social loafing: a meta-analytic review and theoretical ... · pdf filesocial loafing: a...

INTERPERSONAL RELATIONS AND GROUPPROCESSES

Social Loafing: A Meta-Analytic Review and Theoretical Integration

Steven 1 Karau and Kipling D. Williams

Social loafing is the tendency for individuals to expend less effort when working collectively thanwhen working individually. A meta-analysis of 78 studies demonstrates that social loafing is robustand generalizes across tasks and S populations. A large number of variables were found to moderatesocial loafing. Evaluation potential, expectations of co-worker performance, task meaningfulness,and culture had especially strong influence. These findings are interpreted in the light of a Collec-tive Effort Model that integrates elements of expectancy-value, social identity, and self-validationtheories.

Many of life's most important tasks can only be accom-plished in groups, and many group tasks are collective tasksthat require the pooling of individual members' inputs. Govern-ment task forces, sports teams, organizational committees,symphony orchestras, juries, and quality control teams providebut a few examples of groups that combine individual efforts toform a single product. Because collective work settings are sopervasive and indispensable, it is important to determine whichfactors motivate and demotivate individuals within these col-lective contexts. Intuition might lead to the conclusion thatworking with others should inspire individuals to maximizetheir potential and work especially hard. Research on socialloafing, however, has revealed that individuals frequently exertless effort on collective tasks than on individual tasks.

Formally, social loafing is the reduction in motivation andeffort when individuals work collectively compared with whenthey work individually or coactively. When working collectively,individuals work in the real or imagined presence of others withwhom they combine their inputs to form a single group prod-uct. When working coactively, individuals work in the real orimagined presence of others, but their inputs are not combinedwith the inputs of others. Determining the conditions underwhich individuals do or do not engage in social loafing is aproblem of both theoretical and practical importance. At apractical level, the identification of moderating variables may

Parts of this article were presented at the 1992 Midwestern Psycho-logical Association convention in Chicago.

We thank Martin Bourgeois, Donal Carlston, Alice Eagly, RebeccaHenry, Janice Kelly, Norbert Kerr, Brian Mullen, Kristin Sommer, andseveral anonymous reviewers for their valuable comments on drafts ofthis article. We also thank Jeffrey Jackson and William Gabrenya forproviding unpublished data.

Correspondence should be addressed to Steven J. Karau, Depart-ment of Psychology, Clemson University, Clemson, South Carolina29634-1511 or to Kipling D. Williams, Department of Psychology,University of Toledo, Toledo, Ohio 43606-3390. Electronic mail maybe sent to [email protected] or [email protected].

suggest means for devising interventions by which social loaf-ing may be reduced or overcome in everyday groups and organi-zations. Latane, Williams, & Harkins (1979) even suggestedthat social loafing is a type of social disease, having "negativeconsequences for individuals, social institutions, and societies"(p. 831). Perhaps as a result of this characterization, social loaf-ing research has been focused on identifying conditions underwhich the effect can be reduced or eliminated. No studies havebeen designed to determine what factors increase social loaf-ing. This emphasis on eliminating the effect is understandable,given the potential applicability of social loafing research toreal-world contexts. However, this emphasis on studying condi-tions in which the effect is not likely to occur may also result inan underestimation of the magnitude of social loafing across awider range of situations.

At a theoretical level, specifying which variables moderatesocial loafing is central to developing a fuller understanding ofthe dynamics underlying the performance and motivation ofboth individuals and groups. The social loafing literature offersa host of findings relevant to theorists interested in evaluationprocesses, the self, and group dynamics. Indeed, in a recentreview of social motivation, Geen (1991) regarded social loafingas one of only three dominant phenomena addressing this issuein the 1980s. Despite the theoretical importance of this topic,surprisingly little attention has been devoted to systematicallyreviewing or integrating the social loafing research. Although anumber of researchers have discussed social loafing or havepresented theories of particular causes of social loafing (e.g.,Harkins, 1987; Harkins & Szymanski, 1987; Jackson & Wil-liams, 1985; Latane, 1981; Mullen, 1983; Paulus, 1983; Shep-perd, 1993; Stroebe & Frey, 1982), they have drawn their conclu-sions from only small portions of the available empirical stud-ies, which were selected by unspecified criteria. Presently, themagnitude and consistency of social loafing across studies hasnever been estimated, and the extent to which particular vari-ables moderate social loafing has remained unclear. Given thelarge empirical literature that is now available, a thorough inte-gration and analysis of this research is long overdue.

Journal of Personality and Social Psychology, 1993. Vol. 65. No. 4, 681-706Copyright 1993 by the American Psychological Association, Inc. 0022-3514/93/$3.00

681

682 STEVEN J. KARAU AND KIPLING D. WILLIAMS

The current review has three major purposes. First, we pres-ent a model that integrates expectancy theories with theories ofgroup-level social comparison and social identity to account forthe findings of studies examining individual effort in collectivesettings. This model, the Collective Effort Model (CEM), isthen used to generate predictions for the meta-analysis. Sec-ond, we systematically review the available empirical findings.Meta-analytic methods (Cooper, 1989; Glass, McGaw, & Smith,1981; Hedges & Olkin, 1985; Rosenthal, 1991) are used to esti-mate the overall magnitude of social loafing and to compare itwith the magnitude of the effects of other variables on socialbehavior. This quantitative method also allows for an examina-tion of the consistency of social loafing across studies and ac-counts for inconsistencies by identifying moderating variablesthat affect the tendency for individuals to engage in social loaf-ing. The empirical status of such moderating variables is criti-cal to a sophisticated understanding of the dynamics of motiva-tion losses in groups. Third, we use the results of the meta-ana-lysis to identify gaps and ambiguities still existing in the socialloafing literature and discuss the potential of the CEM for gen-erating novel predictions to guide future research toward an-swering these questions.

Brief History of Social Loafing Research and Paradigms

The first experiment to suggest a possible decrement in indi-vidual motivation as a result of working in a group was con-ducted over a hundred years ago by Ringelmann (cited in Kra-vitz & Martin, 1986). Male volunteers were asked to pull on arope, tug-of-war fashion, as hard as they could in groups ofvarying sizes. The rope was connected to a strain gauge thatmeasured the group's total effort. The results showed that asgroup size increased, group performance was increasinglylower than would be expected from the simple addition of indi-vidual performances. However, Steiner (1972) proposed twopossible causes for this performance decrement: (a) reducedindividual motivation or (b) coordination loss. Steiner favoredthe latter cause as most parsimonious, concluding that individ-uals may fail to synchronize their efforts in a maximally effi-cient manner (e.g., pulling while others are pausing), thus evi-dencing less productivity, but not necessarily less effort.

To separate effort reduction from coordination loss, Ingham,Levinger, Graves, and Peckham (1974) had subjects perform arope-pulling task both in actual groups and in pseudogroups inwhich blindfolded male students believed they were pullingwith other group members but were actually pulling alone.Data for the pseudogroup trials showed that performance stilldecreased as perceived group size increased, suggesting thatindividuals exerted less effort when working in groups thanwhen working individually. However, group size covaried in-versely with audience size in Ingham et al.'s studies, leavingopen the possibility that effort was facilitated by larger au-diences in the individual trials.

Latane et al. (1979) conceptually replicated these findingswhile holding audience size constant. They had subjects shoutand clap as loudly as they could, both individually and withothers. Blindfolded and wearing headphones that masked thenoise, college men shouted both in actual groups and in pseu-dogroups in which they shouted alone but believed they were

shouting with others. Latane et al. demonstrated that a substan-tial portion of the decreased performance of groups was attrib-utable to reduced individual effort, distinct from coordinationloss, and that audience size did not account for these results.They also coined the term social loafing for the demotivatingeffects of working in groups.

Since 1974, nearly 80 studies on social loafing have beenconducted in which individuals' coactive efforts were com-pared with individuals' collective efforts. These studies haveused a wide variety of tasks, including physical tasks (e.g., shout-ing, rope-pulling, and swimming), cognitive tasks (e.g., generat-ing ideas), evaluative tasks (e.g., quality ratings of poems, edito-rials, and clinical therapists), and perceptual tasks (e.g., mazeperformance and vigilance tasks on a computer screen). Bothlaboratory experiments and field studies have been conductedusing a range of subject populations varying in age, gender, andculture.

Theoretical Accounts for Social Loafing

Although several researchers have offered theories of socialloafing, these viewpoints have typically been restricted to ex-plaining only one of several possible causal mechanisms andgenerally do not attempt to include the wide range of variablesthat moderate social loafing. In contrast, social loafing re-search has adopted an orientation toward isolating conditionsunder which the effect can be reduced or eliminated. As a re-sult, despite the tendency of researchers to manipulate a widearray of moderating variables across individual studies, therehas not been a concerted attempt to integrate all of these moder-ators into a single theoretical model. Indeed, each moderatingvariable is frequently associated with a separate theory. Never-theless, the theories that have been advanced do provide someinsight into the nature and possible causes of social loafing.Therefore, before presenting the CEM, we first present an over-view of viewpoints offered by prior researchers and discuss theneed for an integrative model.

Social Impact Theory

According to Latane's (1981) social impact theory, people canbe viewed as either sources or targets of social impact. Theamount of social impact experienced in a situation is thought tobe a function of the strength, immediacy, and number ofsources and targets present. In the collective condition of thetypical social loafing experiment, the experimenter serves as asingle source of social impact, whereas the group membersusually serve as multiple targets of social impact. Social impacttheory suggests that the experimenter's request to try as hard aspossible on the task should be divided across targets, resultingin reduced effort as group size increases. The division of impactis predicted to follow an inverse power function, with a negativeexponent having an absolute value less than 1, thereby resultingin marginally decreasing impact as group size increases. In thecoactive condition, however, in which inputs are not combined,each individual feels the full impact of the experimenter's re-quest and works hard. The strength of social impact theoryappears to be its ability to specify group size effects. Thestrength and immediacy factors, however, have been neglected

SOCIAL LOAFING 683

by researchers, and the theory itself has been criticized for notaddressing the underlying psychological processes that it de-scribes (e.g., Mullen, 1985).

Arousal Reduction

Jackson and Williams (1985) proposed a drive explanation toaccompany a social impact theory explanation of social loaf-ing. They argued that the presence of other co-workers is drivereducing because these others serve as cotargets of an outsidesource of social impact. Citing studies in which people facedwith a fearful situation tended to prefer being in the company ofothers (e.g., Schachter, 1959; Wrightsman, 1960), they reasonedthat the presence of others is not necessarily drive inducing (ashas been repeatedly demonstrated in social facilitation re-search). Instead, the presence of others should only be driveinducing when those others serve as sources of impact, butshould be drive reducing when they serve as cotargets. Jacksonand Williams (1985) found support for this logic in an experi-ment that combined features of the social loafing and socialfacilitation paradigms. Subjects completed simple and complexcomputer mazes either alone, coactively, or collectively. On sim-ple tasks, subjects performed better coactively than collectively.On complex tasks, however, subjects performed better collec-tively than coactively. Jackson and Williams argued that work-ing collectively led to reduced drive and effort, resulting in de-creased performance on simple tasks (where the dominant re-sponse is likely to be correct) and increased performance onnovel, difficult tasks (where the dominant response is likely tobe in error).

Evaluation Potential

Many interpretations of social loafing invoke the concept ofevaluation potential (e.g., Harkins, 1987; Harkins & Jackson,1985; Williams et al, 1981). Some researchers (Harkins, 1987;Harkins & Szymanski, 1987,1989; Kerr & Bruun, 1983) haveeven defined social loafing as motivation loss in groups causedby reduced identifiability or evaluation.1 They argue that socialloafing occurs because, in most studies, individuals' inputs canonly be evaluated in the coactive condition. In the collectivecondition, of course, individual inputs are combined into onegroup product. When working on collective tasks, individualscan "hide in the crowd" (Davis, 1969) and avoid taking theblame for a poor group performance. Collective tasks may alsolead individuals to feel "lost in the crowd" (Latane et al., 1979)such that they cannot receive their fair share of the credit for agood group performance. The role of identifiability in socialloafing is illustrated in a study by Williams, Harkins, and La-tane (1981). Individuals performed a shouting task coactivelyand collectively. Evaluation potential was manipulated suchthat individual inputs were either always identifiable, neveridentifiable, or identifiable only in the coactive condition. Indi-viduals only loafed when identifiability and the coactive-col-lective variable were confounded. Recent research by Harkinsand colleagues (Harkins, 1987; Harkins & Jackson, 1985; Har-kins & Szymanski, 1987,1989) has suggested that making indi-viduals' collective inputs evaluable to anyone (including one-self) may be enough to eliminate social loafing in many situa-

tions. This research has also demonstrated that tworequirements must be met for evaluation by any source (theexperimenter, one's co-workers, or oneself) to be possible: (a)the participant's output must be known or identifiable, and (b)there must be a standard (personal, social, or objective) withwhich this output can be compared.

Dispensability of Effort

Kerr and his colleagues (e.g., Kerr, 1983; Kerr & Bruun, 1983)have suggested another possible cause of social loafing: Individ-uals may exert less effort when working collectively becausethey feel that their inputs are not essential to a high-qualitygroup product (i.e., are dispensable). In their "free rider" para-digm, these researchers have found that individuals tend toreduce their collective efforts when working on threshold tasksthat use a disjunctive rule whereby if any one of the groupmembers reaches a certain performance criterion, the groupsucceeds and further effort is unnecessary. Moreover, this re-duction in individual effort occurs even though each groupmember's contribution is made identifiable to themselves, theirpartner, and the experimenter. Thus, on some tasks individualsmay be unwilling to exert effort if they feel their input will havelittle impact on the resulting group product.

Matching of Effort

Jackson and Harkins (1985) proposed that people tend tomatch their co-workers' efforts when working collectively. Ac-cording to this position, social loafing occurs because individ-uals expect others to slack off in groups and, therefore, reducetheir own efforts to maintain equity. In Jackson and Harkins's(1985) experiment, participants' expectations of how hard theirco-worker would work were manipulated such that participantsexpected their co-worker either to try hard or not to try hard ona shouting task. Social loafing was eliminated and participantsmatched their co-worker's anticipated effort. However, in thisstudy, the confederate co-worker's statement of her intendedeffort was confounded with her evaluation of the experiment'sworth. In the high-effort condition, she said that she thoughtthe experiment was interesting and that she was going to tryhard. In the low-effort condition, however, she said that shethought the experiment was boring, and that she was not goingto try hard. Research on job attitudes (see Zalesny & Ford,1990) suggests that workers' perceptions of and motivationstoward their task are highly influenced by co-workers' task as-sessments. Thus, the confederate's assessment of the worth ofthe task could have alone accounted for Jackson and Harkins's

1 Although evaluation is likely to play an important role in socialloafing, it seems premature to define social loafing in these termsbecause (a) defining the phenomenon in terms of its causes prevents,by definition, the discovery of new causes for the same effect, and (b)evidence suggests that there are other causes of social loafing, such asredundancy of contributions (Harkins & Petty, 1982) and dispensabi-lity of effort (Kerr & Bruun, 1983). Therefore, the present reviewadopts a less restrictive definition of social loafing consistent with theoriginal formulation: the tendency to reduce one's effort when workingcollectively compared with coactively on the same task.


results. Moreover, recent research by Williams and Karau(1991) has demonstrated that individuals may actually increasetheir collective effort when they expect their co-workers to per-form poorly on a meaningful task, an effect referred to as socialcompensation.

Self-Attention

Mullen (1983) proposed that self-attention underlies socialloafing. According to this self-attention perspective, workingon a collective task leads to a decrease in self-awareness, lead-ing individuals to disregard salient performance standards andto engage in less self-regulation. Thus, collective performance islower than coactive performance, because participants aremore attentive to task demands and performance standardswhen working alone. Mullen has found evidence consistentwith the notion that self-attention may influence aspects ofsocial behavior in various meta-analyses (see Mullen, 1991, fora summary), but there is currently no evidence that reducedself-attention causes social loafing (see Jackson, 1986). Onlyone study (Stevenson, 1990) has manipulated self-attentionacross coactive and collective conditions, and the results did notsupport the self-attention interpretation. Therefore, althoughthe self-attention hypothesis is provocative, it is in need of em-pirical support.

Conclusions

Prior theories advanced to explain social loafing all appear tohave some limitations. The most significant limitation com-mon to all of these viewpoints is that they tend to offer explana-tions and make predictions about conditions under which so-cial loafing will occur within a limited domain. For example,Harkins (1987) offered an excellent analysis of the impact ofevaluation on social loafing, and Kerr (1983) contributed anequally effective analysis of the role that dispensability plays.However, although current theories of social loafing provideinsights as to why the effect occurs at all, they do not provide aframework that can clearly specify which particular factorsshould moderate social loafing under different conditions.When possible in the current review, we quantitatively examinethe adequacy of some of these viewpoints for explaining socialloafing. More important, we present a unified theory that pro-poses to envelop most if not all of the findings in the existingsocial loafing literature and that offers specific predictions forfuture research.

Integrative Model of Individual Efforton Collective Tasks

The integrated model of individual effort on collective tasks,the CEM, adapts individual-level expectancy-value models ofeffort to collective contexts to highlight the most likely threatsto motivation and uses recent theories of self-evaluation ingroup contexts to clarify which outcomes are likely to be valuedby individuals when working collectively. The resulting modelprovides a framework for clearly specifying the implications ofany key attribute of a given collective setting (e.g., group size andnature of task) for the motivation of individuals in that setting.

We propose that expectancy-value models of effort (e.g.,Heckhausen, 1977; Porter & Lawler, 1968; Vroom, 1964) can beextended to collective contexts by specifying how working in agroup influences individuals' perceptions of the relationshipbetween their effort and their expected outcomes. We believethat social loafing occurs because there is usually a strongerperceived contingency between individual effort and valuedoutcomes when working individually. When working collec-tively, factors other than the individual's effort frequently deter-mine performance, and valued outcomes are often dividedamong all of the group members. The notion that expectancy-value models can be applied to motivation in group contexts isnot entirely new. For example, several researchers have implic-itly used expectancy-value logic to cast light on the findings ofsocial dilemma and social loafing studies (e.g., Kerr, 1983; Ol-son, 1965; Shepperd, 1993; Williams & Karau, 1991). Naylor,Pritchard, and Ilgen (1980) have also suggested that expec-tancy-value logic can be extended into a model that can accountfor organization-level motivational phenomena. Similarly, sev-eral researchers have discussed the potential implications ofself-evaluation for individual motivation (e.g., Goethals & Dar-ley, 1987; Taylor & Brown, 1988; Tesser, 1988). The CEM, how-ever, is unique in specifying additional contingencies betweeneffort and outcomes that are distinctive to collective contexts,and in using group-level social comparison and self-evaluationtheories to specify which factors should influence individualmotivation in collective contexts. This combination allows forthe clear specification of potential threats to collective motiva-tion in a way that is not restricted to mere task-performanceoutcomes.

The key features of the CEM are shown in Figure 1. Themodel suggests that individuals will be willing to exert efforton a collective task only to the degree that they expect theirefforts to be instrumental in obtaining valued outcomes. Thus,a number of factors must be perceived as existing before individ-uals will be willing to exert high levels of effort. Individualeffort must relate to individual performance, which must inturn have some impact on the group's performance. The group'sperformance must then lead to a favorable group outcome,which must be related to a favorable individual outcome. If thetask performance setting disrupts any of these relationships in away that is salient to individuals, then they will not be likely toview their efforts as useful and will not work as hard on thetask. Similarly, individuals will not work as hard when the avail-able outcomes are not valued, even if these outcomes are di-rectly related to individual effort. Relevant individual out-comes include things such as objective outcomes (such as pay),self-evaluation information, and feelings of purpose or belong-ing in one's group. The relative value of these outcomes dependson a number of factors, including the meaningfulness and in-trinsic value of the task, the task's importance to the individualand group, the degree to which the individual is dispositionallypredisposed to view collective outcomes as important, and thedegree to which the outcome provides information relevant tothe individual's self-evaluation.

Like traditional expectancy-value models of effort, the CEMassumes that individuals behave hedonistically and try to max-imize the expected utility of their actions. In Vroom's (1964)original model, individuals' motivational force is dependent on

SOCIAL LOAFING 685

Expectancy

» §3> o

a*

I Instrumentality | X | Valence of Outcome | = | Motivational Force

Individual Effort |-»| Individual Performance ^->flridividual Outcomes | —j,

Individual Effort [»[individual Performance | | Individual Outcomes

• Self-evaluation• Feelings of Belonging• Intrinsic Rewards• Extrinsic Rewards

T| Group Performance |-»j Group Outcomes

• Group Evaluation• Group Cohesiveness• Extrinsic Rewards

Figure 1. The Collective Effort Model (CEM).

->| Outcome Valence |• Task Importance• Task & Reward

Meaningfulness• Individual Differences

(e.g., collectivism,culture, gender)

• Evaluation Apprehension

three factors: (a) expectancy, or the degree to which high levels ofeffort are expected to lead to high levels of performance, (b)instrumentality, or the degree to which high-quality perfor-mance is perceived as instrumental in obtaining an outcome,and (c) valence of the outcome, or the degree to which the out-come is viewed as desirable. The CEM expands on this logic byspecifying that instrumentality on collective tasks is deter-mined by three factors: (a) the perceived relationship betweenindividual performance and group performance, (b) the per-ceived relationship between group performance and group out-comes, and (c) the perceived relationship between group out-comes and individual outcomes. Thus, the model suggests thatworking on a collective task introduces additional contingen-cies between individuals' efforts and their outcomes. For thisreason, an analysis of individual effort is by necessity morecomplex for collective tasks than for coactive tasks.

The CEM also acknowledges that, although individual andgroup performance are strongly related to many of the valuedoutcomes in performance settings, there are some valued out-comes that do not depend on performance. For example, whenworking on intrinsically meaningful tasks or when workingwith highly respected co-workers, exerting high levels of effortmay lead to self-satisfaction, approval from the group, or otherimportant outcomes, even if such high effort has little or noimpact on tangible performance outcomes. However, themodel emphasizes performance-related outcomes becausemany of the most important outcomes available in a perfor-mance context are contingent on actual performance.

The CEM can be described as "cognitive" both because per-ceived rather than actual contingencies are hypothesized to in-fluence behavior and because individuals are hypothesized toeither consciously or subconsciously select a level of effort toexert on the task. However, the model is not necessarily deliber-ative because individuals are unlikely to systematically processall of the available information about the task performancesituation, unless they are particularly motivated to process itbecause of situational constraints or individual differences.Therefore, some situations may lead individuals to respond au-tomatically to a preexisting effort script, whereas other situa-tions may lead individuals to strategically increase or decreasetheir collective effort. The CEM suggests that individuals gener-ally consider salient features of the task setting when attempt-ing to maximize the expected utility of their actions.

The CEM also expands on prior expectancy-value frame-works by specifying which outcomes are likely to be valued incollective settings. Valued outcomes can consist of either objec-tive outcomes such as pay or subjective outcomes such as enjoy-ment, satisfaction, feelings of group spirit and belonging, andfeelings of self-worth. In the case of objective outcomes, how-ever, it is the individual's evaluation of the outcome rather thanthe outcome itself that determines its valence (e.g., Deci, 1975;Lepper & Greene, 1978). Because the CEM is oriented to ex-plaining collective phenomena, the model places particular em-phasis on group-level outcomes that have implications for theindividual's self-evaluation. Research has repeatedly demon-strated that individuals are quite concerned with maintaining a


favorable self-evaluation (e.g., Brown, Collins, & Schmidt, 1988;Greenwald, 1982; Greenwald & Pratkanis, 1984; Suls & Miller,1977; Tesser, 1988). Group performance situations produce thepotential for self-evaluation from a variety of relevant sources(Crocker & Luhtanen, 1990; Leary & Forsyth, 1987). Brecklerand Greenwald (1986) proposed that the self has distinct moti-vational facets that are influenced most strongly by three corre-sponding audiences. They distinguished between (a) a privateself that responds primarily to internal values, (b) a public selfthat responds mainly to others' evaluations, and (c) a collectiveself that is sensitive primarily toward fulfilling one's role withinthe context of the goals of important reference groups. Al-though most social loafing research has been focused on themotivation produced by the experimenter's evaluation of perfor-mance (for prominent exceptions, see Harkins & Szymanski,1988, 1989), Breckler and Greenwald's analysis suggests thatindividuals can be motivated by appeals to any of the threemotivational facets of the self (see also Crocker & Luhtanen,1990). With respect to the CEM, task situations that provideclear information relevant to self-evaluation, whether from one-self, one's co-workers, reference groups, or others, should bemore inherently motivating than situations that do not providesuch information or that make such information ambiguousand less potent.

Social identity theory and group-level versions of social com-parison theory lead to similar conclusions. Social identitytheory (e.g., Abrams & Hogg, 1990; Tajfel & Turner, 1986) pro-poses that individuals gain positive self-identity through theaccomplishments of the groups to which they belong (see alsoCialdini et al., 1976; Levine & Moreland, 1987). Indeed, severalrecent analyses suggest that some social motivations, such as aneed for belonging or a need for social communication, canonly be fulfilled in a collective setting (e.g., Brewer, 1991; Capor-ael, Dawes, Orbell, & van de Kragt, 1989). Similarly, Goethalsand Darley's (1987) group-level revision of social comparisontheory suggests that individuals not only compare themselveswith other individuals to obtain self-evaluation information,but that they also compare the groups to which they belong toother groups for the same reason. Goethals and Darley furtherstated that individuals may seek such group-level comparisoninformation both to obtain self-knowledge (information aboutthe actual level of one's performance compared with others) andself-validation (information that level of one's performance issuperior to that of others). Taken as a whole, the recent theoriesof group level self-evaluation processes all suggest that collec-tive contexts that provide a great deal of information relevant toone's self-validation should be more important to individualsthan contexts that provide less information (or none). Theseperspectives also suggest that self-evaluation information cancome from a variety of sources and that information relevant toone's role in valued reference groups may be especially influen-tial.

Predictions for Meta-Analysis

On the basis of the preceding logic, the CEM provides anumber of predictions for our meta-analysis. First, because themodel suggests that individuals' outcomes are frequently lessreliant on their efforts when working collectively than coacti-

vely, we predicted that individuals would generally tend to en-gage in social loafing and that this effect would be reliableacross studies. Second, the CEM suggests that individuals willwork harder on a collective task when they expect their effortsto be instrumental in obtaining valued outcomes. Therefore,we predicted that, holding other factors constant, social loafingshould be reduced when individuals: (a) believe that their col-lective performances can be evaluated by the experimenter,their co-workers, themselves, or others; (b) work in smallerrather than larger groups; (c) perceive that their contributions tothe collective product are unique, rather than redundant withthe inputs of others; (d) are provided with a standard with whichto compare their group's performance; (e) work on tasks that areeither intrinsically interesting, meaningful to the individual,important to one's reference group or to valued others, or highin personal involvement; (f) work with respected others (highgroup valence; friends, teammates, partners, and respected co-workers) or in a situation that activates a salient group identity;(g) expect their co-workers to perform poorly; and (h) have adispositional tendency to view favorable collective outcomes asvaluable and important.

Regarding the final prediction, several factors are likely toinfluence the degree to which an individual values collectiveoutcomes relative to individual outcomes. Two especially prom-inent factors may be gender and culture. Recent gender re-search seems to suggest that women tend to be more group- orcollectively-oriented than men. For example, research ongender stereotypes (e.g., Broverman, Vogel, Broverman, Clark-son, & Rosenkrantz, 1972; Eagly & Steffen, 1984) has demon-strated that men are expected to possess high levels of agenticqualities (such as being independent and assertive), whereaswomen are expected to possess high levels of communal quali-ties (such as being friendly, unselfish, and concerned withothers; see Bakan, 1966). Similarly, research conducted insmall-group settings has found that men tend to specialize inactivity oriented toward task completion, whereas women tendto specialize in activity oriented toward meeting interpersonaldemands within the group (Anderson & Blanchard, 1982;Carli, 1982). Taken together, these findings suggest that menappear to be oriented toward individualistic and competitiveconcerns, whereas women appear to be oriented toward inter-personal and cooperative concerns. An extension of this logicto social loafing suggests that women are more likely than mento view performing well on collective tasks as important. Thus,we predicted that, although both sexes are likely to engage insocial loafing, women should tend to loaf less than men.

In a similar manner, the degree to which the dominant cul-ture from which subject populations were selected emphasizesindividualistic versus collectivistic concerns may moderate thesocial loafing effect. Eastern or Oriental culture is often de-picted as group- or socially oriented, whereas Western or Amer-ican culture is often depicted as individualistically oriented(e.g., Hsu, 1970; Triandis, 1989; Wheeler, Reis, & Bond, 1989).Therefore, subjects from Eastern cultures are more likely toview performance on collective tasks as important than aresubjects from Western cultures. Thus, we predicted that sub-jects in countries such as Japan, Taiwan, and China should loafless than subjects in the United States and Canada.

Finally, as discussed earlier, Jackson and Williams (1985)

SOCIAL LOAFING 687

proposed that the presence of others is drive reducing whenother group members serve as cotargets of an outside source ofsocial influence. According to drive models of anxiety, workingcollectively should lead to reduced drive, resulting in decreasedperformance on simple tasks and increased performance onnovel, difficult, tasks. In our meta-analysis, we addressed theadequacy of this viewpoint by classifying studies according towhether they used simple or complex tasks.

Some of our predictions mirror well-established effects inthe social-loafing literature, whereas other predictions have ei-ther been examined only sporadically, if at all, or have producedconflicting evidence. For example, researchers generally acceptthe conclusion that individuals tend to engage in social loafingwhen working collectively, and a number of studies have foundthat social loafing can be reduced or eliminated when evalua-tion is provided in the collective condition. However, the effectsof group-level comparison standards, task meaningful ness,and task uniqueness have only been examined in a handful ofstudies. The degree to which these variables consistently influ-ence social loafing across studies is not known, and only tenta-tive conclusions about their effects can currently be drawn. Fi-nally, contradictory results have emerged from studies examin-ing gender, culture, expectations of co-worker performance,group size, and group valence. In the present review, we quanti-tatively examine the influence of all of these variables on socialloafing, using all available comparisons, to provide a clearerunderstanding of what factors moderate social loafing in differ-ent situations. With respect to well-established findings, ourmeta-analytic methods allowed us to estimate the magnitude ofsocial loafing and examine its consistency across studies. Simi-larly, we were able to assess quantitatively the impact of evalua-tion on social loafing and compare it with the influence ofother variables affecting collective effort.

Method

Sample of Studies

Computer-based information searches were conducted using the keywords social loafing, free-rider, Ringelmann, motivation loss, motiva-tion decrement, collective task, and collective performance. These keywords were searched in the following databases: Psychological Ab-stracts (PsycINFO), Dissertation Abstracts International (DISS), Socio-logical Abstracts (SOCA), Educational Resources Information Center(ERIC), and a worldwide business and management database (ABI/INFORM). In addition, the reference lists of numerous review articlesand chapters, as well as the reference lists of all located studies, weresearched. Finally, several researchers provided data from unpublishedstudies that had been presented at regional or national meetings.

Three key criteria were adopted for including studies in the sample.First, the study had to compare the effort of individuals working eitherindividually or coactively with the effort of individuals working collec-tively. Collective performance was denned as a situation in which indi-viduals' inputs were combined into one group score. Thus, studies wereexcluded if they merely compared the effort of individuals across dif-ferent group sizes without including a coactive control condition (e.g.,Kerr& Bruun, 1983, Experiments 1 and 2; Zaccaro, 1984): Theindivid-ual-versus-group work context, which is central to the social loafingconstruct, is not manipulated in such studies. This criterion also ex-cluded studies in which only evaluation potential was manipulated,without providing subjects with an expectation that their inputs would

be combined with those of other subjects (e.g., Price, 1987, Experi-ment 1).

Second, studies were excluded if the number of task performers (i.e.,coactors or co-workers) present was confounded across coactive andcollective conditions (e.g., Huddleston, Doody, & Ruder, 1985; Kerr &Bruun, 1981, Experiment 1, between-subjects comparisons). In suchstudies, effort cannot be separated from distraction and mere presenceeffects. Of course, this criterion excluded most social facilitation stud-ies, because they intentionally varied the number of individuals pres-ent in the alone and group conditions.2

Studies were also omitted if individual effort could not be separatedfrom coordination loss (e.g., Ingham et al., 1974, Experiment 1; seeintroduction). This criterion excluded studies that merely comparedthe average individual performance with the average group perfor-mance without measuring individual performances under collectiveconditions.

Variables Coded From Each Study

The following general information was coded for each study: (a) dateof publication, (b) publication form (journal article, meeting paper,dissertation or master's thesis, or unpublished document), (c) numberof subjects, (d) age of subjects, and (e) status of subjects (third gradestudents or younger, fourth through sixth grade students, junior highor high school students, college students, or organizational employees).The following methodological characteristics were also coded: (a) set-ting of study (laboratory or field), (b) type of coactive-collective com-parison (within-subjects or between-subjects), and (c) type of coverstory used (effort or performance related or other, e.g., studying theeffects of "sensory sound dynamics," improving writing in a collegejournalism department, or obtaining evaluations of clinical thera-pists).

In addition, the following attributes of the task were coded: (a) com-plexity of task (simple or well-learned, complex or novel, or unknownor unclear); (b) type of effort required by task (physical, e.g., shouting orrope-pulling; cognitive, e.g., generating ideas; perceptual, e.g., vigi-lance, maze performance; or evaluative, e.g., quality rating of poems oreditorials); (c) quantity or quality emphasis of task using Steiner's (1972)typology (maximizing, optimizing, and mixed or unclear); and (d)method of combining individual inputs, again using Steiner's typology(additive, compensatory, disjunctive, and mixed or unclear).

Finally, the following predictors relevant to the CEM were alsocoded: (a) the conditions under which individual inputs could be evalu-ated, that is, evaluation potential (none, coactive condition only, coac-tive and collective conditions, and unclear); (b) task valence (high, low,and unclear or unspecified); (c) group valence (high, e.g., close friendsor couples, group cohesiveness manipulation, or teammates; moder-ate, e.g., mere acquaintances; low, e.g., strangers; unknown or unclear,e.g., clear possibility that subjects might be acquainted but not stated inreport); (d) opportunity for group evaluation (present or absent); (e)expectations of co-worker performance (high, low, or unclear or un-specified); (f) uniqueness of individual task contributions (unique, e.g.,brainstorming tasks in which each subject generates uses for a different

2 Several studies (e.g., Harkins & Szymanksi, 1988; Weldon & Gar-gano, 1985) had subjects perform a task "alone" in the collective condi-tion by leading these individuals to believe that their inputs would becombined with those of others who would participate at different ses-sions, thus holding group size constant across individual and collectiveconditions. There were no significant differences between studies thathad subjects work alone (across individual and collective conditions)and studies that had subjects work in the presence of others (acrosscoactive and collective conditions).


object or vigilance tasks in which each subject watches for randomlyoccurring blips in different quadrants of a video terminal; potentiallyredundant, e.g., brainstorming tasks in which all subjects generate usesfor the same object, evaluative tasks that use a compensatory methodof combining inputs; completely redundant, e.g., pulling on a rope,shouting and clapping, or pumping air); (g) group size (i.e., number ofindividual inputs combined into the group product); (h) number oftask performers present at session; (i) sex of subjects (men, women, orboth); and (j) culture of subjects (Eastern, i.e., China, Japan, or Taiwanor Western, i.e., United States or Canada).

Participants' efforts were only coded as evaluable if (a) their outputwas identifiable either to the experimenter, their co-workers, or them-selves, and (b) a personal, social, or objective standard was availablewith which their outputs could be compared. This treatment is consis-tent with widely accepted interpretations of evaluation potential (e.g.,Harkins, 1987; Harkins & Szymanski, 1988,1989).

Task valence was coded as high (or low) if researchers manipulatedtask importance, task meaningfulness, or personal involvement insuch a way that subjects were provided with a clear rationale for view-ing the task as especially important or trivial (e.g., Brickner, Harkins, &Ostrom, 1986; Williams & Karau, 1991). Task valence was also codedas high (or low) if subjects were selected on a personality variable thatwould make them likely to view the task used as especially involving(or aversive), or were selected on the basis of their probable intrinsicinterest (or disinterest) in the task (e.g., subjects high and low in need forcognition performing an idea-generation task—Petty, Cacioppo, &Kasmer, 1985; collegiate swimmers at a competitive meet—Williams,Nida, Baca, & Latane, 1989). In all other cases, task valence was scoredas unclear or unspecified. For practical purposes, tasks that werescored as unclear or unspecified are likely to fall somewhere betweenhigh- and low-valence tasks, and therefore can reasonably be regardedas somewhat moderate in valence.

Expectations of co-worker performance were coded as high (or low)on the basis of researchers' manipulations of co-worker ability or co-worker effort, as well as on the basis of subjects' selection on interper-sonal trust scores (low trusters may expect others to engage in socialloafing, cf. Williams& Karau, 1991). Expectations of co-worker perfor-mance were also coded as high if subjects were experts at the task (e.g.,collegiate swimmers). In all other cases, expectations were coded asunclear or unspecified.

The categories listed for some of these variables were simplifiedfrom an initially more detailed set of categories either because (a) somecategories had few or no entries or (b) prediction of the effect sizes wasnot improved by the more detailed categories. The variables for whichthe initial coding was more detailed are the following: status of sub-jects, culture of subjects, type of effort measured, method of combin-ing individual inputs, and evaluation potential.

All variables were coded by Steven j . Karau. To estimate reliability,Kipling D. Williams independently coded a subsample of 19 studies(24%). The median agreement was 100% (Magreement = 98%, kappa =.95). "Complexity of task" yielded the lowest agreement, 83% (kappa =.71). Disagreements were resolved by discussion. A complete list of thestudies included in the meta-analysis, along with their characteristicsand effect sizes, is presented in Table 1.

Computation and Analysis of Effect SizesThe effect size calculated is g, the difference between individual

effort in the coactive and collective conditions, divided by the pooledstandard deviation (see Hedges & Olkin, 1985). A positive sign indi-cates that subjects exerted more effort coactively than collectively (i.e.,engaged in social loafing), and a negative sign indicates that subjectsexerted more effort collectively than coactively. The effect sizes werecalculated by Steven J. Karau with the aid of a computer program(Johnson, 1989). Although objective effort measures are of central con-

cern to this meta-analysis, effect sizes based on self-report measures ofeffort were also calculated so that supplementary analyses could beconducted. Several studies provided data for more than one relevantdependent variable. When this occurred, separate effect sizes werecomputed for each dependent variable, and these separate effect sizeswere then combined using Rosenthal and Rubin's (1986) suggestedformula.3

Multiple effect sizes from single studies. Under some circum-stances, studies were partitioned and separate effect sizes were com-puted within levels of an independent variable. This partitioning wasundertaken when (a) subjects of different status or from different cul-tures were used (7 studies); (b) group size varied (11 studies); (c) differ-ent tasks were used that varied either in their complexity, the unique-ness of individual inputs to the collective product, or the method usedto combine individual inputs (10 studies); (d) evaluation potential var-ied (17 studies); (e) task valence varied (6 studies); (f) group valencediffered (5 studies); or(g) expectations of co-worker performance weremanipulated (7 studies). Although this partitioning of studies createdsome nonindependence in the data, the strategy was necessary to al-low theory-relevant interactions in the studies to be represented and totest the CEM's predictions about moderating variables. A study thatmanipulated one of these variables was partitioned only if its reportedfindings were sufficient to allow the computation of effect sizes withinthe levels of the variable or to determine the direction or significanceof findings within the levels. By this strategy, the original sample of 78studies obtained from 59 documents produced 166 units, 26 of whichwere intact studies and 140 of which were subdivided parts of studies.For simplicity of exposition, these units are referred to as studies in theremainder of this manuscript. For some analyses, the 166 units werefurther partitioned by sex of subject when data were separately re-ported for male and female subjects. When this latter partitioning wasundertaken, 178 units were available for analysis.

Analysis of effect sizes. The analysis was based on the methodsdetailed by Hedges and Olkin (1985). The gs were converted to ds bycorrecting them for bias. To obtain an overall estimate of the differ-ence in individuals' coactive and collective effort reported in the avail-able research, the relevant study outcomes were then combined byaveraging the ds. All such means were computed with each effect sizeweighted by the reciprocal of its variance, a procedure that gives moreweight to effect sizes that are more reliably estimated.4 A homogeneitystatistic, Q, was also calculated to determine whether each set of dsshared a common effect size (i.e., was consistent across the studies). Totest the CEM's predictions about moderating variables, we accountedfor variability in the effect sizes by relating them to attributes of thestudies using both categorical and continuous models. The categoricalmodels, which are analogous to analyses of variance (ANOVAs), pro-vide a between-classes effect, QB (analogous to a main effect in an

{text continues on page 694)

3 The between-measures correlations that were used when imple-menting this formula were estimated from correlations (a) given inseveral studies in the sample and (b) estimated from data provided byWilliams (from Williams & Williams, 1981, Experiment 1) and Ga-brenya (from Gabrenya, Latane, & Wang, 1983). The resulting correla-tions were .82 for subjective effort, .71 for objective measures based onshouting and clapping tasks, and .49 for objective measures based onevaluative tasks.

4 Because some studies used the group as the unit of analysis,whereas others used individuals or individuals across repeated factorsas the unit of analysis, the number of subjects in the coactive and col-lective conditions were used in all cases to represent sample size inthese calculations in an attempt to avoid arbitrarily weighting someeffect sizes dramatically higher than others despite comparable num-bers of subjects.

SOCIAL LOAFING 689

fN CNCN fN co"co" fN CN

m r m

© © . . gn

-L-L ^.ci J^-LfN CN fN fN CN CN

8888^8 88 TTTT 888^ • " ^ • • ^ • T t — (N t/">v~i ^ - ^ ^^p^- fN CN CN

.©pop© pp ooo© 999•^Tj-'^-rJ-'Tt-fN v-iV) — —. - ^ —i (NfNCN

8888f N CN CN CN

8888 llslsioogopp © p~ ™ CN r4

I I

88

o

H

I CN CN CN fN CN

J"fN fN fN fN fN

I f N CNfS CNfNCNCS f N f N f N (NCNfNCN

CNCNCNfNCNCN CNCN CNfNfNCN CNtNfN CNCNCNCN

f^ CO CO CO CO CO CO ^J fN CN fN CN * ^ ^ ^ "^f fN fN CN ^ ^

rN (N (N fS (N (N

fN (N (N fS (N (N

fN (N (N <N <N

fN fN fN fN fN

fN fN fN CN fN

fN ?T fN" fN" $•

—'fN r*ir*"i ^ ^ ^ - —- H - —H - —H ^ - —i ^ - _ * _ - _ . ^ - - _ ^ ^ - ^ ^ - . . ^ M ^ H fN fNfNCNCNfN

fN fN "TT '^ ' m r *1 fN fN fN fN fN —" fN fN f N f N f N f N —•• ~< —« - • ^ _ ^ —H r^mrîfomr**!

—>CN mm ^-i —> »« _^ —• _ _ - ^ — — c i m m c i — —• - - ^ — , _ * « fîmf^fîr^r^

fN fN fN fN —<

—• r«1 fN - ^ fN

fN <N fN fN fN —

CN fN fN —« —« —

~ ^ — . - ^ — . — , ^ (NfNfN

fNfN r-lfN(NfS - ^ — -H

fNfN fNfN — fN - ^ —• CN

- ^ —• CNCNCNfN —• —• —

(N fN fN fN

O ^

o —O O co O <

— fN Cl —

188 O O P oo O OPOOO QO

qpp^p poppp ^P'•» — Tt<NO O O O O O O O

f o OOfNOO \OÔ^DÔ —

O OOO O

I

O n O O a \ NO o rn rs r- OOO—"OO

o oI I

OCNCT\vifNrn — o\OfNf^fN

O (

1 f

u'B.c

c•5c

'5.a2

IIsi

0 0

c'8.Q,

Cl!

•acC4

OO

c3o

C/5

SO

cS,a

"u• a

ccd0 0

c3O

in

o

5 IIo"Soc

s

15"2'S.o g.

s

3

II

| I'll*E & *-* *—

gsl-s••2 3 - 5• a ' ; - 1 1

R s s 1 15 s § § " -PDOCJ o

CO UCJ


fN rN rN rN rN rN rNc i rN rN c"i c^ c i c i

rN fN <N fN fN TJ- Tf

8 88 88 88fN rN rN rN* rN ^fr -^-ci ci ci rN —• rN —'

(N <Nci ci

rN fN rN rNCI CI CI CI

> O O O _ . . .> p p p — — co coi ci ci ci co co

ci ciO O , _O O — —" " oo oo

ci m

ci cici ci — m

88 888ci ci rn ci ci

88 888 r-j f N r~-i rsi

c m m ci

fN rN rN fN fN fN cN

rN fN fN fN rN rN fNCI —— C I CI CI CI CIfsj — — rN rN rN rN

(N" fN"rN" fN~-^" f N " ^

CNCNfN <NM fNfN t^JfN fN fN

fNfNfN fN fN fNfN fNfN fNfNcVc^c^ cnm" cVm mm mmfNfNCN (NfN fNfN fNfN r-lfNfNfNfN fNfN rNfN fNfN fNfN

fN fS(N fN fNfN fNfNfNfNfNfNfN — fN — fNfNm m m m c~ici m m mfNfNfNfN fNfN (NfNfNCifNCIfN fNfN fNfNfN

fN N tN (N fN — CI rN fN fNrNfNfNfN<N CH rN fN rN rNci ci ci ci ci ci ci ci ci ci

rNrNfNfNrN — c i <N rN rN(N* rT CN" rN" fN~ rT rN CN" fN* rN

fN fN fN fNfN fNci m

fN CN fN fNm m c i r*i

fN fNm m

I —. - ^ - ^ fN fN fN fN

—• fN —« — fN — fN

fN fN fN— fN fN rN fN rN fN fN fN <N fN

8 88 88 88S.S. 5L5L S.5Loo oo oo" do o"oo"<oo oo oo oo pop<oo oofNfN rNfN

oo oooo88

o o oo d doo fN r j q

d o —' o ( N O O' fN—d

CI;CI Od —

CI—

Ooo—'<

rNrNd

— • ci—'•

ootN [ o o

odd—'d do

D. Q.

T3 -OC C

OO 00B C

3 3O O

C/3 C/2

E2inC

"2

a.o.

U

ooc

1o2aa

0ic

'1oc

rai

oa• a

EoC

"2oa

a

ioa:

o oV3 C/5

I

(2

I

. 1 ^^ P s8 =

I"P.

Outs.

ifII

•g,CTv

? «_.

u - -t* w. «E=a

5 uass

ill§

( 0 >•

OT3333^-CS"-CS

; U

oo a a

135c J < 3 3 3 3 3 5 x :| w £ 2 2 2 2 j . 2 P•§, OOOOOo

SOCIAL LOAFING 691

•3Sa,

CN CN fN CN

88 88 888888 888CN CN tTf CN CN CN CN CN CN* CN CN CN

| | | | I l l l l l HICNCN C N ^ f CN CN CN* CN CN CN CN CN CN*CNCN m m CNCNCNCNCNCN m r n m

CN CNm m

88

II

CN CNm m

88CN CN CN CN

8888

IIm rn m m

8888

888^8.0

CN m CN cnm m m m m m CN rn m CN

CN CNm m I!

888odo— , f — — — —.

ICNCN CNCN CNCNfNCNCNCN ^ - _ —com mm -^ — -^mmm mmc^CNCN M M ^^ — -^ — —. —. CNfNCNCNTCN" CNCN" CN"CN"CN"CN"CN"?T m"mm"

CNfNCN(NmmCNCNCN"CN"

CNCNCNCNmmCNCN

CNCNCNCN

mmmm(NCNCNCN

CN CN CN CN CN CNC-J CN CN —* CN CN

CN CN CN CN CN CNCN CN CN m CN CN

CN CN CN CN CN CNCN CN CN CN CN CNmi*i m m r^ en—' m CN CN CN <NCN CN CN CN CN CN

CN CN CN CN CvlCN CN CN CN CNfir^m mm

CN (N CN CN CN <N

"oo

— — CNCN — « —< - —< — _ ^ _ ^ -CNCN —' -^ CNCNCNCNCNCN CNCNCN—•CN —' —- CN CN CN — CN -^ — —* CN

- CN CN CN CNCN fNCN CN

CN CNCN (N

CN CN CN —' —CN CN CN CN CN

|| || Illlll H I |i 8| 8888 888888 88 88 88 888 8CN CN CN CN CN CN

O O oooo oo oo oo oo•* •* CNCN O O O O

CN CN CN CN CN CN

do doC N r jCN —• ^O— — d o — o o — — — O O CN

J

O &)O O

OO -^ O

p q r-#odd

O oo

d o

a

ao

EoC

5

c•5. a.

E3D.

ooc

c

atio

form

co

00

[uno:

« u3 OD. C

00 OOc cE Eo o

c c •3

3XI•c

•a a1

•I .sW BJ W

1s

22

33

P Pn

ID,

o

E

I 2&c WEPc 00

UJ O

" " CN m .B. N . N II || "if

Is'aI*ScN-'f2-

So

(986

Q

1985

D —

1-2

1980

)

"—'

Lat

an

CO

E


— —• m r

88i00"p p

— —- — CO

888 3d 6

88M M

88

M M Mm m m

(N (N NrifN

88 SI

M M

O ^

M —(N M

S88O 'O O O

^ C ^ C1CL CiClTf i" mm mm mm — m

M<N — — <m m m m

CNMtN (NfNM

N ( N (N Ol (N(N (N m rO —«

^ — —> mm

oo P9P^ O O O OQ _ ' M M M—' rsl ^ ^ ^OMM M M M

8 ooo 88

— M ^f T

^2 88 8888 00000cO O O O O <•

O tN

o o

. - \O O— ^- o— o —'•1 1

po

of~i Oo —'• O O O O— —O O — OO(N(N _- — c

aa

3O

C/3

OD

EE

00

00

g

sto

c

Bra

ox:00

S 00

11

(2

- ., oE ^

OO o u

•p > .c•S'3i oM U U

M O O

00

MIL• ^ s

ON

,2 " " | |

•a

01! i n —

o •^'5-S--^

•BE.I

on

- - c^7T^'

ss'§§ f f i i i

S !5 § ^ ^5^£SW

'•S " ^ *?2 **3 ^ ^ M 1> U C

1 K K C M n! •"* ' " " ^

iiiiiifSIIfl~,m

SOCIAL LOAFING 693

I(5

o

IV

I

H

o

XIC/3

fg

£

T3Ct3

Si

iS8

§8

88

88

O O Q OOOOO

8888

o ooo

88

0 0m o - - —< <N

8S 888 88f 4 (N (Nf i o rn

ci ci

r4 (N (N fN (N

• • * ~** ^ ^ . ^ ^ » ^ ^ » ~»»». ^ ^ » ^^ fc ^ * *

8 88 8888 81OOO

r*1 — (N

' 0 0 ~'• o

OOOO

OOOO

O O OO O vi

ing

E£

ooc

aa*OCed00c•£3

Sho

rcl

'o

cI

Dra

ooc

ppi

cooc

ox:en

OD

X it/53cxcO

3

II

H u i> rt

: S "

= = 5 S rt-

9 ^ g S'E.S «<gIII «-<>

O

: E i-s g

3 c c c-•3 E E !'•

c cE E

H I , 1^= OJ LU BJ g UJ W

"IO

I•ac

•o

g

copVIu"O

blei

sri

a

>

thir

d

tt

oc3L*O

c,o

ict

"O

D.OcII

o"cII

<N<A

m<LT

'2%O

IIfN

V)

1'3u

3

•o

linin

gco

:

^-

fj

E• ^VJDrt

va:

o0>

In&)a

.52ia

_ ca

cI -

"2

op

II

ce(l

=le

nva

:ta

sk

sca

o

u

5

r);le

ai

o

o

c

1!

rt

COS)

rtao

>i

d

ot J j

Q

U

a>

a4>

BI.

>

irtt

n

eci

f"5

ivid

u

•g

cSJ2

Xi

rt>

* -

set

o

ntat

s

1 U

i

oner

f

rt

5u

^&cc

1

>-a!5

o o12 .5

(N « •

5 o

2§

ifiHiig x>

ill mum'.lllilll

; .« . „ _i >


ANOVA) and a test of the homogeneity of the effect sizes within eachclass, Qm. The continuous models, which are least squares linear re-gressions calculated with each effect size weighted by the reciprocal ofits variance, provide a test of the significance of the predictor as well asa test of model specification (Hedges & Olkin, 1985). Finally, as asupplementary procedure, we conducted outlier analyses to determinewhether the effect sizes were relatively homogeneous aside from thepresence of a limited number of aberrant values (see Hedges, 1987;Hedges & Olkin, 1985).

Results and Discussion

Characteristics of Studies

Before considering the findings of the social loafing litera-ture, we first present the characteristics of the studies fromwhich our conclusions will be drawn. As shown by the central

tendencies of the characteristics presented in Table 2, studiesgenerally (a) were published recently, (b) were published as jour-nal articles, (c) used a moderate number of subjects, (d) usedcollege-age samples, and (e) only occasionally included organiza-tional or nonadult samples. In addition, studies (a) were gener-ally laboratory experiments with some field studies, (b) usedboth within- and between-subjects comparisons of individuals'effort across coactive and collective conditions, and (c) typicallyinformed subjects that the purpose of the experiment was toexamine effort or task performance. Regarding task character-istics, studies (a) typically used simple tasks, (b) tended to mea-sure either physical or cognitive effort with some studies ofperceptual or evaluative effort, and (c) generally used maximiz-ing, additive tasks.

Finally, regarding key theory-relevant predictors, the last 10variables in Table 1 show that studies generally (a) varied evalua-

Table 2Summary of Study Characteristics

Variable and class

Median date of publicationPublication form

Journal articleMeeting paperDissertation or master's thesisUnpublished document

Median number of subjectsMedian age of subjects (years)Status of subjects

Third grade students or youngerFourth through sixth grade studentsJunior high or high school studentsCollege studentsOrganizational employees

Setting of studyLaboratoryField

Type of coactive-collective comparisonWithin-subjectsBetween-subjects

Type of cover story usedEffort or performance relatedOther

Complexity of taskSimpleComplexUnknown or unclear

Type of effort measuredPhysicalCognitivePerceptualEvaluative

Quantity or quality emphasis of taskMaximizingOptimizingMixed or unclear

Method of combining individual inputsAdditiveCompensatoryOther, mixed, or unclear

Value

1985

12123148

4820

656

1436

14224

6997

11551

1357

24

70532914

127309

147118

Variable and class

Conditions under which individual outputscan be evaluated

NoneCoactive condition onlyCoactive and collective conditionsUnclear

Task valenceHighModerate or no basis for determiningLow

Group valenceHighModerateLowUnknown or unclear

Opportunity for group evaluationPresentAbsentExpectations of co-worker performanceHighModerate or no basis for determiningLowUniqueness of individual contributions

to collective outcomeUniquePotentially redundantCompletely redundant

Median group sizeMedian number of task performerspresent at session

Sex of subjects3

MenWomenBoth

Culture of subjectsEasternWestern

Value

51172816

30131

5

911

12917

12151

14144

8

86494

3

3

2812

126

17148

Note. For categorical variables, numbers in table represent frequency of effort comparisons in each class. Summaries of continuous variables arebased on reports for which information was available on each variable.* When studies were subdivided on sex of subjects, the frequencies were 40 for men, 24 for women, and 114 for both.

SOCIAL LOAFING 695

tion potential across the coactive and collective conditions, (b)did not manipulate task valence, (c) studied low-valence groupsof strangers, (d) did not provide an opportunity for group evalua-tion, (e) did not manipulate expectations of co-worker perfor-mance, (f) used tasks in which individuals' inputs to the collec-tive product were either potentially redundant or completelyredundant with the inputs of other group members, (g) had arelatively small median group size, (h) tended to have all taskperformers physically present at the experimental session, (i)typically included both male and female subjects, and (j) gener-ally examined subjects from Western cultures.

Overall Magnitude of Social Loafing

The summary provided in Table 3 allows one to determinewhether, on the whole, individuals tended to engage in socialloafing and exert less effort collectively than coactively. Anoverall difference in effort across coactive and collective condi-tions is shown by a mean effect size that differed significantlyfrom the 0.00 value that indicates exactly no difference (i.e., bya confidence interval that does not include 0.00). Consistentwith the CEM and with prior reviewers' conclusions, individ-uals tended to engage in social loafing, producing lower effortlevels in the collective condition than in the coactive condition.The weighted mean was very similar to the unweighted mean,whereas the median was somewhat lower. As shown by thehomogeneity statistic given in Table 3, the effect sizes were nothomogeneous (i.e., consistent) across the studies. A very largeproportion (39%) of the effect sizes had to be removed to attainhomogeneity, and the removal of outliers substantially de-creased the value of the weighted mean. Inspection of the out-

Table 3Summary of Effect Sizes (ds)

Criterion

ds with outliers

Sample size («)Mean weighted d {d+f95% CI for d+Homogeneity (Q) ofds making up d±b

Mean unweighted d95% CI for mean unweighted dMedian dDifferences indicating social loafing

ds excluding outliers

Sample size (n)n removed outliers0

Mean weighted d (d+)95% CI for d+Homogeneity (Q) of ds making up d+

Value

1630.44

0.39/0.48964.70**

0.480.36/0.61

0.41128/163 (.79)

9964 (.39)

0.240.19/0.29125.79

Note. Positive effect sizes indicate that subjects worked harder coac-tively than collectively (i.e., engaged in social loafing). CI = confidenceinterval; d= effect size; d+ = mean weighted effect size; Q = homogene-ity of ds." Effect sizes were weighted by the reciprocal of the variance. b Signi-ficance indicates rejection of the hypothesis of homogeneity. c Theproportion appears in parentheses.**;?<.001.

Hers revealed that among the 64 effect sizes removed, 49 wereabove and only 15 were below the median d computed for theoriginal sample. The confidence interval for the weighted meanafter outlier removal showed that the overall tendency towarddecreased collective effort was still significant after outlierswere removed.

The implication of the finding that people tend to engage insocial loafing when working collectively depends in part on itsmagnitude. The several measures of central tendency reportedin Table 3 are relevant to this issue. Taken together, these fig-ures indicate approximately a four-tenths standard deviation inthe direction of reduced collective effort. One means for inter-preting the magnitude of the mean social loafing effect size is tocompare it with effect sizes produced by other quantitative re-views in similar domains. Eagly's (1987) overview of a variety ofmean effect sizes found in meta-analyses of social psychologi-cal research topics suggested that they ranged from about 0.00to 1.20. In a particularly relevant comparison, Bond and Titus's(1983) review of the social facilitation literature reported thateffect sizes ranged between -0.23 and 0.21, depending on thecomplexity of the task and the type of performance measure.Another standard is provided by Cohen (1977), who suggestedthat ds of 0.20 be labeled as small, ds of 0.50 be labeled asmoderate, and ds of 0.80 be labeled as large. On the basis ofthese standards, social loafing might be regarded as small tomoderate in magnitude, within the middle range of effects inthe domain of social behavior, and larger than social facilita-tion.

However, it should be recognized that the tendency of re-searchers to focus on isolating conditions under which socialloafing can be reduced or eliminated is likely to lead to anunderestimation of the magnitude of the effect. Most studieswere designed to test the potential of one or more factors foreliminating social loafing. Therefore, many of the coactive-col-lective comparisons included in our meta-analysis were drawnfrom situations in which social loafing was not predicted. Thus,the mean effect size of d= 0.44 may not necessarily provide thebest measure of the overall magnitude of social loafing. Indeed,a categorical model computed for whether social loafing waspredicted by the original researcher(s) was highly significant,<2B(2) = 310.86, p < .001. Effect sizes were substantially larger(4+. = 0.76, 95% CI = 0.70 to 0.81, n = 89) when social loafingwas predicted than when social loafing was not predicted (d* =-0.01, 95% CI = -0.08 to 0.06, n = 59, contrast p < .001) andwhen researchers did not specify a prediction (d^ = 0.19, 95%CI = 0.09 to 0.29, n = 15, post-hoc contrast p < .005).5 The"predicted" and "not predicted" values may represent esti-mates of upper and lower bounds for the magnitude of socialloafing. In view of these comparisons, the available researchseems to suggest that there is at least a moderate tendency forindividuals to engage in social loafing and to reduce their effortwhen working on collective tasks.

Finally, with regard to the robustness of social loafing, an

5 Contrasts were a priori, unless otherwise indicated. The post-hoccontrast procedure is modeled after Scheffe's method for multiple com-parisons in an analysis of variance (see Hedges & Olkin, 1985).


examination of the consistency of findings across studies isworthwhile. The direction of the effect was very consistent,evidenced by 79% of the comparisons being in the direction ofreduced collective effort. However, outlier analyses showed thatthe effect sizes were highly inconsistent in magnitude: 39% ofthese effect sizes had to be removed to attain homogeneity. Thisfigure is substantially larger than that reported in other meta-analyses of social behavior. It is possible that the focus of socialloafing research on limiting conditions, combined with thewide range of tasks and subject populations used, contributedto this relative inconsistency. It is also possible that this incon-sistency reflects the complex nature of the phenomenon andsuggests that social loafing may represent a class of effects withpsychologically distinct causal mechanisms.

Theory-Relevant Predictors of Social Loafing

The study attributes were examined as predictors of differ-ences in individuals' coactive versus collective effort. Categori-cal models that yielded significant between-classes effects arepresented in Table 4, and significant continuous models arepresented in Table 5.

Evaluation potential. Consistent with the CEM and withevaluation accounts, the tendency for individuals to engage insocial loafing decreased when evaluation potential was heldconstant across the coactive and collective conditions. Thus,individuals were more likely to loaf when their outputs couldonly be evaluated in the coactive condition than when theiroutputs could not be evaluated at all (p < .001) or when theiroutputs could be evaluated in both the coactive and collectiveconditions (p < .001). In fact, social loafing was eliminatedwhen evaluation potential was not varied across coactive andcollective conditions, as evidenced by the confidence intervalspresented in Table 4. These findings are consistent with theclaim made by Harkins (1987) and others that evaluation po-tential is a mediator of social loafing.

Task valence. As hypothesized, the tendency to engage insocial loafing decreased as task valence increased. Contrastsshowed that the tendency to loaf was greater when task valencewas low than when task valence was either high (p < .001) orunspecified (p < .06, post-hoc contrast). Also, the tendency toloaf was greater when task valence was unspecified than whentask valence was high (p < .001, post-hoc contrast). An examina-tion of confidence intervals shows that individuals loafed whentask valence was low or unspecified, but not when it was high.This suggests that high levels of task meaningfulness or per-sonal involvement might eliminate loafing. Also, given that theunspecified tasks were likely relatively moderate in valence,these results suggest a direct relationship between task valenceand individual effort on collective tasks.

Group valence and group-level comparison standards.Individuals did not loaf when group valence was high, butloafed in all other conditions. Thus, as predicted by the CEMand prior theories of self-validation processes in groups, thetendency to loaf was lower when group valence was high thanwhen it was moderate (p < .01), unknown or unclear (p < .05,post-hoc contrast), or low (p < .001). Individuals also loafed less

when group valence was either moderate (p < .001) or un-known (p < .005, post-hoc contrast) than when it was low. Fi-nally, individuals loafed less when group valence was high thanwhen it was low (p < .001). Similarly, the tendency to engage insocial loafing was reduced when participants were providedwith a group-level comparison standard. Taken together, thesefindings suggest that enhancing group cohesiveness or groupidentity might reduce or eliminate social loafing and thatgroup-level outcomes are indeed valued by individuals in cer-tain situations.

Expectations of co-worker performance. Consistent with thepredictions of the CEM, individuals loafed when they expectedtheir co-workers to perform well or when no expectations wereprovided, but did not loaf when they expected their co-workersto perform poorly (low vs. high contrast, p < .005; low vs.unspecified contrast, p < .001, post hoc). These findings, how-ever, do not completely resolve the controversy between thematching of effort (Jackson & Harkins, 1985) and social com-pensation (Williams & Karau, 1991) views. The basic patternof findings is consistent with both the CEM and the socialcompensation research, which both suggest that individuals'outcomes are more contingent on their efforts when workingwith others who are expected to perform poorly. However, ef-fort levels were not significantly higher collectively than coacti-vely when participants expected their co-workers to performpoorly, as has been found in the prior social compensationwork.

Nevertheless, a resolution is possible when the role of taskmeaningfulness is considered. Because social compensationrepresents a motivation gain, individuals must not only in-crease their usual collective effort, they must actually workharder than they would ordinarily work alone. The logic of theCEM suggests that individuals would be unlikely to exert suchextraordinary effort unless they viewed the task as meaningfuland therefore highly valued a favorable outcome. In support ofthis notion, when the interaction between expectations of co-worker performance and task meaningfulness was considered,a significant motivation gain was found only for the four com-parisons in which individuals worked on a meaningful taskwith co-workers who were expected to perform poorly (di+ =-0.69, 95% CI = -1.11 to -0.27). These results highlight thepotential of the CEM for predicting situations in which motiva-tion gains might emerge on collective tasks.

Uniqueness of individual inputs. Individuals worked just ashard collectively as coactively when their individual inputs tothe collective product were unique, but loafed when their inputswere either potentially redundant or completely redundant(ps < .001). Consistent with the CEM and with the results ofKerr(1983; Kerr& Bruun, 1983) and Harkins and Petty (1982),individuals may perceive that their efforts are less instrumentalin obtaining valued outcomes when their inputs to the collec-tive product are either dispensable or largely overlapping withthose of others.6

6 Although uniqueness and dispensability are often closely related,they are conceptually distinct. A group member's inputs can be indis-tinguishable from the inputs of others, yet essential to group perfor-mance. For example, on an additive rope-pulling task, group members'

SOCIAL LOAFING 697

Group size. Simple linear regressions showed that bothgroup size and the number of task performers present werepositively related to social loafing: Effect sizes were larger forstudies that combined more individual inputs into the collec-tive product and for studies that had more individuals performthe task at each session. Most of the theories discussed in theintroduction would predict group size effects, albeit for differ-ent reasons. Social impact theory and self-attention theory de-vote particular attention to group size effects relative to otheraspects of social loafing.

Gender and culture. Consistent with the CEM, the magni-tude of social loafing was larger in studies that used samples ofonly male subjects than for studies that used either mixed sam-ples (p < .05, post-hoc contrast) or samples of only female sub-jects (p < .001). If this difference were indeed attributable to agreater tendency by men to focus on individualistic outcomes,the CEM would predict that men would be more attentive tostrategic concerns than women when working on collectivetasks. Consistent with this reasoning, the group size effect re-ported earlier was only reliable for comparisons involvinggroups of men (b = .13, b* = .36, p < .001). This finding sug-gests that men may be particularly attentive to the reducedefficacy of their inputs when working in larger groups, whereaswomen may tend to exert more consistent levels of collectiveeffort across various group sizes.

Also as predicted, the magnitude of social loafing was largerfor subjects from Western cultures than for subjects from East-ern cultures. These provocative findings suggest that bothwomen and individuals in Eastern cultures are less likely toengage in social loafing, presumably because they have moregroup-oriented priorities than do men and individuals in West-ern cultures.

Task complexity. Individuals performed better coactivelythan collectively when working on simple tasks, but performedat least as well collectively as coactively when working on com-plex tasks (p < .001). This pattern is consistent with the reason-ing presented by Jackson and Williams (1985) and providesinitial, tentative support for the arousal-reduction viewpoint.7

Models Relevant to Evaluating the Generality of SocialLoafing

The status of subjects produced a significant between-classeseffect. Contrasts showed that third graders and younger stu-dents tended to loaf less than either fourth through sixthgraders (p < .05, post-hoc contrast) or college students (p <.005, post-hoc contrast). In addition, adult nonstudents tendedto loaf less than fourth through sixth graders (p < .01, post-hoccontrast), or college students (p < .01, post-hoc contrast). It ispossible that the prior finding may be indicative of a develop-mental trend whereby children do not become attentive to stra-tegic concerns in task performance until they reach a certainage (see Williams & Williams, 1981, for a discussion). The lattertrend is relatively uninterpretable, because the six comparisonsbased on adult nonstudents involved diverse samples.8 The

responses are completely redundant in form, but the effort of eachmember still contributes to the total group performance.

magnitude of social loafing was also larger in laboratory experi-ments than in field studies. Interestingly, the magnitude of theeffect did not vary depending on whether (a) the coactive-col-lective comparison was made within-subjects or between-sub-jects. This finding suggests that social loafing does not occursimply because subjects allocate greater effort to coactive trialsthan to collective trials (see Harkins, Latane, & Williams, 1980;Kerr & Bruun, 1981). Finally, the magnitude of social loafingwas not dependent on whether (a) maximizing or optimizingtasks were used, (b) tasks were additive or compensatory, and (c)physical, cognitive, evaluative, or perceptual effort was mea-sured, suggesting that social loafing is robust across tasks.

Impact of Other Variables on Social Loafing

The trend toward decreased collective effort was enhancedwhen a deceptive cover story was used. Thus, subjects weremore likely to loaf when they were not informed that the pur-pose of the study was to examine effort or performance, sug-gesting that telling subjects the purpose of the experiment mayinhibit their usual tendency to engage in social loafing. Socialloafing was also greater for studies reported in journal articlesthan for dissertations and master's theses or for unpublisheddocuments (ps < .05, post-hoc contrasts). Finally, as shown inTable 5, a simple linear regression showed that date of publica-tion was negatively related to the magnitude of social loafingsuch that effect sizes were larger for earlier studies. However,interpretation of this time trend is clouded when it is recog-nized that the earlier studies both (a) tended to use fewer experi-mental paradigms that were more easily controlled, possiblyresulting in smaller error terms and larger effect sizes and (b)tended to focus on documenting the effect rather than isolatingmoderating variables and were thus more likely to find largereffect sizes across all comparisons.

Multiple Regression Models

Although most of the theory-relevant predictor variableswere relatively independent (rs ranging from —.14 to .33), weestimated several multiple regression models to examine the

7 However, individuals did not perform significantly better collec-tively than coactively when working on complex tasks, as arousal re-duction would predict. Yet, only seven comparisons involving complextasks were available, creating low power to detect a difference. More-over, four of the studies coded as using a complex task used manipula-tions of high task difficulty (Bartis, Szymanski, & Harkins, 1988; Har-kins & Petty, 1982). Although these tasks were not simple, by defini-tion, they may not have been difficult enough to create an inverserelationship between effort and performance. For these studies, themean weighted effect size was -0.09. For the remaining three compari-sons that directly manipulated complexity (Griffith, Fichman, & Mor-eland, 1989; Jackson & Williams, 1985), the mean weighted effect sizewas-0.18.

8 One comparison involved adult residents who worked on a labora-tory brainstorming task (Cooley, 1991), four comparisons involved ei-ther American or Chinese adult managerial trainees working on anin-basket exercise (Earley, 1989), and one comparison involved Japa-nese business managers working on a noise-production task (Wil-liams, Williams, Kawana, & Latane, 1984, Experiment 4).


Table 4Categorical Models for Objective Effort Effect Sizes

Variable and class

Evaluation potential1*None

Coactive condition onlyCoactive and collective conditionsUnclear

Task valenceHighModerate or no basis for determiningLow

Group valenceHighModerateLowUnknown or unclear

Opportunity for group evaluationPresentAbsent

Expectations of co-workerperformance

HighModerateLow

Uniqueness of individualcontributions

UniquePotentially redundantCompletely redundant

Sex of subjects0

MenWomenBoth

Culture of subjectsEasternWestern

Complexity of taskSimpleComplexUnclear

Status of subjectsThird grade students or youngerFourth through sixth grade studentsJunior high or high school studentsCollege studentsOrganizational employees

Setting of studyLaboratoryField

Type of cover story usedEffort or performance relatedOther

Publication formJournal articleMeeting paperDissertation or master's thesisUnpublished document

Between-classeseffect

136.22***

91.52***

45.19***

42.24***

20.54***

16.77***

29.61***

15.44***

21.37***

29.25***

15.39***

10.83***

21.11***

n

5

1152716

30128

5

911

12716

12151

14141

8

86491

3923

113

15148

1337

23

556

1416

14023

11251

11823148

Mean weightedeffect size (di+)

-0.12

0.590.080.32

-0.100.490.90

-0.170.250.500.28

0.030.48

0.330.45

-0.17

0.030.490.44

0.570.220.44

0.190.46

0.47-O.li

0.36

0.140.510.380.47

-0.19

0.470.25

0.390.52

0.500.360.310.05

95%

Lower

-0.33

0.55-0.01

0.20

-0.220.450.57

-0.410.080.460.18

-0.100.44

0.140.41

-0.45

-0.180.410.39

0.480.130.39

0.060.42

0.42-0.37

0.25

-0.020.350.190.42

-0.53

0.430.16

0.340.46

0.450.280.20

-0.22

CI for di+

Upper

0.08

0.640.170.45

0.020.541.22

0.080.420.550.38

0.160.52

0.530.500.10

0.240.560.49

0.650.320.49

0.320.50

0.510.140.47

0.300.680.580.510.15

0.510.35

0.440.59

0.540.450.430.32

Homogeneity withineach class (&J"

8.17

616.34***96.30***

107.68***

12.86***743.32***

2.99

5.48129.91***719.46***64.75***

66.50***855.96***

65.54***853.48***25.14**

3.05355.14***589.75***

261.22***34.66

655.40***

103.96***845.31***

771.08***3.49

168.76***

17.58**24.32***12.13

809.28***72.14***

794.46***154.85***

614.62***339.25***

801.62***52.86***70.79***18.32*

Note. Positive effect sizes indicate that subjects exerted more effort coactively than collectively (i.e., engaged in social loafing). CI = confidenceinterval." Significance indicates rejection of the hypothesis of homogeneity. b Evaluation potential refers to the conditions under which individual outputscan be evaluated. c Calculated on effect sizes that were, whenever possible, partitioned by sex of subject.*p<.05. **p<.01 . ***/?<.001.

SOCIAL LOAFING 699

Table 5Continuous Models for Objective Effort Effect Sizes

Predictor or outcome

Continuous variablesDate of publicationGroup size*Number of task

performers presentb

Categorical variablesPublication formc

Culture of subjects11

Setting of study*Task valencef

Expectations of co-workerperformance8

Evaluation potential11

Additive constantMultiple RSE of estimateGE1

Simple linear regressions

b

-0.02**0.01*

0.02**

b*

-.15.07

.12

Multiple regression

b

-0.01*0.01*

-0.37**0.33**

-0.21**-0.51**

0.54**0.67**

22.19.59

2.09607.41**

b*

-.07.07

-.28.15

-.12-.26

.12

.46

Note. Models are weighted least squares simple linear and multiple regressions calculated with weightsequal to the reciprocal of the variance for each effect size. In the multiple regression model, the predictorswere entered simultaneously. Positive effect sizes indicate higher effort coactively than collectively, b =unstandardized regression coefficient; b* = standardized regression coefficient; n = 163.a Number of individual inputs combined into collective outcome; n = 156. " n = 156. c 0 = journalarticle; 1 = unpublished document. d 0 = Eastern; 1 = Western. "0 = laboratory; 1 = field. f0=low,moderate, or no basis for determining; 1 = high. % 0 = low; 1 = moderate, no basis for determining, orhigh. h Refers to the conditions under which individual inputs can be evaluated; 0 = none, or coactiveand collective conditions (evaluation potential constant across conditions); 1 = coactive only (evaluationpotential confounded across conditions) or unclear. ' Significance indicates model not correctly speci-fied.*p<.0\. **p<.001.

simultaneous impact of several of the continuous and categori-cal variables. For these analyses, categorical variables weredummy coded. The relatively large number of categorical andcontinuous variables that produced significant one-way models(and the presence of three or more classes for several of thecategorical variables) restricted the viability of the multiple re-gression models because the number of potential predictorswas large in relation to the number of effect sizes. Althoughnone of these models was correctly specified, one particularlyinformative model is presented in Table 5. This model entered anumber of the most central predictors—namely, date of publi-cation, group size, publication form, culture of subjects, settingof study, task valence, expectations of co-worker performance,and evaluation potential. All of the predictors in this modelwere significant, with task valence, expectations of co-workerperformance, and evaluation potential producing especiallylarge effects. As reflected in the R of .59, this model was moder-ately successful in accounting for variability in the objectiveeffort effect sizes.

Self-Reported Effort

Although our meta-analysis was primarily concerned withobjective effort measures, effect sizes based on subjects' self-re-

ported effort were also included so that supplementary analysescould be conducted. As shown by the measures of central ten-dency in Table 6, subjects' self-reports were in the direction ofsocial loafing, although the magnitude of this effect was verysmall and was only significantly different from 0.00 when ef-fect sizes were weighted by the reciprocal of the variance andwhen outliers were excluded. Moreover, when nonsignificantreports were considered, the effect became even smaller andwas not significantly different from 0.00. Only slightly morethan half of the reported differences were in the direction ofreduced collective effort. This finding suggests that partici-pants are either unaware of the fact that they engage in socialloafing or are unwilling to report that they engage in socialloafing.9

9 A number of factors had a significant impact on the magnitude ofthe self-reported effort effect sizes. Further information on these analy-ses can be obtained from Steven J. Karau. In general, however, individ-uals were less likely to report that they had engaged in social loafingwhen (a) the cover story informed subjects that the study was interestedin effort or performance, (b) a within-subjects rather than a between-subjects design was used, (c) maximizing rather than additive taskswere used, (d) group size was smaller, and (e) studies were conductedmore recently.


Table 6Summary of Self-Reported Effort Effect Sizes

Criterion Value

ds with outliers

Sample size («)Mean weighted d (d+f95% CI for d+Homogeneity (Q) of ds making up d+

b

Mean unweighted d95% CI for mean unweighted dMedian d

ds excluding outliers

Sample size (n)n removed outliers0

Mean weighted d (d+)95% CI for d+Homogeneity (Q) aids making up d+

All reports

Sample size («)Mean unweighted dDifferences indicating reduced collective effort

280.11

0.00-0.2255.01"0.10

-0.09-0.290.00

2530.16

0.04-0.2733.10

420.07

13/24 (.54)

Note. Positive effect sizes indicate that subjects worked harder coac-tively than collectively (i.e., engaged in social loafing). CI = confidenceinterval; d= effect size; d+= mean weighted effect size; Q= homogene-ity of ds.* Effect sizes were weighted by the reciprocal of the variance. b Sig-nificance indicates rejection of the hypothesis of homogeneity. c Theproportion appears in parentheses.*p<.Q\.

Conclusions

Overview of findings and implications. Our meta-analysisshows that individuals tend to engage in social loafing whenworking in groups. Social loafing appears to be moderate inmagnitude and generalizable across tasks and subject popula-tions. The CEM suggests that social loafing occurs becauseindividuals expect their effort to be less likely to lead to valuedoutcomes when working collectively than when working coacti-vely. The tendency to engage in social loafing was moderated bya large number of variables, and the CEM was quite successfulin accounting for the empirical status of these moderators. Thefindings of this review have a number of implications for futureresearch and practice. In particular, the review highlights anumber of conditions under which social loafing should bemore or less likely to occur. For example, individuals are morelikely to engage in social loafing when their individual outputscannot be evaluated collectively, when working on tasks that areperceived as low in meaningfulness or personal involvement,when a group-level comparison standard is not available, whenworking with strangers, when they expect their co-workers toperform well, and when their inputs to the collective outcomeare redundant with those of other group members. In addition,social loafing is robust across gender, culture, and tasks, al-though the effect is smaller for women and for subjects fromEastern cultures.

A superficial evaluation of these findings might imply that

social loafing is not a very serious problem in the real world,but instead is restricted to artificially constrained laboratorystudies that use trivial tasks. However, a closer examinationreveals that social loafing could indeed have serious conse-quences for a variety of everyday collective settings. First, theeffect was robust across both maximizing and optimizing tasksand across tasks demanding different types of effort (e.g., cogni-tive, physical, or perceptual). Thus, social loafing is not re-stricted to additive, maximizing tasks. Second, although themagnitude of the effect was reduced for field studies, forwomen, and for subjects in Eastern cultures, social loafing wasstill significant under all of these conditions. Thus, the effect isnot restricted to laboratory studies of American men. Third,although social loafing was eliminated when participants'scores could be evaluated collectively and when highly mean-ingful tasks were used, it is important to recognize that, inmany real-world contexts, individual inputs cannot be reliablyidentified or evaluated and people are frequently asked to workon mundane or uninspiring tasks. Even within occupationsthat appear to have a great deal of intrinsic value, at least someimportant aspects of the work are likely to be repetitive or dull(e.g., writing memos or performing accounting and other main-tenance activities). Finally, although social loafing was elimi-nated when participants worked with close friends or team-mates, mere acquaintance with one's co-workers was not suffi-cient to eliminate the effect. In many (but not all) workcontexts, individuals do not necessarily know their co-workersor do not interact with them frequently enough to develop highlevels of cohesiveness. Thus, social loafing appears to be a prob-lem that is likely to manifest itself in a variety of settings.

The tendency of social loafing research (but not theory) tofocus on limiting conditions for the effect, rather than on un-derlying process, may have led inevitably to the erroneous viewthat social loafing is inconsequential. Nevertheless, this ten-dency to focus on moderating variables does have the positiveconsequence of suggesting a number of ways in which socialloafing might be reduced or overcome in natural settings: Pro-viding individuals with feedback about their own performanceor the performance of their work group, monitoring individualperformance or making such performance identifiable, assign-ing meaningful tasks, making tasks unique such that individ-uals feel more responsibility for their work, enhancing the cohe-siveness of work groups, and making individuals feel that theircontributions to the task are necessary and not irrelevant mightall serve, under some conditions, to reduce or eliminate socialloafing. Moreover, the specification of these moderating vari-ables has theoretical as well as practical importance. Althougheach moderating variable may be associated with a distinct setof psychological processes, the CEM suggests that all of thesevariables have one thing in common, that is, they influenceindividuals' perceptions of either the instrumentality of theirefforts for obtaining valued outcomes or the degree to whichoutcomes in a particular setting are likely to be valued.

Summary of novel findings. Because this meta-analysis in-tegrates a rich research literature in which several hypothesesare well-established, some of our findings confirm long-stand-ing predictions. However, our meta-analysis also uncovered anumber of novel findings, some of which are directly relevant toareas of controversy among prior researchers. In particular,

SOCIAL LOAFING 701

only a handful of prior studies have directly examined gender,culture, expectation of co-worker performance, and group va-lence, and those studies have produced contradictory results.However, our meta-analysis revealed that (a) the relative magni-tude of social loafing was reduced for women and for subjectsfrom Eastern cultures, (b) individuals were more likely to loafwhen they expected their co-workers to perform well, and (c)social loafing was reduced when individuals worked with ac-quaintances and was eliminated when individuals worked inhighly valued groups. Similarly, although group-level compari-son standards, task meaningfulness, and task uniqueness haveonly been directly examined in a small handful of studies, mak-ing previous conclusions about these factors tentative, we foundstrong, consistent effects for these variables across studies. Themeta-analysis also identified some key methodological factorsinfluencing the magnitude of social loafing, including the typeof cover story used and the setting of the study (laboratoryversus field). In addition, despite mixed results from a few priorstudies, we found that individuals were generally either unwill-ing or unable to acknowledge that they loafed on collectivetasks. Our findings also highlighted the potential of the CEMfor generating predictions about interactions among moderat-ing variables. For example, group size interacted with gendersuch that the magnitude of social loafing increased in largergroups for men, whereas group size did not influence the mag-nitude of social loafing for women.

With regard to prior hypotheses, our meta-analysis confirmsthe conclusion of many previous researchers that individualstend to engage in social loafing when working collectively andreveals that the effect is consistently obtained and moderate inmagnitude. Our findings also support prior conclusions thatindividuals are more likely to loaf when the potential for evalua-tion is absent. In summary, our meta-analysis contributed anumber of new findings and also showed that several previouslyestablished findings are reliable. In addition, the CEM success-fully accounted for the empirical status of the vast majority ofmoderators of social loafing, both old and new.

Future directions. Future research could profitably proceedat a number of levels. An especially pressing need is to furthersubstantiate social loafing outside of the laboratory in real-world settings such as industry, committees, juries, partner-ships, and interpersonal relationships. Once the effect has beendocumented in such settings, attempts could be made to deter-mine whether the moderating variables isolated in laboratorystudies apply to other contexts as well. Additional researchcould further address the generalizability of social loafing,both in terms of dispositional and situational factors. Amongthe questions that might be more directly addressed are thefollowing: What are the long-term effects of working collec-tively? Do people (in general, or particular subclasses) prefertasks on which they can loaf? Is social loafing adaptive in somecases? Does loafing allow individuals to devote their energiesmore fully to aspects of tasks that are within their particulardomains of mastery and expertise? How might social loafingmanifest itself within interacting groups such as decision-mak-ing teams or juries, and how might individual efforts be mea-sured in these contexts? Is it necessarily the case that the qualityof the work suffers when people loaf? How aware are people of

their social loafing in different settings? Does knowledge thatone loafs in collectives affect the likelihood of loafing?

The CEM generates a number of additional predictionsabout how various factors might interact to determine effort.Unfortunately, these predictions could not be examined in thecurrent data set because of an insufficient number of compari-sons (or none at all) in crucial comparison cells. For example,the CEM suggests that making individual inputs identifiable orproviding a comparison standard or performance feedbackshould have a stronger impact when this information has directimplications for the individual's self-evaluation. Thus, the ef-fects of identifiability of inputs, availability of comparisonstandards, and presence of feedback may interact with factorssuch as personal relevance of the task, task meaningfulness,perceived importance of the task to significant others or toreference groups, and amount of identification with the group.

Similarly, the model suggests that the effects of group size oncollective effort may depend on factors such as relative atten-tion to strategic versus cooperative outcomes, task type, andidentifiability. In particular, whereas increased group sizemight increase the magnitude of social loafing when individ-uals place greater value on strategic considerations (due per-haps to individual differences, cultural or situational norms,importance of the group to the individual, or competition),group size might actually decrease social loafing when individ-uals attach greater value to collective outcomes (as might occurwhen working in highly cohesive groups, or when a strong so-cial desirability or cooperation norm is salient). Increases ingroup size are likely to lead to decreased perceptions that one'sinputs are related to one's outcomes on most additive, compen-satory, and discretionary tasks. However, they may actually in-crease such perceptions on disjunctive (or conjunctive) tasks forhigh ability (or low ability) group members (cf. Kerr & Bruun,1983). In addition, when inputs are identifiable, increases ingroup size might enhance rather than inhibit performanceunder many conditions.

The CEM generates similar predictions regarding possibleinteractions among additional factors such as expectations ofco-worker performance (see Williams & Karau, 1991), groupvalence, and task valence. Among some of the more intriguingimplications of the CEM are the following:

1. Identifiability should enhance effort on superficiallyaversive tasks that are nevertheless important, but should beless likely to enhance effort, and may even undermine effort,on intrinsically meaningful tasks.

2. Gender, culture, and individual differences in collecti-vism should influence the degree to which individual outcomesare valued relative to group outcomes. Thus, men, individualsin Western cultures, and individuals low in collectivism andneed for belonging should attach greater importance to out-comes such as pay, competitive intragroup recognition, andothers' evaluations of one's individual performance. However,women, individuals in Eastern cultures, and individuals high incollectivism and need for belonging should be more likely toattach at least moderate importance to outcomes such as groupharmony, group success, satisfaction of other group members,others' evaluations of the group's performance, and other groupmembers' evaluations of one's individual contributions to thegroup.


3. Even if outcomes are highly valued, high levels of effortare unlikely when individual behaviors are not instrumental inobtaining those outcomes. For example, when individual ef-forts are dispensable, gender, culture, and individual differ-ences in collectivism and need for belonging should have lessimpact.

4. Communication among group members should enhancecollective effort when it enhances perceptions of task impor-tance or social responsibility, but should hinder collective effortwhen it relays negative task attitudes or contributes to feelingsof dispensability.

5. Group structural factors and member roles may pro-foundly influence collective effort in ongoing groups by affect-ing perceptions of the instrumentality of one's inputs and thevalue of various outcomes. For example, leaders and high-sta-tus group members may view their inputs as more instrumentalto group outcomes, and norms encouraging social responsibil-ity and hard work within groups should have a positive effect oncollective effort, especially in cohesive or highly valued groups.These are just a few of the questions that future research canattempt to answer. The research reviewed in this article sup-plies a consistent and robust foundation from which to con-tinue this investigation.

References

Abrams, D., & Hogg, M. (1990). Social identity theory: Constructive andcritical advances. New York: Springer-Verlag.

Anderson, L. R., & Blanchard, P. N. (1982). Sex differences in task andsocial-emotional behavior. Basic and Applied Social Psychology, 3,109-139.

Bakan, D. (1966). The duality of human existence: An essay on psychol-ogy and religion. Chicago: Rand McNally.

Bartis, S., Szymanski, K., & Harkins, S. G. (1988). Evaluation and per-formance: A two-edged knife. Personality and Social PsychologyBulletin, 14, 242-251.

Bond, C. E, & Titus, L. J. (1983). Social facilitation: A meta-analysis of241 studies. Psychological Bulletin, 94, 265-292.

Breckler, S. J., & Greenwald, A. G. (1986). Motivational facets of theself. In R. M. Sorrentino & E. T. Higgins (Eds.), Handbook of motiva-tion and cognition (Vol. 1, pp. 145-164). New York: Guilford Press.

Brewer, M. B. (1991). The social self: On being the same and differentat the same time. Personality and Social Psychology Bulletin, 17,475-482.

Brickner, M. A., Harkins, S. G, & Ostrom, T. M. (1986). Effects ofpersonal involvement: Thought-provoking implications for socialloafing. Journal of Personality and Social Psychology, 51, 763-769.

Brickner, M. A., & Wingard, K. (1988). Social identity and self-motiva-tion: Improvements in productivity. Unpublished manuscript. Akron,OH: University of Akron.

Broverman, I. K., Vogel, S. R., Broverman, D. M., Clarkson, F. E., &Rosenkrantz, P. S. (1972). Sex-role stereotypes: A current appraisal.Journal of Social Issues, 25(2), 59-78.

Brown, J. D., Collins, R. L., & Schmidt, G. W. (1988). Self-esteem anddirect versus indirect formsof self-enhancement. Journal of Personal-ity and Social Psychology, 55, 445-453.

Caporael, L., Dawes, R., Orbell, J., & van de Kragt, A. (1989). Selfish-ness examined: Cooperation in the absence of egoistic incentives.Behavioral and Brain Sciences, 12, 683-699.

Carli, L. L. (1982). Are women more social and men more task oriented?A meta-analytic review of sex differences in group interaction, reward

allocation, coalition formation, and cooperation in the prisoners di-lemma game. Unpublished manuscript.

Cates, K. H. (1986). The effects of evaluation, identifiability, and taskuniqueness on task performance. Dissertation Abstracts Interna-tional, 47,1326B. (University Microfilms No. 86-11258).

Cialdini, R. B., Borden, R. J., Thorne, A., Walker, M. R., Freeman, S., &Sloan, L. R. (1976). Basking in reflected glory: Three (football) fieldstudies. Journal of Personality and Social Psychology, 34, 366-375.

Cohen, C. J. (1988). Social loafing and personality: The effects of indi-vidual differences on collective performance (Doctoral disserta-tion, Fordham University). Dissertation Abstracts International, 49,1430B.

Cohen, J. (1977). Statistical power analysis for the behavioral sciences(rev. ed). San Diego, CA: Academic Press.

Cooley, R. J. (1991). Adult children of alcoholics: Interpersonal trust andsocial compensation. Unpublished master's thesis, University of To-ledo.

Cooper, H. M. (1989). Integrating research: A guide for literature re-views (2nd ed.). Newbury Park, CA: Sage.

Crocker, J., & Luhtanen, R. (1990). Collective self-esteem and ingroupbias. Journal of Personality and Social Psychology, 58, 60-67.

Davis, J. H. (1969). Group performance. Reading, MA: Addison-Wesley.Deci, E. L. (1975). Intrinsic motivation. New York: Plenum Press.Eagly, A. H. (1987). Sex differences in social behavior: A social-role

interpretation. Hillsdale, NJ: Erlbaum.Eagly, A. H., & Steffen, V J. (1984). Gender stereotypes stem from the

distribution of women and men into social roles. Journal of Personal-ity and Social Psychology, 46, 735-754.

Early, P. C. (1989). Social loafing and collectivism: A comparison of theUnited States and the People's Republic of China. AdministrativeScience Quarterly, 34, 565-581.

Gabrenya, W K., Jr., Latane, B., & Wang, Y. E. (1981, August). Socialloafing among Chinese overseas and U.S. students. Paper presented atthe Asian Conference of the International Association for Cross-Cultural Psychology, Taipei, Taiwan.

Gabrenya, W K., Jr., Latane, B., & Wang, Y. E. (1983). Social loafing incross-cultural perspective: Chinese in Taiwan. Journal of Cross-Cul-tural Psychology, 14, 368-384.

Geen, R. G. (1991). Social motivation. Annual Review of Psychology,42, 377-399.

Glass, G. V, McGaw, B., & Smith, M. L. (1981). Meta-analysis in socialresearch. Beverly Hills, CA: Sage.

Goethals, G, & Darley, J. (1987). Social comparison theory: Self-eva-luation and group life. In B. Mullen & G. Goethals (Eds.), Theories ofgroup behavior (pp. 21-47). New York: Springer-Verlag.

Greenwald, A. G. (1982). Ego task analysis: An integration of researchon ego-involvement and self-awareness. In A. H. Hastorf & A. M.Isen (Eds.), Cognitive social psychology (pp. 109-147). New York:Elsevier North-Holland.

Greenwald, A. G, & Pratkanis, A. R. (1984). The self. In R. S. Wyer &T. K. Srull (Eds.), Handbook of social cognition (pp. 129-178). Hills-dale, NJ: Erlbaum.

Griffith, T L., Fichman, M., & Moreland, R. L. (1989). Social loafingand social facilitation: An empirical test of the cognitive-motiva-tional model of performance. Basic and Applied Social Psychology,70.253-271.

Hackman, J. R., & Lawler, E. E., III. (1971). Employee reactions to jobcharacteristics. Journal of Applied Psychology, 55, 259-285.

Hallmark, J. R., & Downs, T. M. (1987, November). Group participationin the organization: Social loafing as a limitation of group effective-ness. Paper presented at the annual meetingof the Speech Communi-cation Association, Boston.

Harcum, E. R., & Badura, L. L. (1990). Social loafing as a response toan appraisal of appropriate effort. Journal of Psychology, 124, 629-637.

SOCIAL LOAFING 703

Hardy, C, & Latane, B. (1986). Social loafing on a cheering task. SocialScience, 77,165-172.

Hardy, C. J., & Latane, B. (1988). Social loafing in cheerleaders: Effectsof team membership and competition. Journal of Sport and ExercisePsychology, 10,109-114.

Harkins, S. G. (1987). Social loafing and social facilitation. Journal ofExperimental Social Psychology, 23,1-18.

Harkins, S. G., & Jackson, J. M. (1985). The role of evaluation in elimi-nating social loafing. Personality and Social Psychology, 11, 575-584.

Harkins, S. G., Latane, B., & Williams, K. D. (1980). Social loafing:Allocating effort or taking it easy? Journal of Experimental SocialPsychology, 16, 457-465.

Harkins, S. G, & Petty, R. E. (1982). Effects of task difficulty and taskuniqueness on social loafing. Journal of Personality and Social Psy-chology, 45,1214-1229.

Harkins, S. G, & Szymanski, K. (1987). Social loafing and social facili-tation: New wine in old bottles. In C. Hendrick (Ed.), Review ofpersonality and social psychology (Vol. 9, pp. 167-188). NewburyPark, CA: Sage.

Harkins, S. G., & Szymanski, K. (1988). Social loafing and self-evalua-tion with an objective standard. Journal of Experimental Social Psy-chology, 24, 354-365.

Harkins, S. G., & Szymanski, K. (1989). Social loafing and group evalu-ation. Journal of Personality and Social Psychology, 56, 934-941.

Heckhausen, H. (1977). Achievement motivation and its constructs: Acognitive model. Motivation and Emotion, 1, 283-329.

Hedges, L. V (1987). How hard is hard science, how soft is soft science?The empirical cumulativeness of research. American Psychologist,42, 443-455.

Hedges, L. V, & Olkin, I. (1985). Statistical methods for meta-analysis.San Diego, CA: Academic Press.

Hsu, F. L. K. (1970). Americans and Chinese: Purposes and fulfillmentin great civilization (2nd ed.). Garden City, NY: Natural HistoryPress.

Huddleston, S., Doody, S. G, & Ruder, M. K. (1985). The effect of priorknowledge of the social loafing phenomenon on performance in agroup. International Journal of Sport Psychology, 16, 176-182.

Ingham, A. G, Levinger, G, Graves, J., & Peckham, V (1974). TheRingelmann effect: Studies of group size and group performance.Journal of Personality and Social Psychology, 10, 371-384.

Jackson, J. M. (1986). In defense of social impact theory: Comment onMullen. Journal of Personality and Social Psychology, 50, 511 -513.

Jackson, J. M., & Harkins, S. G. (1985). Equity in effort: An explanationof the social loafing effect. Journal of Personality and Social Psychol-ogy, 49,1199-1206.

Jackson, J. M., & Latane, B. (1982a, May). Obedience: Strength andimmediacy of authority and social loafing. Paper presented at theannual meeting of the Midwestern Psychological Association, Min-neapolis.

Jackson, 1 M., & Latane, B. (1982b, May). Experimenter immediacyand social loafing. Paper presented at the annual meeting of theMidwestern Psychological Association, Minneapolis.

Jackson, J. M, & Williams, K. D. (1985). Social loafing on difficulttasks: Working collectively can improve performance. Journal of Per-sonality and Social Psychology, 49, 937-942.

Jackson, J. M., Williams, K. D., & Latane, B. (1978, May). The effects ofsexual composition and group size on individual effort. Paper pre-sented at the annual meeting of the Midwestern Psychological Asso-ciation, Chicago.

Johnson, B. T. (1989). DSTAT: Software for the meta-analytic review ofresearch literatures. Hillsdale, NJ: Erlbaum.

Karau, S. J., & Williams, K. D. (1993). The effects of group cohesivenesson social loafing and social compensation. Manuscript submitted forpublication.

Kerr, N. L. (1983). Motivation losses in small groups: A social di-lemma analysis. Journal of Personality and Social Psychology, 45,819-828.

Kerr, N. L., & Bruun, S. (1981). Ringelmann revisited: Alternativeexplanations for the social loafing effect. Journal of Personality andSocial Psychology Bulletin. 7, 224-231.

Kerr, N. L., & Bruun, S. (1983). The dispensability of member effortand group motivation losses: Free rider effects. Journal of Personal-ity and Social Psychology, 44, 78-94.

Kravitz, D. A., & Martin, B. (1986). Ringelmann rediscovered: Theoriginal article. Journal of Personality and Social Psychology, 50,936-941.

Latane, B. (1981). The psychology of social impact. American Psycholo-gist, 36, 343-356.

Latane, B., Williams, K., & Harkins, S. (1979). Many hands make lightthe work: The causes and consequences of social loafing. Journal ofPersonality and Social Psychology, 37, 822-832.

Leary, M. R., & Forsyth, D. R. (1987). Attributions of responsibility forcollective endeavors. InC. Hendrick (Ed.), Review of personality andsocial psychology: Group processes (Vol. 8, pp. 167-188). NewburyPark, CA: Sage.

Lepper, M., & Greene, D. (Eds.). (1978). The hidden costs of reward.Hillsdale, NJ: Erlbaum.

Levine, J. M., & Moreland, R. L. (1987). Social comparison and out-come evaluation in group contexts. In J. C. Masters & W P. Smith(Eds.), Social comparison, social justice, and relativedeprivation: The-oretical, empirical, and policy perspectives. Hillsdale, NJ: Erlbaum.

Martin, L. L., Seta, J. J., & Crelia, R. A. (1990). Assimilation and con-trast as a function of people's willingness and ability to expend effortin forming an impression. Journal of Personality and Social Psychol-ogy, 59, 27-37.

Matsui, T, Kakuyama, X, & Onglatco, M. U. (1987). Effects of goalsand feedback on performance in groups. Journal of Applied Psychol-ogy, 72,407-415.

McGovern, J. M. (1986). The role of insight in the reduction of socialloafing. Unpublished master's thesis, Florida Institute of Technol-ogy.

Mullen, B. (1983). Operationalizing the effect of the group on the indi-vidual: A self-attention perspective. Journal of Experimental SocialPsychology, 19, 295-322.

Mullen, B. (1985). Strength and immediacy of sources: A meta-analyticevaluation of the forgotten elements of social impact theory. Journalof Personality and Social Psychology, 48, 1458-1466.

Mullen, B. (1991). The phenomenology of being in a group: Integra-tionsof social cognition and group processes. Journal of Experimen-tal Social Psychology, 27, 297-323.

Naylor, J. C, Pritchard, R. D., & Ilgen, D. R. (1980). A theory of behaviorin organizations. San Diego, CA: Academic Press.

Olson, M. (1965). The logic of collective action: Public goods and thetheory of groups. Cambridge, MA: Harvard University Press.

Paulus, P. B. (1983). Group influences on individual task performance.In P. B. Paulus (Ed.), Basic group processes (pp. 97-120). New York:Springer-Verlag.

Peterson, C, Zaccaro, S. J., & Daly, D. C. (1986). Learned helplessnessand the generality of social loafing. Cognitive Therapy and Research,10, 563-569.

Petty, R. E., Cacioppo, J. T, & Kasmer, J. A. (1985, May). Individualdifferences in social loafing on cognitive tasks. Paper presented at theannual meeting of the Midwestern Psychological Association, Chi-cago.

Petty, R. E., Harkins, S. G, & Williams, K. D. (1980). The effects ofdiffusion of cognitive effort on attitudes: An information processingview. Journal of Personality and Social Psychology, 38, 81-92.

Petty, R., Harkins, S., Williams, K., & Latane, B. (1977). The effects of


group size on cognitive effort and evaluation. Personality and SocialPsychology Bulletin, 3, 579-582.

Porter, L. W, & Lawler, E. E. (1968). Managerial attitudes and perfor-mance. Homewood, IL: Dorsey Press.

Price, K. H. (1987). Decision responsibility, task responsibility, identi-fiability, and social loafing. Organizational Behavior and HumanDecision Processes, 40, 330-345.

Quintanar, L. R., & Pryor, J. B. (1982, August). Group diffusion of cogni-tive effort as a determinant of attribution. Paper presented at the an-nual meeting of the American Psychological Association, Washing-ton, DC.

Rosenthal, R. (1991). Meta-analylic procedures for social research (2nded.). Newbury Park, CA: Sage.

Rosenthal, R., & Rubin, D. B. (1986). Meta-analytic procedures forcombining studies with multiple effect sizes. Psychological Bulletin,99, 400-406.

Schachter, S. (1959). The psychology of affiliation: Experimental stud-ies of the sources ofgregariousness. Stan ford, CA: Stanford UniversityPress.

Shepperd, J. A. (1993). Productivity loss in performance groups: Amotivation analysis. Psychological Bulletin, 113, 67-81.

Shepperd, J. A., & Wright, R. A. (1989). Individual contributions to acollective effort: An incentive analysis. Personality and Social Psy-chology Bulletin, 15, 141-149.

Shirakashi, S. (1985). Social loafing of Japanese students. HiroshimaForum for Psychology, 10, 35-40.

Sorrentino, R. M., & Sheppard, B. H. (1978). Effects of affiliation-re-lated motives on swimmers in individual versus group competition:A field experiment. Journal of Personality and Social Psychology, 36,704-714.

Steiner, I. D. (1972). Group process and productivity. San Diego, CA:Academic Press.

Stevenson, C. (1990). Motivating effort under social loafing condi-tions: The effects of private self-awareness on task effort. Disserta-tion Abstracts International, 50, 482IB. (University Microfilms No.90-06796)

Stroebe, W, & Frey, B. S. (1982). Self-interest and collective action: Theeconomics and psychology of public goods. British Journal of SocialPsychology, 21,121-137.

Suls, J. M., & Miller, R. L. (Eds.). (1977). Social comparison processes:Theoretical and empirical perspectives. Washington, DC: Hemi-sphere.

Szymanski, K., & Harkins, S. G. (1987). Social loafing and self-evalua-tion with a social standard. Journal of Personality and Social Psychol-ogy, 53,891-897.

Tajfel, H., & Turner, J. (1986). The social identity theory of intergroupbehavior. In S. Worchel & W Austin (Eds.), Psychology of intergrouprelations (pp. 33-48). Chicago: Nelson-Hall.

Taylor, S. E., & Brown, J. (1988). Illusion and well-being: Some socialpsychological contributions to a theory of mental health. Psychologi-cal Bulletin, 103,193-210.

Tesser, A. (1988). Toward a self-evaluation maintenance model of so-cial behavior. In L. Berkowitz (Ed.), Advances in experimental socialpsychology (Vol. 21, pp. 181-227). San Diego, CA: Academic Press.

Thompson, J., Jackson, J., Williams, K., & Latane, B. (1980). The ef-fects ofcoactor ability on social loafing. Paper presented at the annualGraduate Student Conference in Personality and Social Psychology,Columbus, OH.

Triandis, H. C. (1989). The self and social behavior in differing culturalcontexts. Psychological Review, 96, 506-520.

Vroom, V H. (1964). Work and motivation. New York: Wiley.Waller, J. E. (1989). Developmental aspects of self-evaluation and so-

cial loafing in children (Doctoral dissertation, University of Ken-tucky). Dissertation Abstracts International, 49, 5574B.

Weldon, E., & Gargano, G. M. (1985). Cognitive effort in additive taskgroups: The effects of shared responsibility on the quality of multiat-tribute judgments. Organizational Behavior and Human DecisionProcesses, 36, 348-361.

Weldon, E., & Gargano, G. M. (1988). Cognitive loafing: The effects ofaccountability and shared responsibility on cognitive effort. Person-ality and Social Psychology Bulletin, 14, 159-171.

Weldon, E., & Mustari, E. L. (1988). Felt dispensability in groups ofcoactors: The effects of shared responsibility and explicit anonymityon cognitive effort. Organizational Behavior and Human DecisionProcesses, 41, 330-351.

Wheeler, L., Reis, H. T, & Bond, M. H. (1989). Collectivism-individual-ism in everyday social life: The middle kingdom and the meltingpot. Journal of Personality and Social Psychology, 57, 79-86.

Williams, K. D. (1981, May). The effects of group cohesiveness on socialloafing. Paper presented at the annual meeting of the MidwesternPsychological Association, Detroit.

Williams, K. D., & Burmont, S. C. (1981, June). A look at social loafingin a cognitive task: Individual effort in advertising evaluations. Paperpresented at the First Annual Nags Head Conference, Kill DevilHills, NC.

Williams, K., Harkins, S., & Latane, B. (1981). Identifiability as a de-terrent to social loafing: Two cheering experiments. Journal of Per-sonality and Social Psychology, 40, 303-311.

Williams, K. D, & Karau, S. J. (1991). Social loafing and social com-pensation: The effects of expectations of co-worker performance.Journal of Personality and Social Psychology, 61, 570-581.

Williams, K. D, Nida, S. A., Baca, L. D, & Latane, B. (1989). Socialloafing and swimming: Effects of identifiability on individual andrelay performance of intercollegiate swimmers. Basic and AppliedSocial Psychology, 10, 73-81.

Williams, K. D, & Williams, K. B. (1981, August). The development ofsocial loafing in America. Paper presented at the Asian Conferenceof the International Association for Cross-Cultural Psychology, Tai-pei, Taiwan.

Williams, K. D, & Williams, K. B., Kawana, Y, & Latane, B. (1984,May). Collective effort in small groups in the United Stales and Japan.Paper presented at the annual meeting of the Midwestern Psycholog-ical Association, Chicago.

Wrightsman, L. S. (1960). Effects of waiting with others on changes inlevel of felt anxiety. Journal of Abnormal and Social Psychology, 61,216-222.

Yamaguchi, S., Okamoto, K., & Oka, T. (1985). Effects of coactor'spresence: Social loafing and social facilitation. Japanese Psychologi-cal Research, 27, 215-222.

Zaccaro, S. J. (1984). Social loafing: The role of task attractiveness.Personality and Social Psychology Bulletin, 10, 99-106.

Zalesny, M. D, & Ford, J. K. (1990). Extending the social informationprocessing perspective: New links to attitudes, behaviors, and per-ceptions. Organizational Behavior and Human Decision Processes,47, 205-246.

SOCIAL LOAFING 705

AppendixStudies Used in the Meta-Analysis

Bartis, S., Szymanski, K., & Harkins, S. G. (1988). Evaluation and per-formance: A two-edged knife. Personality and Social PsychologyBulletin, 14, 242-251.

Brickner, M. A., Harkins, S. G., & Ostrom, T, M. (1986). Effects ofpersonal involvement: Thought-provoking implications for socialloafing. Journal of Personality and Social Psychology, 51, 763-769.

Brickner, M. A., & Wingard, K. (1988). Social identity and self-motiva-tion: Improvements in productivity. Unpublished manuscript. Akron,OH: University of Akron.

Cates, K. H. (1986). The effects of evaluation, identifiability, and taskuniqueness on task performance. Dissertation Abstracts Interna-tional, 47, 1326B. (University Microfilms No. 86-11258)

Cohen, C. J. (1988). Social loafing and personality: The effects of indi-vidual differences on collective performance (Doctoral disserta-tion, Fordham University). Dissertation Abstracts International, 49,1430B.

Cooley, R. 1 (1991). Adult children of alcoholics: Interpersonal trust andsocial compensation. Unpublished master's thesis, University of To-ledo.

Earley, P. C. (1989). Social loafing and collectivism: A comparison ofthe United States and the People's Republic of China. AdministrativeScience Quarterly, 34, 565-581.

Gabrenya, W K., Jr., Latane, B., & Wang, Y. E. (1981, August). Socialloafing among Chinese overseas and U.S. students. Paper presented atthe Asian Conference of the International Association for Cross-Cultural Psychology, Taipei, Taiwan.

Gabrenya, W K., Jr., Latane, B., & Wang, Y. E. (1983). Social loafing incross-cultural perspective: Chinese in Taiwan. Journal o] Cross-Cul-tural Psychology, 14, 368-384.

Gabrenya, W K., Wang, T. E., & Latane, B. (1985). Social loafing on anoptimizing task: Cross-cultural differences among Chinese andAmericans. Journal of Cross-cultural Psychology, 16, 223-242.

Griffith, T. L., Fichman, M., & Moreland, R. L. (1989). Social loafingand social facilitation: An empirical test of the cognitive-motiva-tional model of performance. Basic and Applied Social Psychology,10, 253-271.

Hallmark, J. R., & Downs, T. M. (1987, November). Group participationin the organization: Social loafing as a limitation of group effective-ness. Paper presented at the annual meeting of the Speech Communi-cation Association, Boston.

Harcum, E. R., & Badura, L. L. (1990). Social loafing as a response toan appraisal of appropriate effort. Journal of Psychology, 124, 629-637.

Hardy, C, & Latane, B. (1986). Social loafing on a cheering task. SocialScience, 77,165-172.

Hardy, C. J., & Latane, B. (1988). Social loafing in cheerleaders: Effectsof team membership and competition. Journal of Sport and ExercisePsychology, 10, 109-114.

Harkins, S. G. (1987). Social loafing and social facilitation. Journal ofExperimental Social Psychology, 23,1-18.

Harkins, S. G., & Jackson, J. M. (1985). The role of evaluation in elimi-nating social loafing. Personality and Social Psychology Bulletin, 11,575-584.

Harkins, S. G., Latane, B., Williams, K. D. (1980). Social loafing: Allo-cating effort or taking it easy? Journal of Experimental Social Psy-chology, 16, 457-465.

Harkins, S. G., & Petty, R. E. (1982). Effects of task difficulty and taskuniqueness on social loafing. Journal of Personality and Social Psy-chology, 43, 1214-1229.

Harkins, S. G., & Szymanski, K. (1988). Social loafing and self-evalua-

tion with an objective standard. Journal of Experimental Social Psy-chology, 24, 354-365.

Harkins, S. G., & Szymanski, K. (1989). Social loafing and group evalu-ation. Journal of Personality and Social Psychology, 56, 934-941.

Ingham, A. G, Levinger, G., Graves, J., & Peckham, V (1974). TheRingelmann effect: Studies of group size and group performance.Journal of Personality and Social Psychology, 10, 371-384.

Jackson, J. M, & Harkins, S. G. (1985). Equity in effort: An explanationof the social loafing effect. Journal of Personality and Social Psychol-ogy, 49,1199-1206.

Jackson, J. M., & Latane, B. (1982a, May). Obedience: Strength andimmediacy of authority and social loafing. Paper presented at theannual meeting of the Midwestern Psychological Association, Min-neapolis.

Jackson, J. M., & Latane, B. (1982b, May). Experimenter immediacyand social loafing. Paper presented at the annual meeting of theMidwestern Psychological Association, Minneapolis.

Jackson, J. M., & Williams, K. D. (1985). Social loafing on difficulttasks: Working collectively can improve performance. Journal of Per-sonality and Social Psychology, 49, 937-942.

Jackson, J. M., Williams, K. D, & Latane, B. (1978, May). The effects ofsexual composition and group size on individual effort. Paper pre-sented at the annual meeting of the Midwestern Psychological Asso-ciation, Chicago.

Karau, S. J., & Williams, K. D. (1993). The effects of group cohesivenesson social loafing and social compensation. Manuscript submitted forpublication.

Kerr, N. L. (1983). Motivation losses in small groups: A social di-lemma analysis. Journal of Personality and Social Psychology, 45,819-828.

Kerr, N. L., & Bruun, S. (1981). Ringelmann revisited: Alternativeexplanations for the social loafing effect. Personality and Social Psy-chology Bulletin, 7, 224-231.

Kerr, N. L., & Bruun, S. (1983). The dispensability of member effortand group motivation losses: Free rider effects. Journal of Personal-ity and Social Psychology, 44, 78-94.

Latane, B., Williams, K., & Harkins, S. (1979). Many hands make lightthe work: The causes and consequences of social loafing. Journal ofPersonality and Social Psychology, 37, 822-832.

Martin, L. L., Seta, J. J., & Crelia, R. A. (1990). Assimilation and con-trast as a function of people's willingness and ability to expend effortin forming an impression. Journal of Personality and Social Psychol-ogy, 59, 27-37.

Matsui, X, Kakuyama, T., & Onglatco, M. U. (1987). Effects of goalsand feedback on performance in groups. Journal of Applied Psychol-ogy, 72, 407-415.

McGovern, J. M. (1986). The role of insight in the reduction of socialloafing. Unpublished master's thesis, Florida Institute of Technol-ogy.

Peterson, C, Zaccaro, S. I, & Daly, D. C. (1986). Learned helplessnessand the generality of social loafing. Cognitive Therapy and Research,10, 563-569.

Petty, R. E., Cacioppo, J. T, & Kasmer, J. A. (1985, May). Individualdifferences in social loafing on cognitive tasks. Paper presented at theannual meeting of the Midwestern Psychological Association, Chi-cago.

Petty, R. E., Harkins, S. G., & Williams, K. D. (1980). The effects ofdiffusion of cognitive effort on attitudes: An information processingview. Journal of Personality and Social Psychology, 38, 81-92.

Petty, R., Harkins, S., Williams, K., & Latane, B. (1977). The effects of


group size on cognitive effort and evaluation. Personality and SocialPsychology Bulletin, 3, 579-582.

Price, K. H. (1987). Decision responsibility, task responsibility, identi-fiability, and social loafing. Organizational Behavior and HumanDecision Processes, 40, 330-345.

Quintanar, L. R., & Pryor, J. B. (1982, August). Group diffusion of cogni-tive effort as a determinant of attribution. Paper presented at the an-nual meeting of the American Psychological Association, Washing-ton, DC.

Shepperd, J. A., & Wright, R. A. (1989). Individual contributions to acollective effort: An incentive analysis. Personality and Social Psy-chology Bulletin, 15, 141-149.

Shirakashi, S. (1985). Social loafing of Japanese students. HiroshimaForum for Psychology, 10, 35-40.

Sorrentino, R. M, & Sheppard, B. H. (1978). Effects of affiliation-re-lated motives on swimmers in individual versus group competition:A field experiment. Journal of Personality and Social Psychology, 36,704-714.

Stevenson, C. (1990). Motivating effort under social loafing condi-tions: The effects of private self-awareness on task effort. Disserta-tion Abstracts International, 50, 482 IB. (University Microfilms No.90-06796)

Szymanski, K., & Harkins, S. G. (1987). Social loafing and self-evalua-tion with a social standard. Journal of Personality and Social Psychol-ogy, 53, S91-897.

Thompson, J., Jackson, J., Williams, K., & Latane, B. (1980). The ef-fects of coactor ability on social loafing. Paper presented at the annualGraduate Student Conference in Personality and Social Psychology,Columbus, OH.

Waller, J. E. (1989). Developmental aspects of self-evaluation and so-cial loafing in children (Doctoral dissertation, University of Ken-tucky). Dissertation Abstracts International, 49, 5574B.

Weldon, E., & Gargano, G. M. (1985). Cognitive effort in additive taskgroups: The effects of shared responsibility on the quality of multiat-tribute judgments. Organizational Behavior and Human DecisionProcesses, 56,348-361.

Weldon, E., & Gargano, G. M. (1988). Cognitive loafing: The effects ofaccountability and shared responsibility on cognitive effort. Person-ality and Social Psychology Bulletin, 14,159-171.

Weldon, E., & Mustari, E. L. (1988). Felt dispensability in groups ofcoactors: The effects of shared responsibility and explicit anonymityon cognitive effort. Organizational Behavior and Human DecisionProcesses, 41, 330-351.

Williams, K. D. (1981, May). The effects of group cohesiveness on socialloafing. Paper presented at the annual meeting of the MidwesternPsychological Association, Detroit.

Williams, K. D, & Burmont, S. C. (1981, June). A look at social loafingin a cognitive task: Individual effort in advertising evaluations. Paperpresented at the 1st Annual Nags Head Conference, Kill Devil Hills,NC.

Williams, K., Harkins, S., & Latane, B. (1981). Identifiability as a de-terrent to social loafing: Two cheering experiments. Journal of Per-sonality and Social Psychology, 40, 303-311.

Williams, K. D, & Karau, S. J. (1991). Social loafing and social com-pensation: The effects of expectations of co-worker performance.Journal of Personality and Social Psychology, 61, 570-581.

Williams, K. D, Nida, S. A., Baca, L. D, & Latane, B. (1989). Socialloafing and swimming: Effects of identifiability on individual andrelay performance of intercollegiate swimmers. Basic and AppliedSocial Psychology, 10, 73-81.

Williams, K. D., & Williams, K. B. (1981, August). The development ofsocial loafing in America. Paper presented at the Asian Conferenceof the International Association for Cross-Cultural Psychology, Tai-pei, Taiwan.

Williams, K. D, & Williams, K. B., Kawana, Y, & Latane, B. (1984,May). Collective effort in small groups in the United States and Japan.Paper presented at the annual meeting of the Midwestern Psycholog-ical Association, Chicago.

Yamaguchi, S., Okamoto, K., & Oka, T. (1985). Effects of coactor'spresence: Social loafing and social facilitation. Japanese Psychologi-cal Research, 27, 215-222.

Received February 18,1992Revision received May 3,1993

Accepted May 11,1993

social loafing: a meta-analytic review and theoretical ... · pdf filesocial loafing: a...

Documents