Beauty is better pursued: Effects of attractiveness inmultiple-face tracking

Chang Hong Liu1 and Wenfeng Chen2

1Department of Psychology, University of Hull, Hull, UK2State Key Laboratory of Brain and Cognitive Science, Institute of Psychology, Chinese Academy of Sciences,Beijing, China

Using the multiple-object tracking paradigm, this study examines how spontaneous appraisal for facialbeauty affects distributed attention to multiple faces in dynamic displays. Observers tracked attractivefaces more effectively than unattractive faces in this task. Tracking performance was only affected bytarget attractiveness, suggesting an absence of appraisal for distractor attractiveness. Attractive malefaces also produced stronger binding of face identity and location for female participants. Together,the results suggest that facial attractiveness was appraised during tracking even though this was task irre-levant. Contrary to the theory that multiple-object tracking is driven by encapsulated low-level vision,our results show that the content of target representation is not only penetrable by social cognition butalso modulates the course of tracking operations.

Keywords: Attractiveness; Multiple-face tracking; Attention.

It is well known that human observers pay greaterattention to faces than to nonface objects (e.g., Ro,Russell, & Lavie, 2001Q1 ). Moreover, some facestimuli attract more attention than others. Apartfrom certain facial expressions such as fear andanger (Palermo & Rhodes, 2007), attractive facesalso trigger greater attention (Sui & Liu, 2009).Popular culture often suggests that a mere glanceof a face elicits spontaneous appraisal of attractive-ness. Humans may be predisposed to direct atten-tion to attractive faces because of their biologicaland social significance. There is well-established

evidence that even newborn babies look at attractivefaces longer (e.g.,Geldart,Maurer,&Carney, 1999;Langlois et al., 1987; Samuels, Butterworth,Roberts, Graupner, & Hole, 1994; Slater et al.,1998). Adults also tend to spend more timelooking at beautiful faces (Aharon et al., 2001;Kranz & Ishai, 2006), which are known to activatedopaminergic regions in the brain that are stronglylinked to the reward system (Kampe, Frith, Dolan,& Frith, 2001).

Appraisal of facial beauty appears to depend onhowwell a face resembles the average in a population

PQJE624186 TECHSET COMPOSITION LTD, SALISBURY, U.K. 9/30/2011

Correspondence should be addressed to ChangHong Liu, Department of Psychology, University of Hull, CottinghamRoad, Hull,

HU6 7RX, UK, or Wenfeng Chen, Institute of Psychology, Chinese Academy of Sciences, 4A Datun Road, Chaoyang District,

Beijing, China. E-mail: C.H.Liu@hull.ac.uk or ChenWF@psych.ac.cn

This research was supported by grants from The Royal Society, KC Wong Foundation, and 973 Program (2011CB302201). We

thank Professor David Perrett for offering the face stimuli and Emma Medford for her thoughtful comments on an earlier version of

this manuscript. We especially thank the reviewers for their constructive comments and suggestions.

# 0000 The Experimental Psychology Society 1http://www.psypress.com/qjep DOI:10.1080/17470218.2011.624186

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY

0000, 00 (0), 1–12

5

10

15

20

25

30

35

40

45

(Langlois & Roggman, 1990). Although symmetryalso plays a role in facial beauty (e.g., Grammer &Thornhill, 1994), there is evidence that it is lessimportant than averageness (Rhodes, Sumich, &Byatt, 1999; also see Rhodes, 2006, for an extensivereview). Hence attentional capture by attractivefaces may rely on preattentive computation of aver-ageness in face stimuli.

However, little is known about how facial attrac-tiveness affects attentional mechanisms. This issuemay be addressed through the existing attentionparadigms. For example, Maner, Gailliot, andDeWall (2007) employed a dot-probe paradigm toassess the effect of attractiveness on disengagingattention. Consistent with their hypothesis, theirobservers were slower at disengaging attention fromattractive female faces than from average-lookingfemale or male faces. In another study, Maner,Gailliot, Rouby, and Miller (2007) used the samemethod and found that participants fixated onhighly attractive faces within the first half secondand took longer to stop looking at them whenasked to shift their attention away. In a differentapproach, Sui and Liu (2009) investigated whetherappraisal for attractiveness occurs when spatial atten-tion has already been directed elsewhere. Theyemployed a spatial cueing task, where participantswere asked to judge the orientation of a cued targetpresented to the left or right visual field while ignor-ing a task-irrelevant face image flashed in the oppo-site field. They found that the presence of attractivefaces significantly lengthened task performance.This suggests that facial beauty can automaticallycompetewith an ongoing task for attention, althoughthis does not necessarily mean that appraisal of facialbeauty is amandatory process (see Schacht,Werheid,&Sommer, 2008). The analysis of eyemovements inSui and Liu’s study revealed that the effect of facialattractiveness was not due to foveal fixation on thetarget or face stimuli; hence there is a strong possi-bility that facial attractiveness can be detectedoutside the fovea. This has been confirmed by arecent investigation, where discrimination of facialbeauty was measured at several eccentricities (Guo,Liu, & Roebuck, 2011).

Both dot-probe and cueing tasks require focusedattention, where attention is directed to a single

spatial location. However, attention in reality canbe distributed to several different targets or spatiallocations. For instance, a group activity requiresattention to be paid to multiple individuals andlocations. The main purpose of this research wasto investigate the effect of facial attractiveness ondistributed attention. Another motivation of thisresearch comes from observation that spontaneousappraisal of attractiveness is rapid and transient(Schacht et al., 2008). It remains unclear whethersuch an appraisal only manifests in brief attentionalshifts. Hence the second purpose of this study wasto investigate whether a similar kind of automaticappraisal for facial attractiveness occurs continu-ously when sustained attention is distributedamong faces in multiple locations.

Pylyshyn and Storm (1988) designed a multiple-object tracking paradigm (MOT) to study howhuman observers maintain attention on multipleobjects across space and time. In recent years, theissue of content addressability in object tracking ishotly debated. This issue was first raised byPylyshyn (2004) who found that observers wereremarkably poor at identifying the features or iden-tity of correctly tracked objects. This phenomenondemonstrates a poor binding of object features andlocation. Pylyshyn’s explanation is that MOT isimplemented by early or low-level vision, wherethe information about individual identity is encap-sulated and inaccessible from higher level cogni-tion. His theory assumes that early vision picksout a small number of objects while ignoring theirvisual properties. Because of this, object identitydifferentiated by visual properties is not coded oraccessible from higher level cognitive processeseven when the objects with those properties areattended. However, recent evidence has challengedthis position by showing that tracking is contentaddressable (Horowitz et al., 2007).

This debate has an important implication forthe present study. If the information content oftracked faces is not processed, tracking perform-ance should not be affected by facial attractive-ness. It is known that information about faceidentity (i.e., individuation of face stimuli by fea-tures and configurations) is not completely lost intracking tasks (Ren, Chen, Liu, & Fu, 2009).

2 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0)

LIU AND HEN

50

55

60

65

70

75

80

85

90

If representations of the tracked faces are to someextent content addressable, there is a chance thatother high-level information such as facial attrac-tiveness is also available for processing. If proces-sing of such information is spontaneous, thiscould in turn affect multiple-face tracking. Theexperiments in this study were designed to testthis hypothesis. The outcome should reveal theextent to which a high-level social cognition canpenetrate and influence the low-level visual per-ception. Appraisal for facial beauty requireshigh-level vision, because a main criterion forbeauty—averageness—cannot be determinedsolely from low-level visual features such as lumi-nance, contrast, and size. Computing averagenessrequires identifying the landmarks of higher levelfacial features such as the eyes, the mouth, andthe face shape. The operations for image normal-ization, alignment, and comparison with theaverage in the hypothetical multidimensionalface space also require high-level vision. Toensure that our results were controlled for thelow-level contribution, we carefully scaled all ourtracking stimuli to the same size, luminance,and contrast.

Other high-level information accompanyingappraisal of facial attractiveness could also contrib-ute to multiple-face tracking performance. AsDion, Berscheid, and Walster (1972) showed intheir seminal study, beautiful people are generallyperceived to possess more positive attributes.They called this the “beauty-is-good” stereotype.Many studies have since produced evidence forthis positive association between attractivenessand goodness (see Eagly, Ashmore, Makhijani, &Longo, 1991, and Langlois et al., 2000, forreviews). It is also referred to as the halo effect ofattractiveness. This well-established duality ofbeauty could predict a similar effect on trackingbased on attractive and valence measures.Moreover, if the reward system is involved in per-ception of attractive faces (Kampe et al., 2001),arousal could also be responsible for any effect ofattractiveness on tracking performance. Toaddress these questions, we compared whethervalence and arousal affect tracking performance inthe same way as attractiveness.

EXPERIMENT 1

The main purpose of Experiment 1 was to investi-gate how target and distractor attractiveness mayaffect tracking performance differently. Given theprior finding that the target face identity is contentaddressable (Ren et al., 2009), we hypothesizedthat the target faces could be automatically appraisedfor their attractiveness. Attractive distractors, on theother hand, could impair tracking performance ifthey were also appraised for attractiveness.

Method

ParticipantsA total of 25 undergraduate students (19 females)from the University of Hull participated in thisstudy. The age of the participants ranged from 18to 28 years (Mdn= 19). All participants hadnormal or corrected-to-normal vision.

StimuliThe face database was obtained from the Universityof St. Andrews. It contains 702 frontal-viewCaucasian faces with no external features (hairand clothing). The size of the faces was normalizedaccording to the face width. The resulting imagemeasured 3.0× 3.8 cm (2.8× 3.6°) on screen. Allimages were scaled to the same mean luminanceand root-mean-square contrast.

All faces in the database were rated by 19 raters(aged between 18 and 29 years, 12 females) forattractiveness on a 7-point scale. To contrast theeffect of attractiveness, only the 149 most attractiveand the 132 least attractive faces were used in thisexperiment. The mean ratings for the two groupsof faces were 3.91 (3.96 for females, 3.72 formales, SD= 0.39) and 2.39 (2.35 for females,2.46 for males, SD= 0.38), respectively. Thesewere significantly different from each other(p, .001). Slightly more than half (58.3%) of thefaces were females. Each attractive or unattractiveface was only used once for each participant; 80were randomly chosen from the pool of attractivefaces, and another 80 were randomly chosen fromthe pool of unattractive faces.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 3

ATTRACTIVENESS AFFECTS MULTIPLE-FACE TRACKING

95

100

105

110

115

120

125

130

135

We performed an additional norming study onthe face stimuli to measure how attractiveness isrelated to valence and arousal. Twenty undergradu-ate students, aged between 18 and 33 years (M=21.1, SD= 4.4), rated a total of 452 faces forvalence and arousal. The image set consisted ofthe attractive and unattractive faces in this exper-iment as well as the additional 171 average facesused in Experiment 3. The faces were rated oneat a time. The presentation order was random.To avoid response bias, the raters were notinformed until the end of the experiment that therating data would be used to study facial attractive-ness. The details of our procedure were identical tothose for the International Affective Picture System(Lang, Bradley, & Cuthbert, 2008). A 9-pointrating scale was used, where 1 represented comple-tely unhappy/calm, whereas 9 completely happy/excited, respectively, for the valence and arousaljudgements. Table 1 shows the mean ratingresults for each face category. To determinewhether valence and arousal ratings vary accordingto levels of attractiveness, we conducted separateanalyses of variance (ANOVAs) for the two vari-ables. The main effect of valence was significant,F(1, 449)= 43.24, MSE= 19.41, p, .001,η2p= .17, where the valence for attractive faceswas higher than that for average-looking and unat-tractive faces (p, .001), and the valence foraverage-looking faces was higher than that forunattractive faces (p, .001). On the other hand,the main effect of arousal was not significant, F(1, 449)= 1.49, MSE= 0.41, p= .23, η2p= .01.

The tracking stimuli were displayed on a 21′′

monitor (SONY Trinitron, GDM-F520). Acentral square area with 22.8× 22.8° of visualangle was designated for stimulus presentation.The background colour of the display was grey.

E-Prime (Version 1.2) was used to generate thedynamic tracking and still displays and to controlthe flow of the experiment.

DesignWe employed a within-subject design. The inde-pendent variables were target attractiveness (attrac-tive vs. unattractive), distractor attractiveness(attractive vs. unattractive), and face gender (malevs. female). Following this design, the targetfaces, which were either all attractive or all unattrac-tive, were tracked among distractor faces, whichwere also either all attractive or all unattractive.

ProcedureParticipants were tested individually. An adjustableheadrest was used to fix the participant’s viewingposition, which was set 60 cm away from the com-puter monitor. The faces presented in each trialwere of the same sex. They were randomly chosenfor each trial from the pool of 281 faces. Thechosen images were not repeated in the subsequenttrials until all faces in the pool had been used.The procedure for each trial of the experiments isillustrated in Figure 1A. Each trial was initiatedby a key press. It began with 10 stationary black rec-tangles on the screen. The location of the rectangleswas randomly assigned, with the constraint thatnone would occlude the others, and the centre-to-centre distance was not less than twice their size.Five of the rectangles would then start to blinktwice for 2 s, signalling the target location.Following this, the rectangles changed abruptlyinto 10 faces and started to move in randomdirections. The faces bounced off each otherwhen the centre-to-centre distance was less thantwice their size. They avoided the edge of thedisplay area when the centre to edge distance wasless than their size. Participants were asked totrack the five moving targets. The velocity of theface images varied between 3.9 and 6.3°/s with amean of 5.1°/s. All faces were moving about thescreen for 5 s. As soon as the motion stopped,all the faces were again occluded by black rec-tangles. The task was to pick out the five targetsby clicking on the rectangles. Once having beenclicked, the rectangle was highlighted with a

Table 1. Mean ratings of valence and arousal

Attractiveness Valence Arousal

Attractive 5.08 (0.65) 4.03 (0.48)

Average 4.68 (0.62) 3.94 (0.52)

Unattractive 4.33 (0.74) 3.99 (0.56)

Note: Values in parentheses represent standard deviations.

4 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0)

LIU AND HEN

140

145

150

155

160

165

170

175

180

yellow border, which could be switched on and offby clicking. Participants were forced to select fiveitems and were allowed to guess. Once five itemshad been selected, they clicked a “finish” buttonto start the next trial. The experiment lastedabout 30 minutes.

A total of 80 experimental trials were run after 4practice trials. Each of the eight conditions—2(target attractiveness)× 2 (distractor attractiveness)× 2 (face gender)—had 10 trials. All 80 trials weremixed in random order in one block.

Results and discussion

An alpha level of .05 was used for all statisticalanalyses in this report. Results of tracking accuracyare shown in Figure 2. The data were analysedusing repeated measures ANOVA. There was a sig-nificant main effect of target attractiveness, F(1,24)= 8.50, MSE= 18.52, p, .01, η2p= .26,where attractive faces were tracked better than unat-tractive faces. However, there was no differencebetween results for attractive and unattractive dis-tractors, F(1, 24)= 1.77, MSE= 27.74, p= .20,η2p= .07. There was also a main effect of facegender, F(1, 24)= 7.47, MSE= 23.31, p, .01,η2p= .24, where male faces were tracked better thanfemale faces. However, face gender did not interactwith other factors, ps. .57. None of the othertwo-way or three-way interactions was significant,ps. .15.

The experiment shows that tracking performancewas affected by appraisal of facial attractiveness.This effect was only present for the target faces,

Figure 1. Illustration of the procedure used in the study. A. Procedure used in Experiments 1 and 2. B. Procedure used in Experiment 3.

Figure 2. Tracking accuracy in Experiment 1 as a function of target

and distractor attractiveness. A. Results for male faces. B. Results for

female faces. Error bars represent standard errors.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 5

ATTRACTIVENESS AFFECTS MULTIPLE-FACE TRACKING

185

190

195

200

205

210

215

220

225

while the influence of attractive distractors was neg-ligible. Althoughmale faces were tracked better thanfemale faces, the effect was not modulated byattractiveness.

The effect of attractive faces on tracking perform-ance could be due to varying levels of involuntaryattention to task-irrelevant information. Some par-ticipants might be able to suppress or control atten-tion to the task-irrelevant information better thanothers. If this is the case, it would be possible thatthose who were poor at the tracking task wouldshow a stronger attractiveness effect, because theyare poor at suppressing the attractiveness infor-mation. To test this prediction, we analysed the cor-relation between the size of attractiveness effect andoverall tracking performance. Consistent with thisprediction, a negative correlation was foundbetween the two, r= –.59, p, .001. Figure 3Ashows a scatter plot of this correlation.

Because the attractive faces had greater valencerating than unattractive faces in our normingstudy, we evaluated how the difference of valencecontributed to the attractiveness effect in thisexperiment. This was done by an item-based analy-sis of covariance (ANCOVA), where facial attrac-tiveness was treated as a random factor, and thevalence a covariate. ANCOVA measures whetherattractiveness had an effect on tracking perform-ance after removal of the covariate valence.Because only the target attractiveness had aneffect on tracking performance, we only usedtarget attractiveness in our ANCOVA as an inde-pendent factor. Hence the target faces were used

as random sample and corresponding valence forthese faces as covariate. Results showed a signifi-cant main effect of target attractiveness, F(1,278)= 3.57, MSE= 0.07, p, .03, η2p= .02,where attractive targets were tracked better thanunattractive targets. However, the covariate wasnot significant, F(1, 278)= 0.59, MSE= 0.001,p= .44, η2p= .002, suggesting that the valencemade no contribution to the attractiveness effectfound in this experiment.

EXPERIMENT 2

In Experiment 1, the target faces were either attrac-tive or unattractive. When the targets had a similarlevel of attractiveness, attention might be more orless evenly distributed among the targets.However, if attractiveness varies considerablyamong targets, attention and consequently trackingperformance for a given target could be affected byits attractiveness relative to other targets. To testthis hypothesis, we varied level of target attractive-ness in this experiment such that the targets con-sisted of attractive, average, and unattractive faces.We then compared the performance for thesedifferent types of targets.

Another issue concerning the use of all attractiveor all unattractive targetswas that attractive facesmaybemore visually similar to one another than unattrac-tive faces. Yantis (1992) showed that similarityamong targets improves tracking. Because of this,the tracking advantage for attractive targets may

Figure 3. Scatter plots for the correlation between the size of the attractiveness effect and tracking performance.

6 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0)

LIU AND HEN

230

235

240

245

250

255

260

265

270

not reflect attractiveness directly, but rather throughfeatural similarity. The design in the present exper-iment was able to eliminate this problem becausean effect of attractiveness would no longer be dueto greater similarity among the attractive targets.

Method

ParticipantsA different group of 61 undergraduate students (48females and 13 males) from the University of Hullparticipated in this study. The age of the participantsranged from 17 to 34 years (Mdn= 20). All partici-pants had normal or corrected-to-normal vision.

StimuliThe attractive and unattractive face stimuli were thesame as those in Experiments 1 and 2. In addition,we also added average-looking faces. The meanratings for the average-looking faces was 2.99(SD= 0.11, N= 171). As in the previous exper-iments, each face category consisted of 80 faces ran-domly selected from the respective pool of stimuli foreach participant. Each trial contained 1 attractive, 1unattractive, and 3 average-looking target faces.The distractors in this experiment consisted ofaverage-looking faces. Because 8 average faces wereused in each trial, and there were 80 trials in total,this requires a total of 640 average faces (80× 8).Since there were 171 average faces available, eachface was repeated about 4 times (640 ÷ 171= 3.7).

DesignThis was again a within-subject design. The inde-pendent variables were target attractiveness (attrac-tive, average, and unattractive) and face gender.Because Experiment 1 showed no effect of distrac-tor attractiveness, this variable was not included inthis and the next experiments.

ProcedureThis was identical to that in Experiment 1.

Results and discussion

The tracking results are shown in Figure 4. Themain effect of target attractiveness was significant,

F(2, 120)= 4.96, MSE= 61.83, p, .01,η2p= .10. No significant effect was found for facegender, F(1, 60)= 0.01, MSE= 93.25, p= .98, orinteraction between face gender and target attrac-tiveness, F(2, 120)= 0.90, MSE= 70.68, p= .41.Post hoc comparisons of means with Bonferronicorrection revealed better tracking performance forattractive targets than for average or unattractivetargets, ts (60). 2.66, ps, .02. There was nodifference between results for average and unattrac-tive targets, t(60)= 0.11, p= 1.00. The resultssuggest that attractive faces were attended morefavourably whereas unattractive and average-looking faces were not treated differently duringtracking.

Consistent with Experiment 1, there was a sig-nificant negative correlation between participants’overall tracking performance and the size of theattractiveness effect, r= –.22, p, .04. A scatterplot of this correlation is shown in Figure 3B.

Results of ANCOVA for Experiment 2 showthat there was a significant main effect of targetattractiveness, F(1, 448)= 7.37, MSE= 0.07,p, .01, η2p= .03, where attractive faces weretracked better than average-looking and unattrac-tive faces (ps, .01), and there was no differencebetween average-looking faces and unattractivefaces (p= .51). The covariate, or valence of thefaces, was also significantly related to the trackingperformance, F(1, 448)= 9.90, MSE= 0.09,p, .01, η2p= .02. This indicates that the valenceof faces also contributed to the attractivenesseffect in this experiment.

Figure 4. Tracking accuracy in Experiment 2 as a function of target

attractiveness and face gender. Error bars represent standard errors.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 7

ATTRACTIVENESS AFFECTS MULTIPLE-FACE TRACKING

275

280

285

290

295

300

305

310

315

EXPERIMENT 3

The tracking task in Experiments 1 and 2 did notexplicitly require binding of a target identity withits location. If target attractiveness is appraised, itis possible that the location of an attractive targetis monitored more accurately than an unattractivetarget. To test this hypothesis, we employed avariant of the MOT paradigm, where one of thefaces was randomly chosen from the targets andprobed at the end of each trial. The task here wasto specify the location of the probe face.

Method

ParticipantsA different group of 50 undergraduate students (25females and 25 males) from the University of Hullparticipated in this study. The age of the partici-pants ranged from 18 to 35 years (Mdn= 20). Allparticipants had normal or corrected-to-normalvision.

StimuliThese were the same as those in Experiment 1.

DesignThis was again a within-subject design. The inde-pendent variables were attractiveness, face gender,and participant gender. Although not originallyplanned, potential participant gender differencewas included in our analysis because the numberof participants in each gender group was fullybalanced.

ProcedureThe procedure of this experiment was identical tothat in Experiment 1 except that at the end ofeach trial one of the faces was randomly chosenfrom the targets and probed (see Figure 1B). Theparticipant was required to specify the specificlocation of the probe face by clicking on one ofthe occluded faces with a mouse. The occluder ofthe chosen face was then removed to reveal theanswer after the mouse click. This feedback wasused to engage the participants. Because this task

was harder than the standard task, the number offaces in each trial was reduced to 8, with thenumber of target faces reduced to 4. The taskbegan with 4 practice trials followed by 40 exper-imental trials (10 trials× 4 conditions).

Results and discussion

Results are shown in Figure 5. Q2The locations ofattractive targets were identified better than thoseof unattractive targets, F(1, 48)= 4.70, MSE=241.50, p, .04, η2p= .11. Neither face gendernor participant gender produced a significantmain effect, Fs(1, 48)= 1.08 and 0.001, ps= .30and .98, respectively. However, this was qualifiedby a significant two-way interaction between par-ticipant gender and attractiveness, F(1, 48)=9.97, MSE= 241.50, p, .01, η2p= .17, as well asa marginally significant interaction between

Figure 5. Tracking accuracy in Experiment 3 as a function of target

attractiveness and face gender. A. Results for male faces. B. Results

for female faces. Error bars represent standard errors.

8 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0)

LIU AND HEN

320

325

330

335

340

345

350

355

360

participant gender and face gender, F(1, 48)=3.80, MSE= 204.27, p= .057, η2p= .07. Thethree-way interaction also approached the level ofsignificance, F(1, 48)= 2.78, MSE= 168.51,p= .10, η2p= .06. The two-way interactionbetween face gender and attractiveness was not sig-nificant, F(1, 48)= 1.63, p= .21, η2p= .03. Theinteraction effects are illustrated in Figure 4.

To investigate the interaction effects, we con-ducted separate ANOVAs for the two participantgenders. Results for the female participantsshowed a significant main effect of attractiveness,F(1, 24)= 14.73, MSE= 246.00, p, .001,η2p= .38, and face gender, F(1, 24)= 4.45,MSE= 478.39, p, .05, η2p= .16. There was alsoa significant interaction between attractiveness andface gender, F(1, 24)= 7.55, MSE= 96.54,p, .01, η2p= .24. Further analyses of the inter-action via pairwise comparisons of means revealedthat female participants identified the spatiallocations for attractive male targets more accuratelythan those for unattractive male targets, t(24)=4.73, p, .001. In contrast, the locations for theseattractive and unattractive male targets were ident-ified equally well by male participants, t(24)= –

0.51, p= .61. In contrast to the results for femaleparticipants, ANOVA did not find any maineffects of face gender and attractiveness, or theinteraction between the two for the male partici-pants, Fs(1, 24), 0.42, ps. .53.

Consistent with Experiments 1 and 2, the per-formance was again negatively correlated with thesize of attractiveness effect, r= –.29, p, .03.Figure 3C shows a scatter plot of this correlation.

To determine whether valence also contributedto tracking performance, we performed anANCOVA on the data following the same statisti-cal procedure as that in Experiment 1. Resultsshowed a significant main effect of target attractive-ness, F(1, 278)= 3.22, MSE= 0.78, p, .04,η2p= .02, where attractive faces were trackedbetter than unattractive faces. The valence, as a cov-ariate, did not significantly affect the tracking per-formance, F(1, 278)= 0.39, MSE= 0.09, p= .53,η2p, .001.

The data in this experiment suggest that anattractive target was more likely to create a stronger

identity–location binding in female participants forattractive male targets. The difference betweenidentity–location binding for attractive and unat-tractive female targets was not significant.Consistent with Experiment 1, valence did notcontribute to the binding advantage of attractivetargets.

GENERAL DISCUSSION

Results in these experiments suggest a facilitativeeffect of facial attractiveness in multiple-face track-ing. Experiment 1 showed a better tracking per-formance when attractive faces were assigned astargets. Experiment 2 showed that attractivetargets were more likely to be tracked successfullywhen they were mixed with unattractive oraverage-looking targets. Finally, Experiment 3found that attractive targets could induce a strongerbinding of face identity and location.

Given that attractive faces may be more visuallysimilar to one another than unattractive faces andthat Yantis (1992) showed that similarity amongtargets improves tracking, one may argue that thetracking advantage for attractive targets may notreflect attractiveness directly, but rather throughfeatural similarity. The replicated attractivenessadvantage of Experiment 2 has eliminated thisequivocal problem. Furthermore, attractive targetsand attractive distractors were likely to be moresimilar than unattractive targets and unattractivedistractors in Experiments 1 and 3. This maymake tracking more difficult (Makovski & Jiang,2009) and predict an opposite unattractivenessadvantage. From this point, the tracking advantagefor attractive targets is less likely to reflect featuralsimilarity of attractive faces.

Facial beauty would not produce an effect onthese tracking tasks if facial features and their hol-istic information were not processed during theprocess of tracking. Therefore, given the patternof our results, the representations of the targetitems must be content addressable. There isalready evidence that face or object identities areprocessed in multiple-target tracking (Horowitzet al., 2007; Oksama & Hyönä, 2008; Ren et al.,

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 9

ATTRACTIVENESS AFFECTS MULTIPLE-FACE TRACKING

365

370

375

380

385

390

395

400

405

2009). The present study shows further that facialattractiveness is also processed during tracking.As the results in Experiment 1 indicate, attractive-ness appeared to be assessed only for the targetitems, because distractors had negligible effects ontracking performance. This finding is consistentwith the observation that only target identity is pro-cessed in multiple-object or multiple-face tracking(Pylyshyn, 2006; Ren et al., 2009).

Contrary to the suggestion that multiple-objecttracking is purely driven by low-level vision, ofwhich higher level cognitive processes are unableto penetrate, tracking performance appeared to bemodulated by the information content of targetitems. Content addressability may depend onwhether the information is important to the obser-ver and whether the observer is predisposed toprocess the type of information. Like face identity,attractive faces may automatically engage the atten-tion system. The appraisal for attractiveness in ourexperiments was spontaneous because it was taskirrelevant. Although the effects of such automaticappraisal in this and other studies (e.g., Sui &Liu, 2009) can be quite small, they are consistentand replicable. A small effect in these studies maysuggest that task-irrelevant processing can belargely suppressed by the central control. Thepresent study revealed that a participant’s trackingperformance was negatively correlated with thesize of attractiveness effect. This suggests thatsome individuals may have weaker central controlthan others: Those who were not as good at track-ing showed a larger attractiveness effect becausethey might not have as much central control andwere more easily distracted by the irrelevantinformation.1

Perhaps due to the extent of the central control,not all evidence to date shows that appraisal ofattractiveness is spontaneous or mandatory.Schacht et al. (2008) found that attractivenessappraisal depended on whether facial attractivenesswas task relevant (e.g., beauty rating). This ledthem to conclude that attractiveness appraisalrequires voluntary attention to the attractivenessdimension. The findings of our study show that

facial attractiveness can be appraised even when itis not task relevant. Our analyses of the correlationbetween the effect of attractiveness and trackingperformance in the three experiments provide pre-liminary evidence that participants with highertracking performance tend to produce a smallerattractiveness effect. This may be due to these par-ticipants’ better central control to suppress task-irrelevant appraisal of facial attractiveness. Task-irrelevant processing of facial beauty has beenreported in brain research. Chatterjee, Thomas,Smith, and Aguirre (2009) found that the ventraloccipital region remained responsive to facialbeauty when the task of their participants was tojudge facial identity rather than to attend explicitlyto attractiveness. They proposed that this region,which includes the fusiform gyrus, the lateral occi-pital cortex, and medially adjacent regions, is acti-vated automatically by beauty and may serve as aneural trigger for pervasive effects of attractivenessin social interactions (see also Kampe et al.,2001). Undoubtedly, the exact level of controlthat the central executive can have over automaticprocessing of certain task-irrelevant stimuli willrequire systematic future research.

The present study also provides preliminary evi-dence that target tracking is spontaneously affectedby gender detection. Experiment 1 showed thatmale faces were tracked better than female faces.This could be due to a gender bias as participantsin this experiment were mainly composed offemales. When we tested equal numbers of maleand female participants in Experiment 3, resultsshowed that female participants tracked thelocation of an attractive male face more accuratelythan that of an unattractive male face. However,it is unclear why the same effect was not foundfor female faces. The gender effect found inExperiment 3 may be similar to that in Bayliss, diPellegrino, and Tipper (2005), who found thataverted eyes of a face or nonpredictive arrowsproduce a stronger reflexive shift of attention infemales than in males. Given that the participantsin our Experiments 1 and 2 were predominantlyfemales, the results in these experiments could be

1 We thank Trafton Drew for suggesting this idea.

10 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0)

LIU AND HEN

410

415

420

425

430

435

440

445

450

largely driven by certain aspects of gender differ-ence. The role of gender difference in attentivetracking is therefore an important area for futureinvestigation.

Consistent with prior research, attractive facestimuli in this study were also rated more positivelyon the valence dimension. Although valence did notcontribute to the tracking advantage of attractivefaces when the tracked faces had similar attractive-ness (Experiments 1 and 3), it did contribute tothe tracking performance when the target faces con-sisted of a full range of attractiveness (Experiment2). Thus, the lack of valence effect in Experiments1 and 3 may be due to the much smaller range ofthis variable associated with the attractive targets.Positive valence is a central trait of facial beauty. Itis therefore not surprising that it produced asimilar effect to attractiveness on distributed atten-tion. It should be noted, however, that the differ-ences in valence ratings of attractive andnonattractive faces were quite small. No significantdifference was found in arousal ratings of attractiveand nonattractive faces. These rating data maysuggest amoderate link between facial attractivenessand valence, whereas the link between facial attrac-tiveness and arousal is negligible.

Although it is difficult to separate attractivenessfrom positive valence due to the very nature ofbeauty, it is possible to study whether valence inother types of stimuli produces a similar effect onmultiple-object tracking. A similar conclusion canbe made about the relationship between attractive-ness and averageness. It would not be possible toseparate attractiveness from averageness becauseaverageness is a defining feature of attractiveness.However, it is possible to study whether averagenessin nonface stimuli also produces a similar advantagein a multiple-object tracking task. There is little evi-dence that averageness in other kinds of stimuli alsoattracts attention, although it is highly likely thatbeautiful things in general attract more attention.Unlike faces, the relationship between averagenessand other types of beautiful things remains to beseen. In face research, this relationship is establishedthrough image morphing techniques. However, it isoften difficult to use the same method that requireswell-defined corresponding features to build an

average for other categories of things such as beau-tiful sunsets, because obvious correspondence insuch images is often absent.

The attentional bias for attractive faces found inthis study suggests that multiple-object tracking ismodulated by underlying biological significance ofthe tracked targets. It may reflect biological interestsof the observer. The preference for attractive facesmay be deeply rooted in evolution (Langlois et al.,2000; Rhodes, 2006). Prior research has demon-strated an automatic reaction to facial beauty infocalized attention. The present study shows asimilar response to beauty in distributed attention.Because multiple-object tracking depends on atten-tional resources in the workingmemory (Oksama&Hyönä, 2004), our results may suggest higherworking memory capacity and greater attentionalresources for attractive faces.

Original manuscript received 4 October 2010

Accepted revision received 8 September 2011

First published online day month year

REFERENCES

Aharon, I., Etcoff, N., Ariely, D., Chabris, C. F.,O’Connor, E., & Breiter, H. C. (2001). Beautifulfaces have variable reward value: fMRI and behavioralevidence. Neuron, 32, 537–551.

Bayliss, A. P., di Pellegrino, G., & Tipper, S. P. (2005).Sex differences in eye gaze and symbolic cueing ofattention. Quarterly Journal of Experimental

Psychology, 58A, 631–650.Chatterjee, A., Thomas, A., Smith, S. E., & Aguirre,

G. K. (2009). The neural response to facial attrac-tiveness. Neuropsychology, 23, 135–143.

Dion, K. K., Berscheid, E., & Walster, E. (1972). Whatis beautiful is good. Journal of Personality & Social

Psychology, 24, 285–290.Eagly, A. H., Ashmore, R. D., Makhijani, M. G., &

Longo, L. C. (1991). What is beautiful is good,but . . . A meta-analytic review of research on thephysical attractiveness stereotype. Psychological

Bulletin, 110, 109–128.Geldart, S., Maurer, D., & Carney, K. (1999). Five-

month-olds’ attention and affective responses tostill-faced emotional expressions. Infant Behavior &Development, 20, 563–568.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 11

ATTRACTIVENESS AFFECTS MULTIPLE-FACE TRACKING

455

460

465

470

475

480

485

490

495

Grammer, K., & Thornhill, R. (1994). Human (Homosapiens) facial attractiveness and sexual selection:The role of symmetry and averageness. Journal of

Comparative Psychology, 108, 233–242.Guo, K., Liu, C. H., & Roebuck, H. (2011). I know you

are beautiful even without looking at you:Discrimination of facial beauty in peripheral vision.Perception, 40, 191–195.

Horowitz, T. S., Klieger, S. B., Fencsik, D. E., Yang,K. K., Alvarez, G. A., & Wolfe, J. M. (2007).Tracking unique objects. Perception & Psychophysics,69, 172–184.

Kampe, K. K. W., Frith, C. D., Dolan, & Frith, R. J., U.(2001). Reward value of attractiveness and gaze.Nature, 413, 589.Q3

Kranz, F., & Ishai, A. (2006). Face perception is modu-lated by sexual preference. Current Biology, 16, 63–68.

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2008).International affective picture system (IAPS): Affective

ratings of pictures and instruction manual (Tech. Rep.No. A-8). University of Florida, Gainesville, FL.

Langlois, J. H., Kalakanis, L. E., Rubenstein, A. J.,Larson, A. D., Hallam, M. J., & Smoot, M. T.(2000). Maxims or myths of beauty: A meta-analyticand theoretical review. Psychological Bulletin, 126,

390–423.Langlois, J. H., & Roggman, L. A. (1990). Attractive

faces are only average. Psychological Science, 1,

115–121.Langlois, J.H.,Roggman, L.A., Casey, R. J., Ritter, J.M.,

Rieser-Danner, L. A., & Jenkins, V. Y. (1987). Infantpreferences for attractive faces: Rudiments of a stereo-type? Developmental Psychology, 23, 363–369.

Makovski, T., & Jiang, Y. V. (2009). Feature binding inattentive tracking of distinct objects. Visual Cognition,17, 180–194.

Maner, J. K., Gailliot, M. T., & DeWall, C. N. (2007).Adaptive attentional attunement: Evidence formating-related perceptual bias. Evolution and

Human Behavior, 28, 28–36.Maner, J.K.,Gailliot,M.T.,Rouby,D.A.,&Miller, S.L.

(2007). Can’t take my eyes off you: Attentionaladhesion to mates and rivals. Journal of Personalityand Social Psychology, 93, 389–401.

Oksama, L., & Hyönä, J. (2004). Is multiple objecttracking carried out automatically by an early vision

mechanism independent of higher-order cognition?An individual difference approach. Visual Cognition,11, 631–671.

Oksama, L., & Hyönä, J. (2008). Dynamic binding ofidentity and location information: A serial model ofmultiple identity tracking. Cognitive Psychology, 56,237–283.

Palermo, R., & Rhodes, G. (2007). Are you always onmy mind? A review of how face perception and atten-tion interact. Neuropsychologia, 45, 75–92.

Pylyshyn, Z. W. (2004). Some puzzling findings in mul-tiple object tracking: I. Tracking without keepingtrack of object identities. Visual Cognition, 11,

801–822.Pylyshyn, Z. W. (2006). Some puzzling findings in mul-

tiple object tracking (MOT): II. Inhibition of movingnontargets. Visual Cognition, 14, 175–198.

Pylyshyn, Z. W., & Storm, R.W. (1988). Tracking mul-tiple independent targets: Evidence for a paralleltracking mechanism. Spatial Vision, 3, 1–19.

Ren, D., Chen, W., Liu, C. H., & Fu, X. (2009).Identity processing in multiple-face tracking. Journalof Vision, 9(5), 1–15, 18. Q4

Rhodes, G. (2006). The evolutionary psychology of facialbeauty. Annual Review of Psychology, 57, 119–226.

Rhodes, G., Sumich, A., & Byatt, G. (1999). Are averagefacial configurations attractive only because of theirsymmetry? Psychological Science, 10, 52–58.

Ro, Russell, & Lavie, 2001. Q1Samuels, C. A., Butterworth, G., Roberts, T., Graupner,

L., &Hole, G. (1994). Facial aesthetics: Babies preferattractiveness to symmetry. Perception, 23, 823–831.

Schacht, A., Werheid, K., & Sommer, W. (2008). Theappraisal of facial beauty is rapid but not mandatory.Cognitive, Affective, and Behavioral Neuroscience, 8,132–142.

Slater, A., von der Schulenberg, C., Brown, E.,Bademoch, M., Butterworth, G., Parsons, S., et al.(1998). Newborn infants prefer attractive faces.Infant Behavior & Development, 21, 345–354.

Sui, J., & Liu, C. H. (2009). Can beauty be ignored?Effects of facial attractiveness on covert attention.Psychonomic Bulletin & Review, 16, 276–281.

Yantis (1992). Multielement visual tracking: Attentionand perceptual organization. Cognitive Psychology,24, 295–340. Q5

12 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0)

LIU AND HEN

500

505

510

515

520

525

530

535

540

PQJE624186Queries

Chang Hong Liu and Wenfeng Chen

Dear AuthorPlease address all the numbered queries on this page which are clearly identified on the proof for yourconvenience.

Thank you for your cooperation

Q1 Ro, Russell, & Lavie, 2001. Not in refs. Please provide full reference.

Q2 There was no text reference to Figure 5. Ok as inserted here?

Q3 Kampe, K. K. W., Frith, C. D., Dolan, Frith, R. J., U.: Initials scrambled? Should this be Kampe,K. K. W., Frith, C. D., Dolan, U., & Frith, R. J.?

Q4 Ren, D., Chen, W., Liu, C. H., & Fu, X. (2009). Presume 18 is a page number?

Q5 Yantis, (1992). Please give initials.

How to make corrections to your proofs using Adobe Acrobat Taylor & Francis now offer you a choice of options to help you make corrections to your proofs. Your PDF proof file has been enabled so that you can edit the proof directly. This is the simplest and best way for you to ensure that your corrections will be incorporated. To do this, please follow these instructions: 1. Check which version of Adobe Acrobat you have on your computer. You can do this by clicking on the “Help” tab, and then “About”. If Adobe Reader is not installed, you can get the latest version free from http://get.adobe.com/reader/.

If you have Adobe Reader 8 (or a later version), go to “Tools”/ “Comments & Markup”/ “Show Comments & Markup”.

If you have Acrobat Professional 7, go to “Tools”/ “Commenting”/ “Show Commenting Toolbar”.

2. Click “Text Edits”. You can then select any text and delete it, replace it, or insert new text as you need to. It is also possible to highlight text and add a note or comment.

3. Make sure that you save the file when you close the document before uploading it to CATS. A full list of the comments and edits you have made can be viewed by clicking on the “Comments” tab in the bottom left-hand corner of the PDF. If you prefer, you can make your corrections using the CATS online correction form.