychology 44A 627-644

Limons E amp Shea J B (1988) Deficient processing in learning andperformance In A M Colley amp R J Beech (Eds) Cognition and ac-tion in skilled behavior (pp 333-347) Amsterdam North-Holland

Locke L F (1990) Why motor learning is ignored A case of ducksnaughty theories and unrequited love Quest 42 136-142

Lotze R H (1852) Medicinische Psychologie oder Physiologie derSeele (Medical psychology or physiology of the soul) Leipzig Weid-mannrsquosche Buchhandlung

Maddox D Wulf G amp Wright D L (2000) The effects of an in-ternal vs external focus of attention on the learning of a tennis back-hand Unpublished manuscript Texas AampM University

Magill R A Chamberlin C J amp Hall K G (1991) Verbalknowledge of results as redundant information for learning an antici-pation timing skill Human Movement Science 10 485-507

Magill R A amp Hall K G (1990) A review of the contextual inter-ference effect in motor skill acquisition Human Movement Science9 241-289

Martens R Burwitz L amp Zuckerman J (1976) Modeling effectson motor performance Research Quarterly 47 277-291

Massaro D W (1990) An information-processing analysis of percep-tion and action In O Neumann amp W Prinz (Eds) Relationships be-tween perception and action Current approaches (pp 133-166)Berlin Springer-Verlag

McCullagh P Weiss M R amp Ross D (1989) Modeling consider-ations in motor skill acquisition and performance An integrated ap-proach In K Pandolf (Ed) Exercise and sport science reviews(Vol 17 pp 475-513) Baltimore Williams amp Wadkins

McNevin N H Shea C H amp Wulf G (in press) Increasing thedistance of an external focus of attention enhances learning Psycho-logical Research

Neisser U (1991) A case of misplaced nostalgia American Psycholo-gist 46 34-36

Neisser U amp Winograd E (1988) Remembering reconsidered Eco-logical and traditional approaches to the study of memory New YorkCambridge University Press

Newell K M (1991) Motor skill acquisition Annual Review of Psy-chology 42 213-237

Newell K M amp Slifkin A B (1996) The nature of movement vari-ability In J Piek (Ed) Motor control and human skill A multidisci-plinary perspective (pp 143-159) Champaign IL Human Kinetics

Nicholson D E amp Schmidt R A (1991) Scheduling informationfeedback to enhance training effectiveness In Proceedings of theHuman Factors Society 35th Annual Meeting (pp 1400-1403) SantaMonica CA Human Factors Society

Park J-H Shea C H amp Wright D L (2000) Reduced-frequencyconcurrent and terminal feedback A test of the guidance hypothesisJournal of Motor Behavior 32 287-296

Pew R W (1966) Acquisition of hierarchical control over the temporalorganization of a skill Journal of Experimental Psychology 71 764-771

Pigott R E amp Shapiro D C (1984) Motor schema The structureof the variability session Research Quarterly for Exercise amp Sport55 41-45

Pinto-Zipp G amp Gentile A M (1995) Practice schedules in motorlearning Children vs adults Society of Neuroscience Abstracts 211620

Povel D-J amp Collard R (1982) Structural factors in patterned fin-ger tapping Acta Psychologica 52 107-123

Prinz W (1990) A common coding approach to perception and actionIn O Neumann amp W Prinz (Eds) Relationships between perceptionand action (pp 167-201) Berlin Springer-Verlag

Prinz W (1992) Why donrsquot we perceive our brain states EuropeanJournal of Cognitive Psychology 4 1-20

Prinz W (1997) Perception and action planning European Journal ofCognitive Psychology 9 129-154

Prinz W Aschersleben G Hommel B amp Vogt S (1995) Hand-lungen als Ereignisse [Actions as events] In D Doumlrner amp E van derMeer (Eds) Das Gedaumlchtnis ProblememdashTrendsmdashPerspektiven(pp 129-168) Goumlttingen Hogrefe

Proctor R W amp Dutta A (1993) Do the same stimulusndashresponserelations influence choice reactions initially and after practice Jour-nal of Experimental Psychology Learning Memory amp Cognition 19922-930

Riley M A Stoffregen T A Grocki M J amp Turvey M T(1999) Postural stabilization for the control of touching HumanMovement Science 18 795-817

Salmoni A W Schmidt R A amp Walter C B (1984) Knowledgeof results and motor learning A review and critical appraisal Psy-chological Bulletin 95 355-386

Sanders A F (1980) Stage analysis of reaction processes In G E Stel-mach amp J Requin (Eds) Tutorials in motor behavior (pp 331-354)Amsterdam North-Holland

Schmidt R A (1988) Motor control and learning A behavioral em-phasis (2nd ed) Champaign IL Human Kinetics

Schmidt R A (1991a) Frequent augmented feedback can degradelearning Evidence and interpretations In J Requin amp G E Stelmach(Eds) Tutorials in motor neuroscience (pp 59-75) Dordrecht TheNetherlands Kluwer

Schmidt R A (1991b) Motor learning and performance From prin-ciples to practice Champaign IL Human Kinetics

Schmidt R A amp Bjork R A (1992) New conceptualizations ofpractice Common principles in three paradigms suggest new princi-ples for training Psychological Science 3 207-217

Schmidt R A Lange C amp Young D E (1990) Optimizing sum-mary knowledge of results for skill learning Human Movement Sci-ence 9 325-348

Schmidt R A amp Wrisberg C A (2000) Motor learning and per-formance A problem-based learning approach (2nd ed) ChampaignIL Human Kinetics

Schmidt R A amp Wulf G (1997) Continuous concurrent feedbackdegrades skill learning Implications for training and simulationHuman Factors 39 509-525

Schmidt R A Young D E Swinnen S amp Shapiro D E (1989)Summary knowledge of results for skill acquisition Support for theguidance hypothesis Journal of Experimental Psychology LearningMemory amp Cognition 15 352-359

Schmidt R A Zelaznik H Hawkins B Frank J S amp Quinn J T(1979) Motor-output variability A theory for the accuracy of rapidmotor acts Psychological Review 47 415-451

Schneider W amp Fisk A D (1983) Attention theory and mechanismsfor skilled performance In R A Magill (Ed) Memory and control ofaction (pp 119-143) Amsterdam North-Holland

Schneider W amp Shiffrin R M (1977) Controlled and automatichuman information processing I Detection search and attentionPsychological Review 84 1-66

Schoumlnfelder-Zohdi B G (1992) Investigating the informational na-ture of a modelled visual demonstration Unpublished doctoral dis-sertation Louisiana State University Baton Rouge

Scully D M amp Newell K M (1985) Observational learning andthe acquisition of motor skills Toward a visual perception perspectiveJournal of Human Movement Studies 11 169-186

Sekiya H Magill R A amp Anderson D I (1996) The contextual in-terference effect in parameter modifications of the same generalizedmotor program Research Quarterly for Exercise amp Sport 67 59-68

Sekiya H Magill R A Sidaway B amp Anderson D I (1994)The contextual interference effect for skill variations from the sameand different generalized motor programs Research Quarterly for Ex-ercise amp Sport 65 330-338

Shapiro D C amp Schmidt R A (1982) The schema theory Recentevidence and developmental implications In J A S Kelso amp J E Clark(Eds) The development of movement control and co-ordination(pp 113-150) New York Wiley

210 WULF AND SHEA

Shea C H Kohl R amp Indermill C (1990) Contextual interfer-ence Contributions of practice Acta Psychologica 73 145-157

Shea C H Lai Q Wright D L Immink M amp Black C (2001)Consistent and variable practice conditions Effects on generalizedmotor program and parameter learning Journal of Motor Behavior33 139-152

Shea C H Wright D L Wulf G amp Whitacre C (2000) Phys-ical and observational practice afford unique learning opportunitiesJournal of Motor Behavior 32 27-36

Shea C H amp Wulf G (1999) Enhancing motor learning through external-focus instructions and feedback Human Movement Science18 553-571

Shea C H Wulf G Whitacre C A (1999) Enhancing training ef-ficiency and effectiveness through the use of dyad training Journal ofMotor Behavior 31 119-125

Shea J B amp Morgan R L (1979) Contextual interference effects onthe acquisition retention and transfer of a motor skill Journal of Ex-perimental Psychology Human Learning amp Memory 5 179-187

Shea J B amp Titzer R C (1993) The influence of reminder trials oncontextual interference effect Journal of Motor Behavior 25 264-274

Shea J B amp Zimny S T (1983) Context effects in learning movementinformation In R A Magill (Ed) Memory and the control of action(pp 345-366) Amsterdam North-Holland

Shebilske W L Regian J W Arthur W amp Jordan J A (1992)A dyadic protocol for training complex skills Human Factors 34369-374

Shiffrin R M amp Schneider W (1977) Controlled and automatichuman information processing II Perceptual learning automatic at-tending and a general theory Psychological Review 84 127-190

Sidaway B amp Hand M J (1993) Frequency of modeling effects onthe acquisition and retention of a motor skill Research Quarterly forExercise amp Sport 64 122-126

Singer R N Lidor R amp Cauraugh J H (1993) To be aware or notaware What to think about while learning and performing a motorskill Sport Psychologist 7 19-30

Smith P J K amp Davies M (1995) Applying contextual interferenceto the Pawlata roll Journal of Sport Sciences 13 455-462

Swanson L amp Lee T D (1992) The effects of aging and schedulingof knowledge of results on motor learning Journal of GerontologyPsychological Sciences 47 406-411

Swinnen S P Lee T D Verschueren S Serrien D J amp Bo-gaerds H (1997) Interlimb coordination Learning and transferunder different feedback conditions Human Movement Science 16749-785

Swinnen S P Schmidt R A Nicholson D E amp Shapiro D C(1990) Information feedback for skill acquisition Instantaneousknowledge of results degrades learning Journal of Experimental Psy-chology Learning Memory amp Cognition 16 706-716

Tarpy R M (1982) Principles of animal learning and motivationGlenview IL Scott Foresman

Todorov E Shadmehr R amp Bizzi E (1997) Augmented feedbackpresented in a virtual environment accelerates learning of a difficultmotor task Journal of Motor Behavior 29 147-158

Tsutsui S Lee T D amp Hodges N J (1998) Contextual interferencein learning new patterns of bimanual coordination Journal of MotorBehavior 30 151-157

Vallacher R R amp Wegner D M (1985) A theory of action identi-fication Hillsdale NJ Erlbaum

Vallacher R R amp Wegner D M (1987) What do people thinktheyrsquore doing Action identification and human behavior Psycholog-ical Review 94 3-15

Vander Linden D W Cauraugh J H amp Greene T A (1993) Theeffect of frequency of kinetic feedback on learning an isometric forceproduction task in nondisabled subjects Physical Therapy 73 79-87

Vereijken B (1991) The dynamics of skill acquisition Meppel KripsRepro

Vereijken B van Emmerik R E A Whiting H T A amp Newell K M (1992) Free(z)ing degrees of freedom in skill acqui-sition Journal of Motor Behavior 24 133-142

Weeks D L amp Anderson L P (2000) The interaction of observa-tional learning and overt practice Effects on motor skill learning ActaPsychologica 104 259-271

Weeks D L amp Kordus R N (1998) Relative frequency of knowl-edge of performance and motor skill learning Research Quarterly forExercise amp Sport 69 224-230

Weiss P (1939) Principles of development New York HoltWelford A T (1968) Fundamentals of skill London MethuenWhitacre C A amp Shea C H (2000) The performance and learning

of generalized motor programs Relative (GMP) and absolute (param-eter) errors Journal of Motor Behavior 32 163-175

Whiting H T A (1988) Imitation and the learning of complex cycli-cal actions In O G Meijer amp K Roth (Eds) Complex motor behav-ior The motor action controversy (pp 381-401) Amsterdam NorthHolland

Wightman D amp Lintern G (1985) Part-task training for tracking andmanual control Human Factors 27 267-283

Willingham D B (1998) A neuropsychological theory of motor skilllearning Psychological Review 105 558-584

Winstein C J Pohl P S Cardinale C Green A Scholtz L ampWaters C S (1996) Learning a partial-weight-bearing skill Effec-tiveness of two forms of feedback Physical Therapy 76 985-993

Winstein C J Pohl P S amp Lewthwaite R (1994) Effects of phys-ical guidance and knowledge of results on motor learning Support forthe guidance hypothesis Research Quarterly for Exercise amp Sport 65316-323

Winstein C J amp Schmidt R A (1990) Reduced frequency of knowl-edge of results enhances motor skill learning Journal of Experimen-tal Psychology Learning Memory amp Cognition 16 677-691

Wright D L Li Y amp Coady W J (1994) Cognitive processes re-lated to contextual interference and observational learning A repli-cation of Blandin Proteau and Alain (1994) Research Quarterly forExercise amp Sport 68 106-109

Wright D L Li Y amp Whitacre C (1992) The contribution of elab-orative processing to the contextual interference effect ResearchQuarterly for Exercise amp Sport 63 30-37

Wright D L amp Shea C H (2001) Manipulating generalized motorprogram difficulty during blocked and random practice does not affectparameter learning Research Quarterly for Exercise amp Sport 72 32-38

Wrisberg C A (1991) A field test of the effect of contextual varietyduring skill acquisition Journal of Teaching in Physical Education11 21-30

Wrisberg C A amp Liu Z (1991) The effect of contextual variety onthe practice retention and transfer of an applied motor skill ResearchQuarterly for Exercise amp Sport 62 406-412

Wulf G (1992) The learning of generalized motor programs andmotor schemata Effects of KR relative frequency and contextual in-terference Journal of Human Movement Studies 23 53-76

Wulf G amp Buumlhner M (1996) Unpublished data Max Planck Insti-tute for Psychological Research Munich

Wulf G Houmlrger M amp Shea C H (1999) Benefits of blocked overserial feedback in complex motor skill learning Journal of Motor Be-havior 31 95-103

Wulf G Houmlszlig M amp Prinz W (1998) Instructions for motor learningDifferential effects of internal vs external focus of attention Journalof Motor Behavior 30 169-179

Wulf G Lauterbach B amp Toole T (1999) Learning advantages ofan external focus of attention in golf Research Quarterly for Exerciseamp Sport 70 120-126

Wulf G amp Lee T D (1993) Contextual interference effects in move-ments of the same class Differential effects on program and parame-ter learning Journal of Motor Behavior 25 254-263

Wulf G Lee T D amp Schmidt R A (1994) Reducing knowledgeof results about relative versus absolute timing Differential effects onlearning Journal of Motor Behavior 26 362-369

Wulf G McConnel N Gaumlrtner M amp Schwarz A (in press)Feedback and attentional focus Enhancing the learning of the volley-ball serve through external-focus feedback Journal of Motor Behavior

Wulf G McNevin N H amp Shea C H (2001) The automaticity of

COMPLEX SKILL LEARNING 211

complex motor skill learning as a function of attentional focus Quar-terly Journal of Experimental Psychology 54A 1143-1154

Wulf G amp Schmidt R A (1989) The learning of generalized motorprograms Reducing the relative frequency of knowledge of resultsenhances memory Journal of Experimental Psychology LearningMemory amp Cognition 15 748-757

Wulf G amp Schmidt R A (1994) Feedback-induced variability andthe learning of generalized motor programs Journal of Motor Behav-ior 26 348-361

Wulf G Schmidt R A amp Deubel H (1993) Reduced feedbackfrequency enhances generalized motor program learning but not pa-rameterization learning Journal of Experimental Psychology Learn-ing Memory amp Cognition 19 1134-1150

Wulf G amp Shea C H (1997) [Unpublished data] Max Planck In-stitute for Psychological Research Munich

Wulf G amp Shea C H (1998) [Unpublished data] Max Planck Insti-tute for Psychological Research Munich

Wulf G Shea C H amp Matschiner S (1998) Frequent feedbackenhances complex motor skill learning Journal of Motor Behavior30 180-192

Wulf G Shea C H amp Park J-H (2001) Attention in motor learn-ing Preferences for and advantages of an external focus ResearchQuarterly for Exercise amp Sport 72 335-344

Wulf G Shea C H amp Whitacre C A (1998) Physical guidancebenefits in learning a complex motor skill Journal of Motor Behav-ior 30 367-380

Wulf G Shea J B amp Rice M (1996) Type of KR and KR frequencyeffects on motor learning Journal of Human Movement Studies 301-18

Wulf G amp Weigelt C (1997) Instructions about physical principlesin learning a complex motor skill To tell or not to tell ResearchQuarterly for Exercise amp Sport 68 362-367

Yao W-X Fischman M G amp Wang Y T (1994) Motor skill ac-quisition and retention as a function of average feedback summaryfeedback and performance variability Journal of Motor Behavior 26273-282

Young D E Cohen M J amp Husak W S (1993) Contextual inter-ference and motor skill acquisition On the processes that influence re-tention Human Movement Science 12 577-600

Zamani M amp Richard J F (2000) Object encoding goal similarityand analogical transfer Memory amp Cognition 28 873-886

(Manuscript received May 31 2000 revision accepted for publication July 2 2001)