eye movement monitoring as a process tracing methodology ......process tracing methods such as...

Eye Movement Monitoring as a Process Tracing Methodology inDecision Making Research

Mackenzie G. Glaholt and Eyal M. ReingoldUniversity of Toronto at Mississauga

Over the past half century, research on human decision making has expanded from apurely behaviorist approach that focuses on decision outcomes, to include a morecognitive approach that focuses on the decision processes that occur prior to theresponse. This newer approach, known as process tracing, has employed variousmethods, such as verbal protocols, information search displays, and eye movementmonitoring, to identify and track psychological events that occur prior to the response(such as cognitive states, stages, or processes). In the present article, we reviewempirical studies that have employed eye movement monitoring as a process tracingmethod in decision making research, and we examine the potential of eye movementmonitoring as a process tracing methodology. We also present an experiment thatfurther illustrates the experimental manipulations and analysis techniques that arepossible with modern eye tracking technology. In this experiment, a gaze-contingentdisplay was used to manipulate stimulus exposure during decision making, whichallowed us to test a specific hypothesis about the role of eye movements in preferencedecisions (the Gaze Cascade model; Shimojo, Simion, Shimojo, & Scheier, 2003). Theresults of the experiment did not confirm the predictions of the Gaze Cascade model,but instead support the idea that eye movements in these decisions reflect the screeningand evaluation of decision alternatives. In summary, we argue that eye movementmonitoring is a valuable tool for capturing decision makers’ information searchbehaviors, and that modern eye tracking technology is highly compatible with otherprocess tracing methods such as retrospective verbal protocols and neuroimagingtechniques, and hence it is poised to be an integral part of the next wave of decisionresearch.

Keywords: eye movements, process tracing, decision making

Early research in decision making focused onexplaining and predicting decision outcomes,such as choices and judgments. This approach,known as the structural approach, produces sta-tistical models that account for decision mak-ers’ responses (outcomes) as a function of thestimulus information and parameters of the de-cision (inputs). While this approach has beenquite successful in probing the strategies thatdecision makers apply, the modeling of out-comes cannot identify different stages of the

decision process, or changes in the decisionstrategy that might occur prior to the final re-sponse. To shed light on these aspects of deci-sion making, researchers have sought ways toobserve more directly the cognitive processesthat occur prior to the final behavioral response.This approach, which was originally applied toresearch on problem solving, is known as pro-cess tracing (for discussion regarding these twoapproaches, see Abelson & Levi, 1985; Billings& Marcus, 1983; Einhorn, Kleinmuntz, &Kleinmuntz, 1979; Harte & Koele, 1995; Payne,Braunstein, & Carrol, 1978; Svenson, 1979,1996).

There are several methods of process tracingthat have been applied in decision making re-search, primarily involving the use of verbalprotocol techniques or information search dis-play paradigms (for reviews see Ford, Schmitt,Schechtman, Hults, & Doherty, 1989; Riedl,

Mackenzie G. Glaholt and Eyal M. Reingold, Departmentof Psychology, University of Toronto at Mississauga, Mis-sissauga, Ontario, Canada.

Correspondence concerning this article should be ad-dressed to Mackenzie G. Glaholt, Department of Psychol-ogy, University of Toronto at Mississauga, 3359 Missis-sauga Road N. RM 2037B, Mississauga, Ontario, CanadaL5L 1C6. E-mail: [email protected]

Journal of Neuroscience, Psychology, and Economics © 2011 American Psychological Association2011, Vol. 4, No. 2, 125–146 1937-321X/11/$12.00 DOI: 10.1037/a0020692

125

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Brandstatter, & Roithmayr, 2008). The maingoal of the present article is to evaluate andillustrate the potential utility of eye movementmonitoring as a process tracing method. Ac-cordingly, we begin by briefly introducing andcontrasting several methods of process tracing.We then review research that employed eyemovement recordings to study processes under-lying decision making. Finally, we report anexperiment that demonstrates the unique possi-bilities afforded by eye movement monitoring.In particular, we tested a hypothesis about therole of eye movements in preference decisions(the Gaze Cascade model; Shimojo et al., 2003;Simion & Shimojo, 2006, 2007). Using a so-phisticated method of updating the displaybased on eye position (gaze-contingent display),stimulus exposure was manipulated during on-going decisions, allowing for a direct test of thepredictions of the model proposed by Shimojoand colleagues. In addition, the present studydemonstrates several new techniques for ana-lyzing eye movement data in the context ofmultialternative decisions.

Process Tracing and MonitoringInformation Search

Early process tracing research employedverbal protocols in an attempt to observe thecognitive processes that occur during decisionmaking. Verbal protocols involve having thedecision maker describe what they are think-ing or doing (i.e., “think aloud”) while mak-ing their decision (concurrent verbal proto-col), or having them recall their decision pro-cess after the decision has been made(retrospective verbal protocol). While the ver-bal protocol methodology has been shown toprovide information regarding the sequenceof information sampled, and can often suggestthe decision strategy that is employed by par-ticipants, this method has several potentialshortcomings. Concurrent verbalization is ef-fectively a secondary task that the decisionmaker conducts in parallel with the ongoingdecision, the burden of which has been shownto reduce decision accuracy (Russo, Johnson,& Stephens, 1989). Retrospective verbal pro-tocols rely on the decision maker having ac-curate memory for the decision process as itunfolded. However, it has been demonstratedthat decision makers’ retrospective protocols

reflect substantial forgetting and confabula-tion (Russo et al., 1989). To corroborate andexpand the information revealed by verbalreports, methods have been developed tomore directly observe the patterns of infor-mation search that are employed by decisionmakers.

The traditional method of monitoring infor-mation search is through the use of informationsearch displays. In these paradigms (also re-ferred to as information display boards, infor-mation display matrices, or computer processtracing), decision makers are presented with amatrix of stimulus information (alternatives bycolumn and attributes by row, or vice versa) andthey are tasked with making a decision aboutthe alternatives according to some decision rule.Of note, the information search display para-digm constrains the way in which the decisionmaker samples information from the display.The information in each cell in the matrix ishidden. Decision makers must access cells in-dividually, and the displayed information isconcealed when the decision maker selects an-other cell. In this way the decision maker’spattern of information search is made explicit.A variety of measures of information search canbe derived from this method, such as the depthof search, variability of search, and pattern ofsearch (see Riedl et al., 2008 for a review ofthese measures). These measures can be used toinfer the presence of certain decision strategies,and can detect transitions between differentstages of processing (Ball, 1997; Billings &Marcus, 1983; Levin, Huneke, & Jasper, 2000;Payne, 1976; Payne, Bettman, & Johnson,1993).

In the early versions of this paradigm, theinformation in each cell was written on a card inan envelope, and the decision maker would re-move, view, and then replace each card individ-ually. Computerized versions of this paradigmpresent the information matrix on a computerscreen, and participants use a pointing device(often a mouse) to reveal individual cells in thematrix. Computerized versions of the paradigmoffer obvious advantages in both ease of use forthe decision maker, and in precision of themeasurements obtained. However, all versionsof this method suffer from a potential shortcom-ing. In these paradigms the decision maker mustexecute a deliberate manual act in order to sam-ple each piece of information (i.e., select it by

126 GLAHOLT AND REINGOLD

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

hand). In natural decision situations in which adecision maker is presented with multiple alter-natives, decision makers sample informationfrom those alternatives by directing their gaze tothem. In general, people make rapid eye move-ments (called saccades) roughly three to fourtimes each second. The periods between sac-cades where the eye is relatively still, and visualinformation is extracted, are known as fixations(for a review of eye movement measures, seeRayner, 1998, 2009). Eye movements are con-siderably faster than movements of the hand,and they require less deliberate effort to exe-cute. This difference has important implicationsfor process tracing in decision making becauseit has been argued that effort, and basic infor-mation processing limitations such as memory,play a significant role in the way that decisionsare made (Payne et al., 1993), and hence anymethod of process tracing that produces artifi-cial demands of these kind may actually alterthe decision process.

This issue was investigated by Lohse andJohnson (1996), in a direct comparison of theinformation search display and eye movementmonitoring process tracing methods (specifi-cally, comparing a computerized informationsearch display tool called Mouselab, and an eyemovement monitoring system called Eyegaze).Consistent with prior findings (Russo, 1978;van Raaij, 1977), they found significant differ-ences in the process data obtained from thesetwo process tracing methodologies. In particu-lar, Lohse and Johnson found that the informa-tion search display paradigm produced longertotal time per decision, longer time per infor-mation acquisition, lower rate of reacquisition,more information searched, and reduced deci-sion accuracy. Furthermore, these differencesbecame more pronounced in decisions that aremore complex (i.e., more alternatives or attri-butes). Lohse and Johnson concluded that com-pared with eye movement monitoring, the in-formation search display methodology imposesgreater demands in effort and working memory.

These findings strongly advocate the use ofeye movement recordings as a process tracingmeasure (but see Reisen, Hoffrage, & Mast,2008, for a critique). In addition, there are sev-eral other reasons why eye movement monitor-ing may be desirable over other methods ofmonitoring information search. Of note, unlikeinformation search displays that are primarily

sensitive to deliberate information sampling,eye movement recordings capture a broaderrange of information sampling acts, which areexecuted with or without conscious awareness(e.g., eye movements that are elicited exoge-nously). In addition, eye movement recordingsmight be very useful in supplementing and dis-ambiguating concurrent verbal reports or byserving as powerful cues for retrospective ver-bal protocols. For example, in a recent study,Eger, Ball, Stevens, and Dodd (2007) found thateye movements collected passively duringongoing performance were particularly infor-mative when replayed to participants as cuesduring a retrospective verbal protocol (eyemovement-cued retrospective verbal protocol;see also Hansen, 1991; van Gog, Paas, vanMerrienboer, & Witte, 2005). This combinationof methods has also been successful in the con-text of usability research, where the pattern ofeye movements can yield insight into the pro-cessing steps taken by users (for a review of eyetracking in the field of usability see Ehmke &Wilson, 2007; for a more general review of eyetracking in applied contexts, see Duchowski,2002).

One caveat in using eye tracking as a processtracing measure must be acknowledged. Theeye tracking methodology assumes that the de-cision maker’s attention is focused at the pointof fixation, though research in visual attentionhas shown that people are able to direct theirattention covertly to areas of the visual fieldaway from their point of gaze (Posner, Snyder,& Davison, 1980). However, during naturalviewing, attention and eye movements aretightly coupled: the focus of attention tends toshift to a new location just prior to a shift ingaze to that location and consequently thespatial distribution of eye fixations is a goodindirect measure of the distribution of visualattention (for a review see Hoffman, 1998).Furthermore, the link between the eye move-ment and attentional systems is supported byneurophysiological data (e.g., Goldberg &Wurtz, 1972; Kustov & Robinson, 1996;Mohler & Wurtz, 1976; Wurtz & Mohler, 1976)and work with neuropsychological populationssuch as neglect patients (e.g., Johnston & Diller,1986; Walker & Young, 1996). Hence, duringdecision making tasks where participants areallowed to freely view the decision information,eye movements may be generally considered to

127SPECIAL ISSUE: PROCESS TRACING WITH EYE MOVEMENTS

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

provide a valid measure of the spatial distribu-tion of attention.

In the past, the use of eye tracking in processtracing has been dissuaded by the cost of theequipment, the relatively low fidelity of the dataobtained, and the strict requirements for head-stabilization during recording (e.g., a bite-bar).Current video-based eye tracking technologyallows for head-free eye movement monitoringwhile participants view a computer or a projec-tion screen, conditions that are ideal for observ-ing computer-based decisions (e.g., onlineshopping). Also, lightweight portable eye track-ing equipment that is mounted on goggles oreye glass frames is available, allowing for therecording of eye movements while people makedecisions in natural everyday decision environ-ments (e.g., store shopping). In addition thespatial and temporal resolution as well as theaccuracy of present day eye trackers is vastlysuperior to that of their predecessors. Theseimprovements in eye tracking technology allowfor the possibility of a real time manipulation ofthe stimulus display as a function of the partic-ipant’s gaze location. This technique, known asgaze-contingent display (McConkie & Rayner,1975; Rayner, 1975), allows for a variety ofexperimental manipulations (as is illustrated inthe present experiment). Hence, eye trackingtechnology has advanced to the point where itcan yield high fidelity process tracing data withminimal intrusion upon the natural behavior ofthe decision maker.

Process Tracing With Eye MovementRecordings

Prior research may be coarsely divided ac-cording to the approach used in analyzing theeye movement record. Because of limitationsin the eye tracking technology available at thetime, early studies tended to focus on thespatial distribution of eye movements (e.g.,where the decision maker looks; the order inwhich information is sampled). Advances ineye tracking technology (in particular, in-creases in the rate at which gaze position issampled) have allowed for more sophisticatedanalyses of the temporal information con-tained in the eye movement record (e.g., theduration of individual fixations).

Russo and colleagues (Russo, 1978; Russo &Dosher, 1983; Russo & Leclerc, 1994; Russo &

Rosen, 1975) pioneered the use of eye move-ment monitoring as a process tracing method.These early eye movement studies attempted toidentify natural indices of decision processes inthe eye movement record. For example, Russoand Rosen (1975) observed that certain patternsof gaze transitions were a prominent feature ofthe eye movement record while participantschose one of six cars (where each car wasdescribed by three attributes). Paired compari-sons were identified as sequences in the eyemovement record where participants lookedback and forth between two alternatives (wherethe sequence A-B-A was classified as a “weak”pair and the sequence A-B-A-B was classifiedas a “strong” pair). Participants’ verbal reportscorroborated the claim that such transition se-quences in fact reflected paired comparisons.Russo and Rosen found that these sequencestended to involve alternatives that were similar(i.e., sharing attributes). In addition, alternativesinvolved in paired comparisons tended havehigher subjective utility (as revealed by a sep-arate subjective rating task in which participantsrated alternatives on a continuum between“worst” and “best”), and were refixated morefrequently than alternatives that were not in-volved in paired comparisons. Together thesefindings suggest that decision makers may ef-fectively narrow a multialternative decision to adecision among a smaller set of competitivealternatives. However, in follow-up analyses re-garding the relative utility of the alternativeswithin each pair, it was not possible to deter-mine the exact nature of the processing stepstaking place in paired comparisons (such asevaluation vs. elimination). Hence while the eyemovement record was useful in revealing the setof decision alternatives that are actively beingprocessed, other sources of information mightbe required to identify the actual (cognitive)processing steps that take place.

Russo and Leclerc (1994) analyzed the se-quence of eye fixations in order to identifydifferent stages of the decision process. Eyemovements were recorded through a one-waymirror while participants made consumer deci-sions among groups of up to 16 everyday house-hold items placed on a mock store shelf. Thedecision period was separated into stages basedon the first and last time that the participant’sgaze refixated a decision alternative. Specifi-cally, the first stage was identified as the period


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

prior to the first refixation on an alternative. Thefirst refixation marked the onset of the interme-diate (second) stage. The termination of thesecond stage, and the onset of the third stage ofprocessing, was identified by the first refixationon an alternative when counting back from theresponse (i.e., identical to the method for iden-tifying the first stage, but looking at the se-quence of fixations in reverse). Russo andLeclerc hypothesized that the first stage mightinvolve either a screening process, where deci-sion alternatives are processed selectively basedon their relevance and inferior alternatives areexcluded from further processing, or else anorientation process where alternatives are ini-tially surveyed prior to further processing. Thesecond stage was linked to an evaluative stageof processing where competitive alternativesare compared and the majority of the decision“work” takes place. The final segment of thedecision time course was described as a stageinvolving a review of competitive alternativesjust prior to the final response. Aspects of thisstage structure have been corroborated by re-search using the information search displaymethodology (Wedell & Senter, 1997).

Measurements of the spatial distribution ofeye movements have also been used to derivemeasures that are analogous to those obtainedfrom information search displays, such as thepattern of information search, variability of in-formation search, and depth of informationsearch (Day, Lin, Huang, & Chuang, 2009;Lohse & Johnson, 1996; Pieters & Warlop,1999; Reisen et al., 2008; Rosen & Rosenkoet-ter, 1976; Russo & Dosher, 1983; Selart,Kuvaas, Boe, & Takemura, 2006). Specifically,to index the pattern of information search (de-veloped by Payne, 1976), researchers comparethe number of alternative-wise gaze transitions(i.e., moving one’s gaze from one attribute toanother within a single alternative) and thenumber of attribute-wise gaze transitions (i.e.,moving one’s gaze from one attribute in a de-cision alternative to the same attribute in an-other alternative). In particular, these measurescan describe the extent to which the partici-pant’s decision strategy involves the holisticencoding and evaluation of decision alterna-tives. When confronted with a complex decisionscenario (e.g., with many alternatives and manyattributes), limitations in information process-ing capacity may prevent decision makers from

encoding each decision alternative holisticallyalong all attributes. Instead, participants mayadopt heuristic strategies that result in changesin the pattern of search, depth of search, andvariability of search (for a review, see Payne etal., 1993). Specifically, increased complexitybrings about a shift toward an attribute-wisesearch pattern, where alternatives are processedalong particular attributes (e.g., the most impor-tant attributes). In addition, participants mayignore some decision information altogether(reduced depth of search), or process some al-ternatives or attributes more extensively thanothers (increased variability of search).

In general, measures of search pattern, depthof search, and variability of search, obtainedfrom eye movement recordings have providedconvergent evidence to prior findings using in-formation search displays. However, some dif-ferences between these two process tracingmethodologies have been documented. Lohseand Johnson (1996) reported that comparedwith eye movement monitoring, informationsearch displays induced a more alternative-wisesearch, with greater depth (i.e., more informa-tion sampled), but less variability. In contrast,Reisen et al. (2008) found the informationsearch displays produced a more attribute-wisesearch, with less variability, but with compara-ble depth of search to the eye movement mon-itoring method. Further research is clearly re-quired to tease apart the differences betweenthese two methods of monitoring informationsearch.

Since the first wave of research in this do-main, there has been an increasing trend towardderiving estimates of the temporal characteris-tics of eye movements during decision making(e.g., fixation duration). For example, Pietersand Warlop (1999) monitored participants’ eyemovements while they chose one of six differ-ent brands within a single product category(rice, shampoo, canned soup, or salad dressing).The six alternatives for each trial were dis-played simultaneously on a computer screen.One goal was to test the commonly held beliefregarding consumer decisions that products re-ceiving more attention are more likely to bechosen. In addition, Pieters and Warlop askedwhether two factors external to the decisionwould affect the pattern of eye movements ob-served during the decision: the motivation of thedecision maker (decision rewarded or not), and


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

time constraints imposed upon the decisionmaker (7 second time limit or unlimited). Piet-ers and Warlop obtained an estimate of theduration of individual eye fixations, and theyfound a bias toward the item that was chosen,where participants spent more time fixatingitems that were chosen compared with itemsthat were not. This was driven by longer, andmore frequent, fixations on the chosen item. Inaddition, while the manipulations of task moti-vation and of time constraints did have an im-pact on eye movements overall (e.g., fixationdurations), they did not have a significant effecton the bias in eye movements toward the itemthat was chosen.

More recently, advances in eye tracking tech-nology have allowed for precise analysis of thetemporal information contained in the eyemovement record. The rate of sampling of gazeposition has improved dramatically, allowingfor precise estimates of the duration that stim-ulus information is viewed. This has broughtabout an increase in the application of eye track-ing to decision making (Bee, Prendinger, An-dre, & Ishizuka, 2006; Glaholt & Reingold,2009a, 2009b; Glaholt, Wu, & Reingold, 2009,2010; Schotter, Berry, McKenzie, & Rayner,2010; Shimojo et al., 2003; Simion & Shimojo,2006, 2007; Sutterlin, Brunner, & Opwis, 2008)as well as associated areas including problemsolving and reasoning (Ball, Lucas, Miles, &Gale, 2003; Ball, Phillips, Wade, & Quayle,2006; Ellis, Glaholt, & Reingold, in press), cat-egorization (Rehder & Hoffman, 2005a,2005b); and visual marketing and advertising(Goldberg, Probart, & Zak, 1999; Pieters,Rosenbergen, & Wedel, 1999; Pieters & Wedel,2004, 2007; Radach, Lemmer, Vorstius, Heller,& Radach, 2003; Rayner, Miller, & Rotello,2008; Rayner, Rotello, Stewart, Keir, & Duffy,2001; Wedel & Pieters, 2000; Wedel, Pieters, &Liechty, 2008; for a review, see Wedel & Piet-ers, 2008). These recent studies have introducedseveral new methods of analyzing the temporalinformation contained in the eye movementrecord.

For example, Shimojo et al. (2003) intro-duced an analysis of the decision time courseprior to the response, known as “gaze likelihoodanalysis.” This analysis plots, for each time binover a period prior to the response (where eachbin spans a certain number of samples of gazeposition), the likelihood that observers’ gaze

was directed toward the stimulus that was even-tually chosen (Figure 3a and b for an illustrationof gaze likelihood analysis). Shimojo et al.(2003) monitored gaze position while partici-pants made two-alternative forced choice (2-AFC) preference decisions between pairs offaces that were presented simultaneously onscreen. Gaze likelihood analysis revealed thatover the period just prior to the response therewas a progressively increasing bias in the like-lihood that observers’ gaze was directed towardthe chosen stimulus. This bias was dubbed ‘thegaze cascade effect,’ and led to the formulationof a specific hypothesis about the role of eyemovements in preference decisions (the GazeCascade model; discussed later in the context ofour experiment).

Simion and Shimojo (2006, 2007) considereda potential hurdle for interpreting the gaze cas-cade effect that is of general interest to processtracing research in decision making. They wereconcerned that the effect might partly reflect atendency to direct gaze to the chosen item at thepoint of decision, or even after the decision hadbeen (cognitively) resolved. Such a bias mightoccur while the response is being held in mem-ory, or while the motor response is beingprogrammed. Indeed, related issues were con-sidered by Russo and Leclerc (1994), who iden-tified processing events that are very near to theresponse, including review/verification justprior to the response where the decision wasnearly complete, or additional verification fol-lowing the response (e.g., ‘second-guessing’).Thus it is of general interest for process tracingresearch to dissociate components of the deci-sion process that occur over the duration of thedecision and prior to the response, reflecting theextraction and evaluation of information fromdecision alternatives, and those that are tied tothe announcement of the decision outcome.This issue is relevant for all process tracingmethods that monitor information search, but itis particularly important to consider in the con-text of eye movement recordings, as one mightexpect there to be a significant interval wherethe decision has been made, and the eye is still“waiting” for another effector (such as the hand)to announce the response (see Ball et al. (2003),for a related discussion in the context of re-search on problem solving and reasoning).

To rule out response-related explanations forthe gaze cascade effect, Simion and Shimojo


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

(2006) employed a sophisticated experimentalmanipulation that has only recently becomepossible with high fidelity eye movement mon-itoring. This technique, known as the gaze-contingent window, updates the display in realtime such that there is a window that is contin-uously centered on the viewer’s point of gaze.Stimulus information is visible inside the win-dow and it is masked outside the window,which restricts the viewer to processing stimu-lus information within the central region of vi-sion (for a review see Rayner, 1998, 2009; for areview of applied uses of gaze-contingent win-dow methodology see also Reingold, Loschky,McConkie, & Stampe, 2003). In the context ofdecision making research, the gaze-contingentwindow can be used to force the decision makerto process individual decision alternatives oneat a time, much like the deliberate sampling thatoccurs with information search displays. Simionand Shimojo (2006) expected that this mode ofviewing would substantially lengthen the deci-sion time course, and that if the bias could bedemonstrated a sufficient amount of time inadvance of the response, then it would be pos-sible to rule out response-related explanationsfor the effect. Indeed, they found that gaze-contingent window viewing mode greatly in-creased decision times compared with decisionsmade under conditions of free viewing, and thatin the gaze-contingent mode the gaze cascadeeffect began several seconds prior to the re-sponse, compared with �1.5 seconds prior tothe response in free viewing. Based on thisfinding, Simion and Shimojo argued that thebias in gaze likelihood could not be attributedto response-related factors alone, but ratherwas likely related to the ongoing decisionprocess.

However, Glaholt and Reingold (2009a)observed that in both two-alternative and eight-alternative forced-choice decisions (under prefer-ence and nonpreference decision instructions),there was a strong and significant tendency for thechosen alternative to be the last alternative viewedbefore the response. Based on this finding, it be-came clear that in order to rule out response-related explanations for any observed gaze bias,it was insufficient to demonstrate a bias a cer-tain amount of time before the response, butrather it was necessary at least to show a biasprior to the last alternative viewed during thedecision. This prompted the development of a

novel analysis, known as ‘dwell sequence anal-ysis’, which identifies the sequence of dwellsthat occur over the course of the decision, wherea dwell is a run of consecutive fixations on adecision alternative (similar to an individual“look” in an information search display). Ofnote, the dwell sequence analysis distinguishesbetween how often gaze is directed to a decisionalternative and how long gaze dwells on a de-cision alternative, two components that are con-founded in gaze likelihood analysis.

Glaholt and Reingold (2009a) found thattwo-alternative decisions (under both prefer-ence and nonpreference decision instructions;Figure 1a) were composed of very few dwells,making it difficult to depict the decision timecourse, and even harder to rule out responserelated explanations for observed biases. In ad-dition, Glaholt and Reingold (2009a) examinedthe effect of a gaze-contingent window manip-ulation on 2-AFC decisions (see Figure 1b) andreplicated the finding of Simion and Shimojo(2006) of a gaze bias that appeared much earlierin the decision time course. However, it wasfound that rather than extending the decisiontime course in terms of the number of dwells,the gaze-contingent window manipulationmerely extended the length of individual dwells.Hence even under these conditions it was diffi-cult to rule out response-related explanations forthe gaze bias effect. In search of a decision taskthat would produce a longer dwell sequence,Glaholt and Reingold (2009a) examined eight-alternative decisions (8-AFC; Figure 1c). The8-AFC decisions were found to be composed ofan extended sequence of dwells, over which twoseparate choice-related gaze biases were mani-fest (we use “gaze bias” generally to refer to thefinding of an eye movement measure that showsdifferentiation between the chosen item anditems that were not chosen). Specifically, thelast few dwells in a decision tended to have ahigh probability of being directed toward thechosen item (a bias in dwell frequency). Inaddition, dwells on the chosen item were longerthan dwells on other items, from the very firstdwell and throughout the decision period (a biasin dwell duration). Of note, both of these biaseswere manifest well in advance of the response,in temporal distance, but more importantly inthe unit of dwells. This constitutes clear evi-dence that these biases cannot be accounted forby response-related explanations, and instead


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

they likely reflect aspects of the decision pro-cess itself.

The dwell duration bias and the dwell fre-quency bias that emerged from this analysis also

exhibited certain dissociations that suggest dif-ferent processing roles. For one, the dwell du-ration bias was present from the beginning ofthe decision period while the dwell frequency

Figure 1. Examples of the decision scenarios employed by Glaholt and Reingold (2009a).Participants made decisions about photographic art images. (a) Two-alternative forced choice(2-AFC); (b) 2-AFC with gaze-contingent window. The window is continuously centeredabout the participant’s point of gaze; (c) eight alternative forced choice (8-AFC).


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

bias occurred only toward the end of the deci-sion period. As such, these biases might maponto different stages of the decision process,such as an early screening/orientation stage anda later evaluation/comparison stage as describedby Russo and Leclerc (1994). Second, in an-other study, Glaholt and Reingold (2009b) ma-nipulated exposure to decision alternatives bypreexposing decision makers to a subset of thealternatives prior to a multialternative decision.Dwell sequence analysis showed that while thedwell duration bias was quite sensitive to priorexposure of decision alternatives, the dwell fre-quency bias was entirely insensitive to this ma-nipulation. This provides further evidence thatthe dwell duration bias might reflect a processrelated to selective encoding of decision alter-natives while the bias in dwell frequency mightreflect an evaluation or comparison process thatoccurs after the alternatives have been initiallyencoded.

Additional evidence from eye movements forthe presence of a screening process during mul-tialternative visual decision tasks was reportedby Glaholt, Wu, and Reingold (2009, 2010). Forexample, Glaholt, Wu, and Reingold (2010)manipulated decision task instructions such thatparticipants chose between two sets of threeitems (set selection) or one out of six individualitems (item selection), where the items weredrawn from a set of images of everyday prod-ucts (belts, sunglasses, shirts, or shoes). Of im-portance, the stimulus displays were identicalbetween the two tasks, allowing for a directexamination of the impact of decision task in-structions on patterns of eye movements. Bycomparing the first half and the second half ofthe decision period, it was shown that under theitem selection instructions there was a signifi-cant reduction in the number of different itemsviewed from the first to the second half of thetrial. Of interest, in contrast, there was no dif-ference in the number of items viewed from thefirst to the second half of the trial under setselection instructions. This finding is consistentwith the idea that in the item selection task, ascreening process takes place whereby the num-ber of items being actively considered is nar-rowed over the course of the decision.

In summary, the body of existing researchemploying eye movements for process tracingin decision research is relatively small. How-ever, collectively this research has clearly dem-

onstrated that eye movement measures providean effective way to capture information search,and that eye movement measures are indeedsensitive to decision processes. Future researchmight seek to move beyond the more conven-tional use of eye movement monitoring in thisdomain, namely to observe decision makers’pattern of information search, and to take ad-vantage of the unique experimental manipula-tions that are possible with modern eye trackingtechnology. Specifically, eye tracking allowsfor powerful gaze-contingent manipulationsthat allow researchers to answer research ques-tions that might otherwise be difficult or impos-sible to address. In the following section wepresent an experiment that illustrates some ofthese interesting possibilities.

Experiment: Testing the Gaze CascadeHypothesis

In the present experiment we tested a specifichypothesis about the role of eye movements inpreference decisions. By employing a gaze-contingent manipulation of exposure to decisionalternatives, we were able to provide a directtest of the Gaze Cascade model proposed byShimojo and colleagues (Shimojo et al., 2003;Simion & Shimojo, 2006, 2007). As a second-ary goal, we further investigated the dissocia-tion between the dwell frequency and dwellduration biases that was documented previ-ously. In particular, Glaholt and Reingold(2009b) found that while the dwell duration biaswas present throughout the decision timecourse, and was quite sensitive to whether ornot decision alternatives had been exposed priorto the decision, the dwell frequency bias oc-curred later in the trial and was insensitive to themanipulation of stimulus exposure. Glaholt andReingold (2009b) speculated that the two biasesmight map onto different stages of the decisionprocess (following Russo & Leclerc, 1994),where the dwell duration bias might be relatedto a screening process that occurs during stim-ulus encoding, and the frequency bias mightreflect an evaluative process that occurs later inthe trial. The present manipulation providesconvergent evidence for this dissociation in thecontext of the analysis techniques developed byRusso and Leclerc (1994). In addition, this ex-periment demonstrates the utility of eye track-ing technology to provide experimental manip-


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

ulations (e.g., gaze-contingent display) that areimpossible using other techniques.

To explain the gaze cascade effect (describedin the previous section), Shimojo and col-leagues (Shimojo et al., 2003; Simion & Shi-mojo, 2006, 2007) proposed the Gaze Cascademodel. This model specifies two componentprocesses related to looking behavior that inter-act during preference decisions. The first pro-cess is preferential looking, where one tends tolook longer at the stimulus that one likes (Birch,Shimojo & Held, 1985). The second process isthe mere exposure effect, where merely lookingat a stimulus increases preference for that stim-ulus (Kunst-Wilson & Zajonc, 1980; Moreland& Zajonc, 1977, 1982; Zajonc, 1968). Shimojoet al. (2003) suggested that these two processescan combine to create a positive feedbackloop (dubbed a Gaze Cascade) that progres-sively increases the activation of one of thedecision options until it exceeds the thresholdfor response. This model represents a depar-ture from previous decision making research,in that rather than assuming that eye move-ments reflect the sampling and processing ofdecision information (i.e., informationsearch), the Gaze Cascade model holds thatgaze itself plays an active role in the decisionprocess. The model argues that for preferencedecisions, gaze both reflects the commitmentto a decision alternative (e.g., the preferencefor that alternative), but also actively in-creases the commitment to that decision al-ternative (Shimojo et al., 2003).

Glaholt and Reingold (2009b) investigatedpredictions derived from the Gaze Cascademodel. Specifically, the model makes the pre-diction that decision alternatives that are ex-posed for longer should be subject to a largergaze cascade effect, and hence be more likely tobe chosen, by way of the hypothesized positivefeedback loop between mere exposure and pref-erential looking. However, the model argues forstrong preference-specificity in this effect, andhence it should only hold for preference deci-sions and should be reduced or absent for non-preference decisions. In order to manipulate thedegree of exposure to decision alternatives, Gla-holt and Reingold (2009b) preexposed a subsetof the decision alternatives prior to an 8-AFCdecision (under preference or nonpreference in-structions). Consistent with our prior findings

(Glaholt & Reingold, 2009a), we observed abias in dwell duration on the chosen item overthe entire decision time course, and a bias in thefrequency of dwells on the chosen item in thelast few dwells in the trial. However, contrary tothe prediction of the model, we found that thebias in dwell duration was actually larger fordecision alternatives that were not preexposedthan for those that were, and that the bias indwell frequency was almost entirely insensitiveto the exposure manipulation. We also foundthat pattern of gaze biases was extremely simi-lar under preference and nonpreference decisioninstructions.

These findings are difficult to reconcile withthe decision process specified by the Gaze Cas-cade model. However, there remains a specificset of conditions where increased stimulus ex-posure might affect gaze biases in the way pre-dicted by the model. Our previous manipulationof stimulus exposure (Glaholt & Reingold,2009b) involved a central presentation of thepreexposed stimuli prior to the eight-alternativedecision. The Gaze Cascade model describes amechanism that is sensitive to differential stim-ulus exposure occurring during the ongoing de-cision. Consequently it remains possible that amanipulation of exposure to decision alterna-tives that occurs dynamically during the ongo-ing 8-AFC decision would produce results con-sistent with the gaze cascade hypothesis. In theexperiment that follows we manipulated stimu-lus exposure in a way that addresses this poten-tial concern. Using a gaze-contingent method-ology, we controlled the maximum stimulusexposure duration within individual dwells dur-ing 8-AFC decisions. When participants di-rected their gaze to a stimulus alternative, thestimulus was removed from the display(blanked) after either a short duration (200 ms)or a long duration (400 ms). Stimuli that hadbeen removed reappeared after the participantdirected their gaze to another stimulus alterna-tive. If the Gaze Cascade model is correct, stim-uli with the longer exposure duration shouldexhibit stronger biases in looking behavior, andshould be more likely to be chosen than stimuliwith the short exposure duration. Critically, themodel predicts that these exposure effectsshould be present in preference decisions andweak or absent in nonpreference decisions.


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Method

Participants

All 16 participants were undergraduate stu-dents at the University of Toronto Mississauga,and each received $10 for their participation.

Apparatus

The eye tracker employed in this researchwas SR Research Ltd. EyeLink 1000 system.Following calibration, gaze-position error wasless than 0.5°. Gaze position was sampled at1000 Hz. Stimulus displays were presented on a19-in. monitor. The participant’s monitor wasset to a resolution of 1600 � 1200 and a refreshrate of 85 Hz. Participants were seated 65 cmfrom the display and used a chinrest with a headsupport.

Materials and Design

Stimuli consisted of a set of 400 grayscaleimages of photographic art (see Glaholt & Re-ingold, 2009a, 2009b). For each participant,half of the images were used in the Preferencedecision and half were used in the nonprefer-ence control decision (Typicality decision). Wefound no correlation (across images) betweenthe number of times an image was chosen in thePreference decision and the number of times itwas chosen in the Typicality decision (r � .01,ns), and hence the two decisions were likely toinvolve different decision criteria, possibly re-quiring participants to sample different infor-mation from the stimuli. The order of the deci-sion tasks and the assignment of stimuli to eachdecision type were counterbalanced across par-ticipants. Within each decision task, the 200photographs were randomly divided into 25 setsof 8, where each set corresponded to an 8-AFCtrial. This process was then repeated yielding atotal of 50 trials per decision type.

Procedure

In the Preference task, the participant wasinstructed to select, from the eight alternatives,the image that he or she liked the most. In theTypicality task (nonpreference control), the par-ticipant had to select the image that he or shejudged to be most unusual (i.e., most out of the

ordinary, least typical). Participants were in-formed that upon viewing an image, it woulddisappear after a short time and that it wouldreappear after they had moved on to anotherimage, but that they should try to make the bestdecision they could.

At the beginning of each trial, the eight stim-uli were presented in a 3 � 3 array, where eachcell measured 8° � 8° of visual angle (400 �400 pixels). The middle cell was empty exceptfor a fixation circle. Direct foveal viewing timewas limited using a gaze-contingent display:when the participant’s gaze entered a gridsquare containing a stimulus, that stimulus wasremoved from the display (blanked) after eithera short duration (200 ms) or a long duration(400 ms) (Figure 2). In each trial, four of theeight alternatives were randomly assigned tothe short exposure duration and the other fourwere assigned to the long exposure duration.The blanked stimulus reappeared 80 ms afterthe participant’s gaze had moved to another gridsquare. Participants were instructed that havinghad reached a decision, they should look at thegray circle located in the center square and pressa button on a video game controller. Thiscaused the circle to turn green, which signaledthe participant that the selection-by-looking toolwas active, and that he or she should then fixatethe chosen item in order to select it. Participantsadvanced to the next trial by fixating the centralgrid square and pressing a button on the videogame controller.

Results and Discussion

Our analysis is divided into two sections. Inthe first section we tested predictions derivedfrom the Gaze Cascade model. To reiterate, theGaze Cascade model predicts that greater stim-ulus exposure should produce stronger biases inlooking behavior in preference decisions, andthese exposure effects should be weak or absentin nonpreference decisions. To quantify theseeffects, we employed analysis methods devel-oped by Shimojo and colleagues (gaze likeli-hood analysis; Shimojo et al., 2003; Simion &Shimojo, 2006, 2007) as well as techniques wedeveloped subsequently (dwell sequence analy-sis). In the second section, we investigated theapparent dissociation between the selectivity inthe placement of dwells (e.g., the dwell fre-quency bias) and selectivity expressed in the


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

duration of dwells (e.g., the dwell durationbias). To accomplish this, we adapted the anal-ysis technique introduced by Russo and Leclerc(1994) to separate the decision time course intostages based on the eye movement record. Inparticular, Russo and Leclerc argued that theonset of revisits to decision alternatives markeda transition to an evaluative stage of processing.Based on this model, we hypothesized that se-lectivity in the placement of dwells (i.e., thelikelihood of viewing the chosen item) wouldincrease greatly among revisit dwells comparedwith first encounters with a decision alternative.

Testing the Gaze Cascade Model

Prior to analyzing the eye movement data, weanalyzed the probability that long exposureitems were chosen over short exposure itemsand found that in the Preference task partici-pants chose the long exposure items signifi-cantly more often than the short exposure items(chance � 0.50; M � 0.54, p � .05). However,in contrast to the preference-specificity pre-dicted by the Gaze Cascade model, this effectwas also present in the Typicality task(chance � 0.50; M � 0.55, p � .05). Wespeculate that participants may be slightly lesslikely to extract enough information from theshort exposure items for them to be chosen,resulting in a slightly increased likelihood ofchoosing a long exposure item.

Following Shimojo et al. (2003) we producedgaze likelihood plots that display the proportionof time that gaze was directed at the chosen itemover the 2-s period just prior to the response.The gaze likelihood plots are shown in Figure 3(Preference in panel a, Typicality in panel b). Acomparison of the plots by exposure durationand decision type indicated a very similar pat-tern across conditions and across decision types,with all plots showing an increasing tendencyfor the eyes to be directed toward the chosen

Figure 2. Schematic diagram of the gaze-contingent lim-ited exposure manipulation employed for 8-alternativeforced choice (8-AFC) decisions in this experiment. Whenthe participant’s gaze entered a grid square containing adecision alternative, the alternative was blanked after a shortexposure duration (200 ms) or a long exposure duration(400 ms), and it reappeared 80 ms after gaze was directed toanother alternative.


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

item prior to the response. This runs contrary tothe prediction of the Gaze Cascade model thatthe bias in gaze likelihood should be more pro-nounced for long exposure items, and specifi-cally for preference decisions. There are,however, some slight differences in the gazelikelihood plots that are apparent. Over the lastsecond prior to the response, in both tasks, thelong exposure item appeared to reach a highervalue earlier in the time course, than the shortexposure item. On the other hand, the short

exposure items reached a higher final gaze like-lihood value than the long exposure items forboth tasks. These differences in the gaze likeli-hood plots are difficult to interpret. In previousstudies (Glaholt & Reingold, 2009a, 2009b), wefound that the chosen item tended to be the lastitem fixated in a trial. Indeed, the value of thefinal point in the gaze likelihood curve is theprobability of fixating the chosen item last. As aresult, the shape of the final portion of the gazelikelihood curve was strongly influenced by the

Figure 3. Testing the Gaze Cascade model: (1) Gaze likelihood analysis, plotting theproportion of time that gaze was directed toward the chosen item, going back 2 s from theresponse, as a function of stimulus exposure duration (Preference in panel a, Typicality inpanel b). We obtained 95% confidence intervals about each point in the time series using abootstrapping procedure (Efron & Tibshirani, 1994); (2) Analysis of dwell frequency bias;proportion of dwells directed toward the chosen item, as a function of stimulus exposureduration, for each of 4 bins prior to the response, and for the first dwell bin (preference in panelc, Typicality in panel d); (3) Analysis of dwell duration bias: Mean dwell duration for the chosenand other items, as a function of stimulus exposure duration, for each of 4 dwell bins prior to theresponse, and for the first dwell bin (Preference in panel e, Typicality in panel f).


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

duration of this last viewing of the chosen item(i.e., the last dwell). In the context of the presentresults, a possible additional source of variancein the appearance of gaze likelihood curvesmight lie in the differences in the duration ofdwells on long and short exposure items. Ifdwells on long exposure items were longer onaverage than dwells on short exposure items,the observed bias in the gaze likelihood plotsmight be expected to extend further back fromthe response for long exposure items comparedwith short exposure items. This illustrates theproblem with gaze likelihood analysis that wasmentioned earlier in the review section, namelythat it confounds the frequency with whichitems are viewed and the duration of thoseviews.

Because of these difficulties in interpretingthe appearance of the gaze likelihood plots, weexpanded on the gaze likelihood analysis byconducting a dwell sequence analysis. For eachtrial in each decision task, we identified a se-quence of dwells, where a dwell was defined asthe cumulative duration of all consecutive fix-ations from the moment the participant’s gazeenters a grid square containing an image, anduntil it exits that square. Preference and Typi-cality decisions were quite similar in terms oftotal number of dwells (Preference: M � 19.4,SD � 7.8; Typicality: M � 23.2, SD � 7.7),with the Typicality decisions having a slightlylarger number of dwells, t(15) � 2.57, p � .05.As shown in Figure 3 (c, d, e, and f), wecreated 4 different dwell position bins goingback from the response: the last two dwell po-sitions (Last), positions 3–4 prior to the re-sponse (�1), positions 5–6 prior to response(�2), and positions 7–8 prior to response (�3).In addition, we created a dwell position bincorresponding to the first two dwell positions inthe trial (First). For each bin, we computed theproportion of dwells directed toward the chosenitem (dwell frequency: panels c and d), and wealso computed the mean dwell duration fordwells on the chosen item and dwells on theother items (dwell duration: panels e and f).These measures were computed separately forlong exposure and short exposure items and foreach decision type.

To analyze dwell frequency, we conducted a2 � 2 � 4 analysis of variance (ANOVA) thatcrossed Decision Type (Preference, Typicality),Exposure Duration (Long, Short), and Dwell

Position Bin (�3, �2, �1, Last) as within-participant variables. The first dwell positionbin was analyzed in a separate 2 � 2 ANOVAcrossing Decision Type and Exposure Duration.In the first dwell position bin, no effects orinteractions reached significance, and in all con-ditions the likelihood of dwells being directedto the chosen item in the first bin did not differfrom chance (all ts � 1.76, all ps � 0.1). Con-sistent with the results of the gaze likelihoodanalysis, there was an increase in the bias indwell frequency toward the chosen item in thelast four dwell position bins, F(3, 45) � 57.56,MSE � 0.012, p � .001. However, unlike theprediction of the Gaze Cascade model the biasin dwell frequency was actually slightly largeroverall for items with short exposure durationthan for items with long exposure duration, F(1,15) � 6.55, MSE � 0.01, p � .05, and theinfluence of exposure on the gaze bias in dwellfrequency was not stronger in the Preferencetask than the Typicality task (F � 1). The onlyeffect of decision type on dwell frequency wasa trend toward a larger overall dwell frequencybias in the Typicality task, F(1, 15) � 4.26,MSE � 0.01, p � .057.

In analyzing dwell duration, we conducted a2 � 2 � 2 � 4 ANOVA that crossed DecisionType (Preference, Typicality), Choice (Chosen,Other), Exposure Duration (Long, Short), andDwell Position Bin (�3, �2, �1, Last). Thefirst dwell bin was analyzed in a separate 2 �2 � 2 ANOVA crossing Decision Type, Expo-sure Duration, and Choice. Our analysis of thefirst dwell position bin revealed a two-way in-teraction between Choice and Exposure Dura-tion, F(1, 15) � 11.72, MSE � 1.66 � 103, p �.01. Post hoc analyses confirmed that for thefirst dwell position bin, dwells on the chosenitem were significantly longer than dwells onother items when the items were long exposure,F(1, 15) � 10.64, MSE � 3.70 � 103, p � .01,but not when and they were short exposureitems (F � 1). In our previous studies (Glaholt& Reingold, 2009a, 2009b) we found a robustbias in dwell duration in the first dwell positionin the trial. Our present findings indicate thatthis may only occur if a sufficient amount ofstimulus information is available.

Our analysis of the last four dwell positionbins revealed that dwells on the chosen itemwere longer than dwells on other items, F(1,15) � 177.56, MSE � 1.22 � 104, p � .001.


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Consistent with the predictions of the Gaze Cas-cade model, the bias in dwell duration wassignificantly larger for long exposure itemscompared with short exposure items, F(1,15) � 10.76, MSE � 5.55 �103, p � .01, andthis interaction tended to be more pronounced inthe later dwell position bins, F(3, 45) � 2.80,MSE � 3.52 � 103, p � .051. However, wefound no evidence in support of the criticalprediction of the model that the effect of stim-ulus exposure duration on the gaze bias wouldbe stronger in the Preference task than in thenon-Preference control task (the three-way in-teraction between Exposure Duration, Choice,and Decision Type was not significant: F(1,15) � 1.34, MSE � 5.40 � 103, p � .25. Theonly evidence of task-specificity in the dwellduration bias appeared in a tendency for the biasto be larger overall in the Typicality task than inthe Preference task, F(1, 15) � 4.51,MSE � 3.98 � 103, p � .051.

Thus, together with our previous findings(Glaholt & Reingold, 2009b), we have failed todemonstrate the predicted effects of varyingstimulus exposure regardless of whether in-creased exposure duration occurred as a resultof prior exposure to decision alternatives, orthrough a relative increase during the ongoingdecision. Furthermore, in the present study andin both of our previous investigations of thegaze bias phenomenon (Glaholt & Reingold,2009a, 2009b) we have found no evidence ofthe preference-specificity that is central to theGaze Cascade model. Accordingly, we wouldargue that these findings rule out the Gaze Cas-cade model as a viable account of the observedbiases in looking behavior. Instead, such biasesappear to be a more general characteristic ofmultialternative visual decision making.

Gaze Bias in First Dwells and inRevisit Dwells

A second goal of the present study was tofurther investigate the apparent dissociation be-tween the bias in dwell frequency and the biasin the duration of dwells. In addition, we exam-ined the possible link between these biases inlooking behavior and the suggestion of multiplestages in multialternative decision making. Spe-cifically, Russo and Leclerc (1994) suggestedthat during multialternative decisions, the pointat which the decision maker begins to revisit de-

cision alternatives marks a transition in the deci-sion process from an early screening stage to alater evaluative stage where stimuli are compareddirectly (for a similar approach see Schotter,Berry, McKenzie, & Rayner, 2010). Accordingly,we contrasted two classes of dwells: first dwells,defined as dwells occurring in the period priorto the first refixation on a decision alternative ina trial, and revisit dwells which occurred later inthe trial and were directed toward stimulus al-ternatives that were previously fixated duringthat trial (on average, participants viewed 5.12different decision alternatives before revisitingany one of them). In an additional exploratoryanalysis, we examined the composition ofdwells in terms of individual fixations by com-puting the mean number of fixations, and themean fixation duration, for dwells on the chosenitem and dwells on other items. Each measurewas analyzed in a 2 � 2 � 2 � 2 within-subjects ANOVA that crossed Choice (Chosen,Other), Exposure Duration (200 ms, 400 ms),Encounter Type (First, Revisit), and DecisionType (Preference, Typicality). Given that therewas a remarkably similar pattern of findingsunder the Preference and Typicality decisioninstructions, the results are described collapsingacross Decision Type.

First we examined the bias in the placementof dwells (dwell frequency; Figure 4a). Whilethere was a very small but significant increase inthe probability of directing dwells to the chosenitem compared with other items during firstdwells, F(1, 15) � 10.20, MSE � 0.001, p �.01, this effect increased dramatically amongrevisit dwells, F(1, 15) � 95.79, MSE � 0.002,p � .001. The fact that the frequency of firstdwells is roughly equivalent across chosen andother items might indicate that for the stimuliand decision tasks that we employed, parafovealinformation is ineffective in guiding dwells to-ward promising decision alternatives early inthe trial. To investigate this further, we com-puted the average serial position of the firstdwell on the chosen item and the first dwell onother items and found that they did not differsignificantly (chosen � 5.01, other � 5.27,t(15) � 1.76, p � .1).

Next we analyzed dwell duration. As can beseen in Figure 4b, dwells on the chosen itemwere longer in duration than dwells on otheritems (all ts � 2.27, all ps � 0.05), and this


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

effect was larger for revisit dwells comparedwith first encounters, F(1, 15) � 65.34, MSE �893.56, p � .001, and for long exposure stimulithan for short exposure stimuli, F(1,15) � 18.61, MSE � 967.06, p � .001. Morespecifically, this pattern of effects is also pres-ent in mean fixation duration (Figure 4d) wherethe bias in mean fixation duration was larger forrevisit dwells compared with first encounters,F(1, 15) � 12.21, MSE � 249.83, p � .01, andfor long exposure stimuli compared with shortexposure stimuli, F(1, 15) � 5.96, MSE �

143.67, p � .05. A similar pattern was presentin the number of fixations per dwell (Figure 4c),where the bias was larger for revisit dwells thanfor first encounters, F(1, 15) � 33.80,MSE � 0.01, p � .001, and there was a trendtoward a larger bias for long exposure stimulicompared with short exposure stimuli, F(1,15) � 3.40, MSE � 0.02, p � .1. In summary,our results showed that the bias in dwell du-ration increases when more stimulus informa-tion is available within a dwell (i.e., longexposure vs. short exposure), and it also in-

Figure 4. Gaze bias in first dwells and revisit dwells: In each panel, measurements aredisplayed as a function of whether the dwell was directed to the chosen item or another item,whether the dwell was directed to a short (200 ms) or a long (400 ms) exposure stimulus, andwhether the dwell was among first dwells (i.e., prior to the first revisit) or a revisit dwell. Errorbars represent the standard error on the mean. (a) Proportion of dwells directed to the chosenitem, and to other items. (b) Mean dwell duration. (c) Number of fixations per dwell. (d) Meanfixation duration.


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

creases when the dwell constitutes a return toa previously viewed stimulus (i.e., first dwellsvs. revisit dwells). The lengthening of dwellsreflects both a lengthening of individual fix-ations and an increase in the number of fixa-tions per dwell.

The present findings and the findings fromour previous experiments establish a pattern ofan initial portion of the decision period wherethe dwell duration bias is present and the fre-quency bias is largely absent, and a later periodthat is marked by the emergence of the dwellfrequency bias and an increase in the dwellduration bias. As suggested previously (Glaholt& Reingold, 2009b), this dissociation betweenthe early and late portions of the decision periodmay map onto different stages of the decisionprocess. In particular, it has been proposed(Beach, 1993; Russo & Leclerc, 1994; Senter &Wedell, 1999; Wedell & Senter, 1997) that inmultialternative decisions participants employan early “screening” process where the degreeof encoding is not uniform across decision al-ternatives; rather, highly relevant alternativesare processed more deeply, and poor alterna-tives are processed to a lesser extent, or possiblyeven excluded from further processing. Consis-tent with this hypothesis, prior studies examin-ing eye movements have found evidence of anarrowing in the set of decision alternativesconsidered over course of the decision period(Glaholt, Wu, & Reingold, 2009, 2010). In thepresent study we found additional evidence forthe operation of such a screening process. Spe-cifically, we found that the dwell duration biasis sensitive to the amount of stimulus informa-tion that is available within a dwell (i.e., longvs. short exposure duration). When more stim-ulus information is available, we observed agreater differentiation in dwell duration be-tween the chosen and not chosen items. Thismight reflect an early stage of processing wheredecision alternatives are initially encoded, andrelevant alternatives (e.g., the chosen item) areencoded to a greater extent than poor alterna-tives. The later stage of the decision periodmight involve deeper evaluation of relevant al-ternatives, and direct comparisons between al-ternatives, reflected in an increased dwell dura-tion bias and an increase in the frequency ofdwells on the chosen item.

General Discussion

In this article, we reviewed the status of eyemovement recordings as a process tracing meth-odology in decision making research. Since thelate 1970s, eye tracking technology has becomean increasingly valuable method of monitoringdecision makers’ information search behavior.Early research demonstrated that eye movementmonitoring can provide a variety of processtracing measures that corroborate those ob-tained from the more traditional methods suchas information search displays and verbal pro-tocols. More recently, advances in eye trackingtechnology have afforded several unique advan-tages for this methodology. In particular, mod-ern eye tracking technology allows for ex-tremely precise measurement of the spatial andtemporal profile of decision makers’ informa-tion sampling. Of importance, in comparison toother methods, eye movement monitoring im-poses very few external demands on the deci-sion maker, and as such is likely to portraydecision processes as they occur naturally. Be-yond providing precise monitoring of informa-tion search, eye movement recordings can beused to manipulate the information presented tothe decision maker in gaze-contingent fashion.By updating the display based on the viewer’sgaze position, a variety of unique experimentalmanipulations are possible, such as precise con-trol over exposure to individual stimuli. As il-lustrated in the present experiment, this sort ofmanipulation can allow researchers to test spe-cific hypotheses about decision processes thatmight otherwise be difficult or impossible toaddress.

One potential problem with eye movementmonitoring as a process tracing method is thatwhen decision makers’ are allowed to freelyview a display, they may be able to processinformation outside of the focus of their gaze,which might undercut the ability of eye move-ment monitoring to capture the true pattern ofinformation search. However, as has been ar-gued elsewhere, the focus of gaze and the focusof attention tend to be tightly coupled duringnatural viewing. But moreover, eye tracking canbe applied to tackle this issue directly. For ex-ample, evidence of the processing of peripheralinformation can be derived from the pattern ofinformation sampling. In the present experi-ment, we found that the bias in the placement of


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

dwells did not appear until later in the decisionperiod, indicating that decision makers wereinitially unable to direct their gaze to relevantalternatives based on peripheral informationalone. Furthermore, the processing of informa-tion peripheral to the point of gaze can be re-stricted using the gaze-contingent window par-adigm (e.g., Glaholt & Reingold, 2009a; Simion& Shimojo, 2006). In addition to blocking pe-ripheral information, this technique can be usedto control the amount, and quality, of peripheralinformation available to the viewer (e.g., Mc-Conkie & Rayner, 1975; Pomplun, Reingold, &Shen, 2001a, 2001b; Rayner & Bertera, 1979;Reingold, Charness, Pomplun, & Stampe, 2001;Reingold et al., 2003), and similar methodsmight be applied in the context of process trac-ing research in decision making.

In addition to these methodological advan-tages, modern eye tracking technology has en-couraged researchers to ask new questionsabout the processes that occur during decisionmaking. For instance, Shimojo and colleaguesraised the possibility that eye movements mayhave an active role in preference decisions, be-yond the traditionally assumed role of informa-tion sampling. This hypothesis was motivatedby the discovery of the gaze cascade effect,which was observed in a fine-grained temporalanalysis of gaze behavior just prior to the re-sponse (gaze likelihood analysis). In order tofurther investigate this phenomenon, Glaholtand Reingold (2009a) developed an analysis ofdwell sequence, which revealed dissociable bi-ases in dwell frequency and dwell duration overthe decision time course.

Rather than reflecting a Gaze Cascade mech-anism, we interpreted these two biases in thecontext of prior process tracing research. Spe-cifically, the bias in dwell duration and the biasin dwell frequency might reflect different stagesof the decision process. It has been suggestedthat in multialternative decisions, because oflimits in information processing capacity it maybe inconvenient or impossible for decision mak-ers to completely encode (i.e., process holisti-cally) all of the alternatives and compare themsimultaneously (Ford et al., 1989; Payne, 1976;Payne et al., 1993). Consequently, decisionmakers may engage in a “screening” stage char-acterized by selective processing, where rele-

vant alternatives are processed in greater depth,and poor alternatives may be subject to shallowencoding or excluded from further processingaltogether (Beach, 1993; Russo & Leclerc,1994; Senter & Wedell, 1999; Wedell & Senter,1997). A key characteristic of such a screeningprocess is that stimuli are encoded to a differentdegree depending on their relevance to the de-cision task. The bias in dwell duration is con-sistent with differential encoding according totask relevance, and hence it might reflect theoperation of such a screening process duringmultialternative decisions. In contrast, the biasin dwell frequency might reflect a later evalua-tive stage of processing that involves the directcomparison of alternatives and results in ahigher frequency of dwells on the chosen item.Consistent with this possibility, in the presentexperiment and in prior work (Glaholt & Rein-gold, 2009b), we found the dwell duration biasto be quite sensitive to manipulations of stimu-lus exposure duration while the bias in dwellfrequency was not. Together these findings sug-gest that the dwell duration bias is somehowrelated to the encoding of decision alternatives.In contrast, the dwell frequency bias is likely tobe less related to the encoding of alternatives,and perhaps instead reflects another process(possibly postencoding) such as the evaluationand comparison of relevant alternatives. How-ever, further research is clearly required to testthese ideas.

At present, it is apparent that eye movementrecordings might be best employed in a com-bined approach with other process tracing meth-ods. As has been argued by Riedl et al. (2008)with regards to information search display par-adigms, convergent evidence from multipletechniques (including verbal reports, analysis ofoutcomes, etc.) may be required to specificallyidentify the nature of the cognitive processesthat occur during decision making. In this re-gard, there is a possible avenue of future re-search that has yet to be explored, but that hasthe potential to provide additional informationabout decision processes. Recently, there hasbeen an explosion of research in the area ofdecision neuroscience (cf., O’Doherty &Bossaerts, 2008; Sanfey, 2007; for reviews seeRangel, Camerer, & Montague, 2008; andGlimcher, 2003). In line with the process trac-ing approach, decision neuroscience seeks todetect cognitive processes that occur over the


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

decision time course prior to the final behavioralresponse. Modern eye tracking technology isnow available to sample gaze position at a ratethat matches or exceeds the rate of sampling ofbrain activity available with current neuroimag-ing methods. In particular, eye movement mon-itoring may be especially useful in providingtemporal markers of decision processes forwhich neural activity can then be observed inthe neuroimaging record. This combinedmethod has recently been successfully em-ployed in the study of visual cognition, andholds considerable promise as a future processtracing method in decision research.

In summary, our review indicates that eyemovement monitoring is poised to be an increas-ingly valuable process tracing technique in thenext wave of decision making research. Recentadvances in eye tracking technology have broughtabout a dramatic improvement in the quality ofeye movement data and in the ecological validityof this methodology. Eye movement monitoringnow affords a variety of experimental manipula-tions, and in particular, rigorous experimental con-trol can be achieved by the use of gaze-contingenttechniques. Finally, the combination of eye move-ments and other process tracing techniques havestrong potential for future research.

References

Abelson, R. P., & Levi, A. (1985). Decision makingand decision theory. In G. Lindzey & E. Aronson(Eds.), Handbook of social psychology (Vol. 1):Theory and method (pp. 231–309). New York:Random House.

Ball, C. (1997). A comparison of single-step andmultiple-step transition analyses of multiattributedecision strategies. Organizational Behavior andHuman Decision Processes, 69(3), 195–204.

Ball, L. J., Lucas, E. J., Miles, J. N. V., & Gale, A. G.(2003). Inspection times and the selection task:What do eye movements reveal about relevanceeffects? The Quarterly Journal of ExperimentalPsychology, 56A, 1053–1077.

Ball, L. J., Phillips, P., Wade, C. N., & Quayle, J. D.(2006). Effects of belief and logic on syllogisticreasoning. Experimental Psychology, 53, 77–86.

Beach, L. R. (1993). Broadening the definition ofdecision making: The role of prechoice screeningof options. Psychological Science, 4, 215–220.

Bee, N., Prendinger, H., Andre, E., & Ishizuka, M.(2006, September). Automatic preference detec-tion by analyzing the gaze “cascade effect”. Pro-

ceedings of the 2nd Conference on Communicationby Gaze Interaction, Turin, Italy.

Billings, R. S., & Marcus, S. A. (1983). Measures ofcompensatory and noncompensatory models of de-cision behavior: Process tracing versus policy cap-turing. Organizational Behavior and Human Per-formance, 31, 331–352.

Birch, E. E., Shimojo, S., & Held, R. (1985). Pref-erential looking assessment of fusion and stereop-sis in infants aged 1 to 6 months. InvestigativeOphthalmology and Visual Science, 26, 366–370.

Day, R., Lin, C., Huang, W., & Chuang, S. (2009).Effects of music tempo and task difficulty onmulti-attribute decision making: An eye-trackingapproach. Computers in Human Behavior, 25,130–143.

Duchowski, A. T. (2002). A breadth-first survey ofeye-tracking applications. Behavior ResearchMethods, Instruments, & Computers, 34, 455–470.

Efron, B., & Tibshirani, R. J. (1994). An Introductionto the bootstrap. Boca Raton, FL: Chapman &Hall.

Eger, N., Ball, L. J., Stevens, R., & Dodd, J. (2007).Cueing retrospective verbal reports in usabilitytesting through eye-movement replay. chap. inL. J. Ball, M. A. Sasse, C. Sas, T. C. Ormerod, A.Dix, P. Bagnall, & T. McEwan (Eds.), People andcomputers XXI – HCI. . . but not as we know it:Proceedings of HCI 2007. Swindon: The BritishComputer Society.

Ehmke, C., & Wilson, S. (2007). Identifying webusability problems from eye-tracking data. In Pro-ceedings of the 21st British CHI Group AnnualConference on HCI 2007: People and ComputersXXI: HCL . . . but not as we know it. Vol. 1. (pp.12). University of Lancaster, United Kingdom:The British Computer Society.

Einhorn, H. J., Kleinmuntz, D. N., & Kleinmuntz, B.(1979). Linear regression and process-tracingmodels of judgment. Psychological Review, 86,465–485.

Ellis, J. J., Glaholt, M. G., & Reingold, E. M. (inpress). Eye movements reveal solution knowledgeprior to insight. Consciousness and Cognition.

Ford, J. K., Schmitt, N., Schechtman, S. L., Hults,B. M., & Doherty, M. L. (1989). Process tracingmethods: Contributions, problems, and neglectedresearch questions. Organizational Behavior andHuman Decision Processes, 43, 75–117.

Glaholt, M. G., & Reingold, E. M. (2009a). The timecourse of gaze bias in visual decision tasks. VisualCognition, 17, 1228–1243.

Glaholt, M. G., & Reingold, E. M. (2009b). Stimulusexposure and gaze bias: A further test of the gazecascade model. Attention, Perception & Psycho-physics, 71, 445–450.


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Glaholt, M. G., Wu, M., & Reingold, E. M. (2009).Predicting preference from fixations. PsychNologyJournal, 7, 141–158.

Glaholt, M. G., Wu, M., & Reingold, E. M. (2010).Evidence for top-down control of eye movementsduring visual decision making. Journal of Vision,10(5):15, 1–10.

Glimcher, P. W. (2003). The neurobiology of visual-saccadic decision making. Annual Review of Neu-roscience, 26, 133–179.

Goldberg, J. H., Probart, C. K., & Zak, R. E. (1999).Visual search of food nutrition labels. Human Fac-tors, 41, 425–437.

Goldberg, M. E., & Wurtz, R. H. (1972). Activity ofsuperior colliculus in behaving monkey: 1. Visualreceptive fields of single neurons. Journal of Neu-rophysiology, 35, 542–559.

Hansen, J. P. (1991). The use of eye mark recordingsto support verbal retrospection in software testing.Acta Psychologica, 76, 31–49.

Harte, J. M., & Koele, P. (1995). A comparison ofdifferent methods for the elicitation of attributeweights: Structural modeling, process tracing, andself-reports. Organizational Behavior and HumanDecision Processes, 64, 49–64.

Hoffman, J. E. (1998). Visual attention and eyemovements. In H. Pashler (Ed.), Attention (pp.119–150). Hove, United Kingdom: PsychologyPress.

Johnston, C. W., & Diller, L. (1986). Exploratory eyemovements and visual hemi-neglect. Journal ofClinical and Experimental Neuropsychology, 8,93–101.

Kunst-Wilson, W. R., & Zajonc, R. B. (1980). Af-fective discrimination of stimuli that cannot berecognized. Science, 207, 557–558.

Kustov, A. A., & Robinson, D. L. (1996). Sharedneural control of attentional shifts and eye move-ments. Nature, 384, 74–77.

Levin, I. P., Huneke, M. E., & Jasper, J. D. (2000).Information processing at successive stages of de-cision making: Need for cognition and inclusion-exclusion effects. Organizational Behavior andHuman Decision Processes, 82, 171–193.

Lohse, G. L., & Johnson, E. J. (1996). A comparisonof two process tracing methods for choice tasks.Organizational Behavior and Human DecisionProcesses, 68, 28–43.

McConkie, G. W., & Rayner, K. (1975). The span ofthe effective stimulus during a fixation in reading.Perception & Psychophysics, 117, 578–586.

Mohler, C. W., & Wurtz, R. (1976). Organization ofmonkey superior colliculus: Intermediate layercells discharging before eye movements. Journalof Neurophysiology, 39, 722–744.

Moreland, R. L., & Zajonc, R. B. (1977). Is stimulusrecognition a necessary condition for the occur-

rence of exposure effects? Journal of Personalityand Social Psychology, 35, 191–199.

Moreland, R. L., & Zajonc, R. B. (1982). Exposureeffects in person perception: Familiarity, similar-ity, and attraction. Journal of Experimental SocialPsychology, 18, 395–415.

O’Doherty, J. P., & Bossaerts, P. (2008). Toward amechanistic understanding of human decisionmaking: Contributions of functional neuroimag-ing. Current Directions in Psychological Sci-ence, 17, 119–123.

Payne, J. W. (1976). Task complexity and contingentprocessing in decision making: An informationsearch and protocol analysis. Organizational Be-havior and Human Performance, 16, 366–387.

Payne, J. W., Bettman, J. R., & Johnson, E. J. (1993).The adaptive decision maker. New York: Cam-bridge University Press.

Payne, J. W., Braunstein, M. L., & Carroll, J. S.(1978). Exploring predecisional behavior: Analternative approach to decision research. Organi-zational Behavior and Human Performance, 22,17–44.

Pieters, R., Rosenbergen, E., & Wedel, M. (1999).Visual attention to repeated print advertising: Atest of scanpath theory. Journal of Marketing Re-search, 36, 424–438.

Pieters, R., & Warlop, L. (1999). Visual attentionduring brand choice: The impact of time pressureand task motivation. International Journal of Re-search in Marketing, 16, 1–16.

Pieters, R., & Wedel, M. (2004). Attention captureand transfer in advertising: Brand, pictorial, andtext-size effects. Journal of Marketing, 68, 36–50.

Pieters, R., & Wedel, M. (2007). Goal control ofattention to advertising: The Yarbus implication.Journal of Consumer Research, 34, 224–233.

Pomplun, M., Reingold, E. M., & Shen, J. (2001a).Investigating the visual span in comparativesearch: The effects of task difficulty and dividedattention. Cognition, 81, B57–B67.

Pomplun, M., Reingold, E. M., & Shen, J. (2001b).Peripheral and parafoveal cueing and masking onsaccadic selectivity in a gaze-contingent windowparadigm. Vision Research, 41, 2757–2769.

Posner, M. I., Snyder, C. R. R., & Davidson, B. J.(1980). Attention and the detection of signals.Journal of Experimental Psychology: General,109, 160–174.

Radach, R., Lemmer, S., Vorstius, C., Heller, D., &Radach, K. (2003). Eye movements in the process-ing of print advertisements. In J. HyOna, R.Radach, & D. Heller (Eds.), The mind’s eye: Cog-nitive and applied aspects of eye movement re-search (pp. 609 – 632). Amsterdam: North-Holland.

Rangel, A., Camerer, C., & Montague, P. R. (2008).A framework for studying the neurobiology of


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

value-based decision making. Nature ReviewsNeuroscience, 9, 545–556.

Rayner, K. (1975). The perceptual span and peripheralcues in reading. Cognitive Psychology, 7, 65–81.

Rayner, K. (1998). Eye movements in reading andinformation processing: 20 years of research. Psy-chological Bulletin, 102, 21–38.

Rayner, K. (2009). Eye movements and attention inreading, scene perception, and visual search. TheQuarterly Journal of Experimental Psychol-ogy, 62, 1457–1506.

Rayner, K., & Bertera, J. H. (1979). Reading withouta fovea. Science, 206(4417), 468–469.

Rayner, K., Miller, B., & Rotello, C. M. (2008). Eyemovements when looking at print advertisements:The goal of the viewer matters. Applied CognitivePsychology, 22, 697–707.

Rayner, K., Rotello, C. M., Stewart, A. J., Keir, J., &Duffy, S. A. (2001). Integrating text and pictorialinformation: Eye movements when looking at printadvertisements. Journal of Experimental Psychol-ogy: Applied, 3, 219–226.

Rehder, B., & Hoffman, A. B. (2005a). Eyetrackingand selective attention in category learning. Cog-nitive Psychology, 51, 1–41.

Rehder, B., & Hoffman, A. B. (2005b). Thirty-something categorization results explained: Selec-tive attention, eyetracking, and models of categorylearning. Journal of Experimental Psychology:Learning, Memory, and Cognition, 31, 811–829.

Reingold, E. M., Charness, N., Pomplun, M., &Stampe, D. M. (2001). Visual span in expert chessplayers: Evidence from eye movements. Psycho-logical Science, 12, 49–56.

Reingold, E. M., Loschky, L. C., McConkie, G. W.,& Stampe, D. M. (2003). Gaze-contingent multi-resolutional displays: An integrative review. Hu-man Factors, 45, 307–328.

Reisen, N., Hoffrage, U., & Mast, F. W. (2008).Identifying decision strategies in a consumerchoice situation. Judgment and Decision Mak-ing, 3, 641–658.

Riedl, R., Brandstatter, E., & Roithmayr, F. (2008).Identifying decision strategies: A process and out-come-based classification method. Behavior Re-search Methods, 40, 795–807.

Rosen, L. D., & Rosenkoetter, P. (1976). An eyefixation analysis of choice and judgment with mul-tiattribute stimuli. Memory & Cognition, 4, 747–752.

Russo, J. E. (1978). Eye fixations can save the world:A critical evaluation and a comparison betweeneye fixations and other information processingmethodologies. Advances in Consumer Re-search, 5, 561–570.

Russo, J. E., & Dosher, B. A. (1983). Strategies formultiattribute binary choice. Journal of Experi-

mental Psychology: Learning, Memory, and Cog-nition, 9, 676–696.

Russo, J. E., Johnson, E. J., & Stephens, D. L. (1989).The validity of verbal protocols. Memory & Cog-nition, 17, 759–769.

Russo, J. E., & Leclerc, F. (1994). An eye-fixationanalysis of choice processes for consumer nondu-rables. Journal of Consumer Research, 21, 274–290.

Russo, J. E., & Rosen, L. D. (1975). An eye fixationanalysis of multialternative choice. Memory &Cognition, 3, 267–276.

Sanfey, A. G. (2007). Decision neuroscience: Newdirections in studies of judgment and decisionmaking. Current Directions in Psychological Sci-ence, 16, 151–155.

Schotter, E. R., Berry, R. W., McKenzie, C. R., &Rayner, K. (2010). Gaze bias: Selective encodingand liking effects. Visual Cognition, 18, 1113–1132.

Selart, M., Kuvaas, B., Boe, O., & Takemura, K.(2006). The influence of decision heuristics andoverconfidence on multiattribute choice: A pro-cess-tracing study. European Journal of CognitivePsychology, 18, 437–453.

Senter, S. M., & Wedell, D. H. (1999). Informationpresentation constraints and the adaptive decisionmaker hypothesis. Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 25,428–446.

Shimojo, S., Simion, C., Shimojo, E., & Scheier, C.(2003). Gaze bias both reflects and influence pref-erence. Nature Neuroscience, 6, 1317–1322.

Simion, C., & Shimojo, S. (2006). Early interactionsbetween orienting, visual sampling and decisionmaking in facial preference. Vision Research, 46,3331–3335.

Simion, C., & Shimojo, S. (2007). Interrupting thecascade: Orienting contributes to decision makingeven in the absence of visual stimulation. Percep-tion & Psychophysics, 69, 591–595.

Sutterlin, B., Brunner, T. A., & Opwis, K. (2008).Eye-tracking the cancellation and focus model forpreference judgments. Journal of ExperimentalSocial Psychology, 44, 904–911.

Svenson, O. (1979). Process descriptions of decisionmaking. Organizational Behaviour and HumanPerformance, 23, 86–112.

Svenson, O. (1996). Decision making and the searchfor fundamental psychological regularities: Whatcan be learned from a process perspective? Orga-nizational Behavior and Human Decision Pro-cesses, 65, 252–267.

van Gog, T., Paas, F., van Merrienboer, J. J. G., &Witte, P. (2005). Uncovering the problem-solvingprocess: Cued retrospective versus concurrent andretrospective reporting. Journal of ExperimentalPsychology: Applied, 11, 237–244.


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

van Raaij, F. W. (1977). Consumer information pro-cessing for different information structures andformats. Advances in Consumer Research, 4, 176–184.

Walker, R., & Young, A. W. (1996). Object-basedneglect: An investigation of the contributions ofeye movements and perceptual completion. Cor-tex, 32, 279–295.

Wedel, M., & Pieters, R. (2000). Eye fixations onadvertisements and memory for brands: A modeland findings. Marketing Science, 19, 297–312.

Wedel, M., & Pieters, R. (2008). Visual marketing:From attention to action. New York: Erlbaum,Taylor and Francis Group.

Wedel, M., Pieters, R., & Liechty, J. (2008). Atten-tion switching during scene perception: How goalsinfluence the time course of eye movements across

advertisements. Journal of Experimental Psychol-ogy: Applied, 14, 129–138.

Wedell, D. H., & Senter, S. M. (1997). Looking andweighting in judgment and choice. OrganizationalBehaviour and Human Decision Processes, 70,41–64.

Wurtz, R., & Mohler, C. W. (1976). Organization ofmonkey superior colliculus: Enhanced visual re-sponse of superficial layer cells. Journal of Neu-rophysiology, 39, 745–765.

Zajonc, R. B. (1968). Attitudinal effects of mereexposure. Journal of Personality and Social Psy-chology, 9(2P2), 1.

Received June 14, 2009Revision received April 9, 2010

Accepted April 14, 2010 �

Showcase your work in APA’s newest database.

Make your tests available to other researchers and students; get wider recognition for your work.

“PsycTESTS is going to be an outstanding resource for psychology,” said Ronald F. Levant, PhD. “I was among the first to provide some of my tests and was happy to do so. They will be available for others to use—and will relieve me of the administrative tasks of providing them to individuals.”

Visit http://www.apa.org/pubs/databases/psyctests/call-for-tests.aspx to learn more about PsycTESTS and how you can participate.

Questions? Call 1-800-374-2722 or write to [email protected] since PsycARTICLES has a database been so eagerly anticipated!


This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

eye movement monitoring as a process tracing methodology ......process tracing methods such as...

Documents