SOLUTION: Social categorization | My Assignment Tutor

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/7576338Social categorization and the perception of facial affect: Target racemoderates the response latency advantage for happy facesArticle in Emotion · October 2005DOI: 10.1037/1528-3542.5.3.267 · Source: PubMedCITATIONS102READS9401 author:Some of the authors of this publication are also working on these related projects:Social categorisation and emotion perception View projectKurt HugenbergIndiana University Bloomington100 PUBLICATIONS 4,150 CITATIONSSEE PROFILEAll content following this page was uploaded by Kurt Hugenberg on 30 October 2014.The user has requested enhancement of the downloaded file.Social Categorization and the Perception of Facial Affect: Target RaceModerates the Response Latency Advantage for Happy FacesKurt HugenbergMiami UniversityTwo experiments competitively test 3 potential mechanisms (negativity inhibiting responses, featurebased accounts, and evaluative context) for the response latency advantage for recognizing happyexpressions by investigating how the race of a target can moderate the strength of the effect. Bothexperiments indicate that target race modulates the happy face advantage, such that European Americanparticipants displayed the happy face advantage for White target faces, but displayed a response latencyadvantage for angry (Experiments 1 and 2) and sad (Experiment 2) Black target faces. This pattern offindings is consistent with an evaluative context mechanism and inconsistent with negativity inhibitionand feature-based accounts of the happy face advantage. Thus, the race of a target face provides anevaluative context in which facial expressions are categorized.Keywords: happy face advantage, facial expressions, prejudice, stereotypesFacial expressions are at the core of human social life. Theyencode a vast wealth of information about the affective states ofindividuals around us and when successfully decoded can substantially facilitate social interaction (Ekman, 2003). Given the ubiquity of facial expressions in human life, and the potential windowthey offer into the internal states of those around us, many theoristshave argued that the survival benefits of rapidly and accuratelyparsing facial affect are of such evolutionary significance thathumans have evolved particular modules for the processing offacial affect (e.g., Fridlund, 1994). Considering the importance thatunderstanding facial affect plays in our everyday lives, it is perhaps not surprising that there has been an increasing amount ofresearch involved in understanding how we attend to, categorize,and process facial affect (Adolphs, 2002; Haxby, Hoffman, &Gobbini, 2002).For example, an increasing body of research suggests that thehuman cognitive system is evolutionarily predisposed to attend tonegative, and particularly hostile, expressions. Indeed, the processing of facial displays that pose a direct threat to an individual, suchas anger, have been found to be processed with particular efficiency (Rolls, 1992). Hansen and Hansen (1988) found that angryfaces are found more quickly than other expressions in visualsearch tasks. In their face-in-the-crowd paradigm, in which participants search for a singleton expression among a matrix of otherexpressions (e.g., an angry face in a matrix of happy faces), angryexpressions appear to “pop out” of visual displays, powerfullycapturing visual attention, which Hansen and Hansen interpretedas being due to a preattentive scan for threat (see O¨ hman, Lundqvist, & Esteves, 2001, for a conceptual replication). Such apreattentive scan for threat is consistent with other research regarding vigilance for threat (e.g., LeDoux, 1996), and it wouldcertainly be a very functional mechanism, considering the clearevolutionary importance of recognizing threatening facial displaysquickly and efficiently. Other visual attention tasks show thatangry expressions not only attract, but also hold visual attention.For example, Fox, Russo, Bowles, and Dutton (2001) used anexogenous cuing paradigm (adapted from Posner, Inhoff,Friedrich, & Cohen, 1987) in which participants were tasked withdetecting a visual target whose location is accurately or inaccurately cued by different facial expressions. In particular, when anangry face and the visual cue appear in separate locations, responselatencies are particularly long. Fox et al. interpreted these longerlatencies as being due to a particular difficulty in shifting attentionaway from the location of angry faces, indicating that angry faceshold visual attention more powerfully than other expressions.Many theorists agree that because of the particular importance ofunderstanding and decoding facial information from others(O¨ hman, 1992), emotional, and particularly angry, faces seem topowerfully attract (Vuilleumier & Schwartz, 2001) and hold visualattention (e.g., Fox, Russo, & Dutton, 2002).Despite the sizable literature on the attention grabbing andholding power of negative, and particularly angry, facial expressions, a somewhat different pattern of data seems to be emergingregarding the categorization of facial expressions. An increasingnumber of studies seem to indicate that there is a speed advantagefor the recognition of happy facial expressions. When participantsare asked to simply categorize a series of single target expressions,happy expressions are categorized faster than angry (e.g., Billings,Harrison, & Alden, 1993; Hugdahl, Iversen, & Johnsen, 1993),disgusted (e.g., Ducci, 1981), sad (e.g., Kirita & Endo, 1995), andeven neutral expressions (e.g., Hugdahl et al., 1993). In a review ofthis literature, Leppa¨nen and Hietanen (2003) noted that the response time advantage for happy faces is a very robust phenomenon drawing an increasing amount of interest, in particular beI thank Amanda Diekman, Galen Bodenhausen, and Heather Claypoolfor their thoughtful comments on drafts of this article. I also thank JenMiller, Megan Erb, and Andy McGurk for their assistance with datacollection.Correspondence concerning this article should be addressed to KurtHugenberg, Department of Psychology, Benton Hall, Miami University,Oxford, OH 45056. E-mail: hugenbk@muohio.eduEmotion Copyright 2005 by the American Psychological Association2005, Vol. 5, No. 3, 267–276 1528-3542/05/$12.00 DOI: 10.1037/1528-3542.5.3.267267cause of its seeming disagreement with the findings of attentionaladvantage for negative and particularly angry faces.Although seemingly contradictory, Leppa¨nen, Tenhunen, andHietanen (2003) argued that the two different patterns of findingsfor angry and happy expressions differ critically in the cognitivemechanisms observers likely engage while performing attentionaltasks, such as search tasks and attention holding tasks, versuscategorization tasks in which facial expressions are presentedindividually. The response time advantage shown for angry expressions has been found almost exclusively in tasks that measurehow different expressions differentially grab and hold visual attention. Such visual search (e.g., O¨ hman et al., 2001) and attentiondisengagement (e.g., Fox et al., 2002) tasks measure the speed withwhich attention can focus on or shift away from a particularstimulus, respectively. Categorization tasks, which robustly showa response time advantage for happy faces, however, do notrequire the rapid allocation or reallocation of visual attention.Because of the nature of categorization tasks, attention allocationis fixed by presenting a target at a fixation point to which attentionhas already been drawn. Instead, such tasks measure only thecategorization time of the target expression. Thus, although negative expressions may grab and hold attention, happy expressionsseem to be processed more quickly.Despite the increasing amount of research investigating thisresponse latency advantage for recognizing happy faces, the reasons for the effect have remained somewhat unclear (see Leppa¨nenet al., 2003, for a review). Initial claims regarding happy versussad expressions being processed at coarse versus fine spatialfrequencies (which themselves are processed at different speeds)proved untenable (Kirita & Endo, 1995), given that the happy faceadvantage was robust for both real and schematic faces, the latterof which do not differ in spatial frequencies. Another more plausible explanation for this effect is that of negativity inhibition.Such an account argues that negative expressions engage a relatively extensive and thus slow cognitive analysis (see Baumeister,Bratslavsky, Finkenauer, & Vohs, 2001), or perhaps hold attentionlonger (e.g., Fox et al., 2002), inhibiting response times to negativeexpressions. Thus, the happy face advantage may be due not tohappy expressions being processed particularly quickly, but instead to negative expressions being responded to or processedmore slowly. Although certainly a plausible account, this potentialmechanism has a more difficult time explaining why the happinessadvantage occurs even when compared with neutral expressions,which should neither inhibit processing nor hold visual attention.Another account for this speed advantage for recognizing happyfaces has to do with the differential configuration of facial featuresin displays of positive and negative affect. Specifically, it may bepossible that happiness can be categorized based on the solefeature of a smiling mouth. Negatively valenced facial expressions, however, may require the perceiver to attend to the configuration among the facial features, including eyes, brow, and mouth(Adolphs, 2002). Thus, this feature-based account suggests that,while negative expressions may require thorough attention to theentire configuration of features of the face in question, happinessmay be processed by simply attending in a more piecemeal fashionto the shape of the mouth, potentially facilitating faster responses.Because of these potential feature-based differences in the ease ofcategorization of happy versus negative faces, Adolphs suggestedthat emotion categories may come to be mentally represented atdifferent levels. Specifically, he argued that basic emotion categories have a hierarchical organization of happy versus unhappy asthe basic level of categorization. Thus, whereas happiness mayitself be a basic level category, other negative expressions such asfear, anger, and disgust are actually subordinate categories to thebroader, basic level category of unhappy. This leads to a relatedlevel of categorization account, which argues that these initialdifferences in the ease of perception of positive and negative facialexpressions may lead to happiness and negative emotions beingrepresented at different levels of categorization (basic vs. subordinate, respectively), which may itself lead to or exacerbate thespeed advantage for recognizing happy faces.Recently, Leppa¨nen and Hietanen (2003) presented a third account for this response latency advantage for happy faces. Theyargued that the response latency advantage for recognizing happyfaces may be due to different expectancies about facial affectcreated by the typical experimental context. In particular, theydrew on the research of Cacioppo and colleagues (e.g., Cacioppo,Gardner, & Bernston, 1999), which suggested that in neutralenvironmental conditions, there are higher levels of activation ofthe positive affect system than of the negative affect system, whichwill actually lead to mildly positive affect. Cacioppo et al. referredto this higher level of baseline activation of the positive affectsystem as “positivity offset.” In the intentionally neutral conditionsthat exist in most emotion categorization experiments, most individuals will likely have a higher activation of the positive affectsystem, which may lead to overall differences in the expectanciesfor and facilitation in the processing of evaluatively positive stimuli. If participants are experiencing positivity offset at baseline,this may lead to the relatively quick recognition of happy faces anda longer processing time of negative faces because they contrastwith the normative affective state of the perceiver. Leppa¨nen andHietanen found the recognition time advantage for happy facesreversed, because of an inhibition of response latencies to happyexpressions, when participants were put in a negative evaluativecontext during categorization, induced by smelling an unpleasantodor. Thus, when the negative affective system is activated byodors, recognition speed for happy expressions is inhibited, leading to negative facial affect (i.e., disgust) being recognized morequickly than happiness.Although there are a number of interesting explanations for thecategorization advantage for happy faces, the importance of understanding which mechanism actually mediates the effect becomes increasingly important when researchers consider the trulysocial nature of emotion perception. For example, it is of criticalimportance for a police officer to quickly and accurately determinewhether a potentially dangerous suspect means well or ill. Curiously, given the inherently social nature of encoding and decodingfacial affect, relatively little research has investigated how criticalsocial variables, such as social categorization, may influence theprocess of emotion perception (but see Hugenberg & Bodenhausen, 2003, 2004).A consideration of the social categories of an individual displaying an expression and, in particular the race or ethnicity ofsuch an individual, may yield quite interesting results. Not onlywould such an investigation yield a more nuanced understandingof how facial expressions are perceived across racial lines, but mayalso serve as a fertile testing ground for efficacy of the differentproposed mechanisms by which the categorization advantage for268 HUGENBERGhappy faces may occur. In this vein, the current studies investigatethe extent to which the perceived race of a target moderates thestrength of the response latency advantage for happy expressions.One common prejudice among the European American majoritygroup in the United States is a negative evaluation of AfricanAmericans (Devine, 1989). By manipulating target race on awithin-subject basis, it is possible to not only test the extent towhich race moderates the response latency advantage for happyfaces, but in doing so also test the potential mechanisms by whichthis effect may occur.Anti-Black prejudice is a particularly viable prejudice to use inthis instance because it is consensually acknowledged, if notendorsed, by most individuals (Devine, 1989). Common manifestations of this pervasive, negative evaluation of African Americansinvolve an overall negativity in many evaluations, leading to anumber of harmful sequelae, such as lower ratings of job performance and lower willingness to hire African Americans (e.g.,Dovidio & Gaertner, 2000), increased ratings of guilt in many jurydecisions (e.g., Sommers & Ellsworth, 2001), and lower levels ofnonverbal positivity and increased seating distance (e.g., Word,Zanna, & Cooper, 1974). Although there have been relatively largechanges over the past few decades in explicit endorsement ofanti-Black prejudice, it seems that the effects of anti-Black prejudice have not abated. For example, Dovidio and Gaertner foundthat self-reported anti-Black prejudice has shown substantial reductions from 1989 to 1999. They also found, however, thatdespite reductions in reports of anti-Black prejudice, White respondents in 1999 were just as likely as White respondents in 1989to favor fellow European Americans in hiring evaluations anddecisions. Indeed, many theorists argue that contemporary formsof prejudice may be more subtle, but no less prevalent, than moreovert traditional anti-Black prejudice (e.g., Dovidio & Gaertner,1988).In the present experiments, unambiguously Black and unambiguously White targets displaying happiness and anger were presented to European American participants. A straightforward replication of the response latency advantage for happy faces amongWhite targets is predicted by all of the current proposed mechanisms for the response latency advantage for happy faces. Thenegativity inhibition account would predict that angry expressionswould slow responses as compared with happy expressions, leading to the happiness advantage. The feature-based (or the relatedlevel of categorization) account predicts that the happiness advantage will result from the easier piecemeal processing of happyfaces (or speedier categorization of the basic level category ofhappiness) as compared with angry faces. The evaluative contextaccount predicts that the neutral context of the experiment, pairedwith the evaluative positivity of White faces, should lead to agreater activation of the positive affect system, yielding a processing advantage for happy faces.The predictions for Black targets, however, are a bit morecomplex. In fact, the three different proposed mechanisms predictthree different patterns of data. If the response latency advantagefor happy faces occurs because negative stimuli tend to slowresponses, rather than happy expressions facilitating responses,then a member of a negatively evaluated category (e.g., a Blacktarget) displaying a negatively evaluated expression (e.g., anger)should slow processing and responses even more powerfully thana positively evaluated category (e.g., a White target) displaying thesame angry expression. Thus, a negativity inhibition hypothesiswould predict that the response latency advantage for happy facesshould be exacerbated for Black targets.Feature-based accounts, on the other hand, are based on thedifference between the ease of processing of happiness, which maybe recognized just by the mouth shape, and of negative expressions, which require attention to and processing of the entireconstellation of facial features. Such feature-based accounts predict that, insofar as White and Black faces displayed an expressionidentically, the response latency advantage should be identical forWhite and Black faces. Indeed, if differences in the ease or modeof processing of happiness and negative expressions are sufficientto explain the effect, the magnitude of the effect should be unaffected by the social categories of the perceptual targets, insofar asthose targets display identical expressions. The related level ofcategorization account predicts an identical pattern of data, albeitby a different mechanism. If happiness is simply easier to categorize because it is cognitively represented as a basic level emotion,and negative affective states are more difficult to categorize because they are represented as subordinate categories (to the basiclevel negative category), then the race of a target displaying theemotion should be irrelevant to the speed of categorization.Finally, if evaluative context drives the typical response latencyadvantage for recognizing happy faces, this response latency advantage for happy faces may be eliminated or even reversed forBlack targets. Indeed, insofar as Black targets are negativelyevaluated, this should create a relatively negative evaluative context in which the facial expressions are categorized. If Blacktargets create a negative evaluative context, this should facilitatethe processing of negative expressions and inhibit the processingof positive expressions, leading to an elimination, or in the case ofthe strong version of this hypothesis, a reversal of the responselatency advantage for happy faces among Black targets.Experiment 1Experiment 1 was designed to competitively test the threepotential mechanisms for the response latency advantage for happyfaces discussed earlier by manipulating the perceived race of andexpressions displayed by the perceptual targets. In Experiment 1,participants engaged in a speeded binary decision task in whichthey saw a series of colorized computer-generated Black andWhite race faces appear at the center of a computer screen. Eachtarget face was a man who displayed either happiness or anger.Participants were asked to categorize as quickly as possible each ofthe target faces as either happy or angry. As previously noted, areplication of the response latency advantage for happy faces waspredicted for White faces. Of particular interest was the pattern ofresponse latencies to Black faces, because of the different patternsof predictions drawn from the three mechanisms describedpreviously.MethodParticipants and design. Twenty-two European American undergraduates (13 women) participated in this experiment for partial course credit.Target race (2: Black vs. White) and target expression (2: happy vs. angry)were manipulated on a within-subject basis. Two participants were excluded from the analysis because, during debriefing, they admitted to notTARGET RACE MODERATES HAPPY FACE ADVANTAGE 269following task instructions and had error rates of nearly 50%. Thus, allanalyses were conducted with the remaining 20 participants.Materials. Four computer-generated facial structures were constructedusing the Poser 4 (2000) three-dimensional character animation software.These four core facial structures were then differentiated into eight different stimulus faces, one Black and one White face adapted from each corefacial structure; all eight targets were male faces. The Poser 4 softwarepermitted the Black face and White face adapted from a core facialstructure to have identical facial physiognomies, differing only in skintone, eye color, hair style, and hair color (see also Hugenberg & Bodenhausen, 2003, 2004). This matching ensured that differences in the facialphysiognomy of Black and White targets did not influence the way anexpression was displayed. Additionally, it rules out the possibility thatparticular facial features (other than those manipulated) are themselvesevaluatively laden (Livingston & Brewer, 2002) while ensuring a similarlevel of facial attractiveness.As past research found the happy face speed advantage using both actualfaces (e.g., Leppa¨nen & Hietanen, 2003), as well as schematic faces (e.g.,Kirita & Endo, 1995), it was not predicted that realistic computergenerated faces would show a distinctly different pattern of data than eitherreal or schematic faces. Moreover, by constructing realistic computergenerated faces, one can simultaneously pursue the ecological validityyielded by the use of naturalistic rather than schematic faces, as well as thetight control over the nature of the stimuli yielded by the use of schematicrather than actual faces. Pretesting was conducted to ensure that both Blackand White versions of each core facial structure were easily categorized byrace and were perceived to be unambiguously Black or White, respectively.Twelve European American participants were first asked to categorize eachof the eight (four Black and four White) target faces as either AfricanAmerican or European American. There was 100% consensus that theBlack targets were African American and the White targets were EuropeanAmerican. These same participants were then asked to rate each of thetargets on a 7-point response scale with responses ranging from 1 (definitely European American) to 7 (definitely African American). A pairedsamples t test confirmed that the White targets were rated as quite prototypic of the category European American (M 1.67, SE 0.2), whereasthe Black targets were rated as prototypic of the category African American (M 6.46, SE 0.2), t(11) 19.34, p .001.Each of these 8 stimulus faces was then further manipulated to createtwo versions: one with a clearly happy facial expression and a second witha clearly angry facial expression, yielding a total of 16 total target faces.The software allowed for manipulation of expression without changing thefacial physiognomy or race of the targets, providing two sets of matchedBlack and White targets displaying identical angry and happy expressions,differing only in skin tone, eye color, hair style, and hair color (see Figure1, left and center panels).Further pretesting was conducted to ensure that the angry and happyexpressions were clearly and easily identifiable as such on all eight stimulus faces. Forty-one European American participants were shown eitherthe eight Black (n 21) or the eight White (n 20) target faces and wereasked to answer the following question for each face in a free responseformat: “What emotion is this person displaying?” Participants respondedwith equal levels of either “happy” or happy-related words (e.g., “joy”) forboth Black (90%) and White (93%) happy faces, 2(1, N 41) 0.10, ns.There were also high levels of consensus for the angry faces, with Black(82%) and White (83%) faces eliciting equal rates of “anger” or angerrelated (e.g., “furious”) responses, 2(1, N 41) 0.00, ns.Procedure. Participants arrived in the laboratory in groups of up tofour. After giving informed consent, participants were seated at computersin individual cubicles and were instructed that they would engage in avisual attention task that involved speeded judgments of facial expressions.The procedure for Experiment 1 was adapted from Leppa¨nen and Hietanen(2003). The experiment consisted of two blocks of 80 trials each, separatedby a brief break. Within each block, each of the 16 stimulus faces (4 BlackFigure 1. An example of the Black and White happy (left panels) and angry (center panels) stimuli used inExperiment 1. Sad stimuli (right panels) were added in Experiment 2.270 HUGENBERGand 4 White faces, each displaying happiness and anger) was displayed fivetimes in a random order. Participants were tasked with categorizing each ofthe stimulus faces as happy or angry using the a key on the keyboard (lefthand) and the 5 key on the number pad (right hand). Response mappingswere counterbalanced on a within-subject basis; the mappings were reversed from the first to the second block. Starting order of responsemappings was counterbalanced on a between-subjects basis. Before beginning each block, participants engaged in 16 practice trials that familiarizedthem with all stimulus faces used in the task (one trial per stimulus face),as well as the particular response mappings for that block.Within the two blocks, each trial consisted of the presentation of thefixation point for 1,000 ms followed by the appearance of a stimulus facethat occluded the fixation point. Each stimulus face was displayed for 200ms. The stimulus face was then occluded by a gray box until participantsresponded, after which the next trial began with the presentation of thefixation point. Participants were asked to identify as quickly as possiblewhich of the two emotions (happy or angry) each of the target facesdisplayed while making as few errors as possible. Once participants hadcompleted both experimental blocks, they were thanked and debriefed.Results and DiscussionThe primary dependent measure of interest in this study was themean time taken by participants to correctly categorize the happyand angry Black and White faces. Incorrect responses were infrequent ( 9%) and were removed from analyses of responselatencies. Additionally, responses slower than two standard deviations above the mean of correct responses ( 3%) were eliminated. The remaining response latencies were aggregated intomean response times for happy Black, happy White, angry Black,and angry White faces separately for each participant. Initialanalyses yielded no significant effects for participant sex. As such,this variable is not discussed further.These response latencies were subjected to a 2 (target race:Black vs. White) 2 (target expression: happy vs. angry)repeated-measures analysis of variance (ANOVA). The ANOVAyielded a significant main effect of target race, F(1, 19) 4.55,p .05, which was qualified by the Target Race TargetExpression interaction predicted by the evaluative context hypothesis, F(1, 19) 30.00, p .001. As can be seen in Figure 2,among White faces, the predicted response latency advantage forhappy faces (M 547.1, SE 15.7), as compared with angryfaces (M 579.0, SE 17.8), was clearly in evidence, F(1, 19) 29.93, p .001, d 1.17. Among Black faces, however, thepattern of data showed a reversal, such that angry faces (M 548.7, SE 15.7) were recognized significantly faster than happyfaces (M 561.9, SE 15.6), F(1, 19) 5.13, p .04, d 0.49.Responses to angry expressions on Black faces were made substantially faster than responses to matched White faces, F(1, 19) 27.00, p .001, d 1.12, whereas responses to happy expressionson Black faces were made slower than responses to matched Whitefaces, F(1, 19) 6.44, p .02, d 0.54. The response latencydata seem to yield clear support for the evaluative context accountof the response latency advantage for recognizing happy faces. Thespeed of categorization of negative, angry expressions was substantially facilitated on Black as compared with White targets, andthe speed of categorization of happy faces was significantly sloweron Black as compared with White targets. This pattern of dataconflicts with both negativity inhibition and feature-based (or levelof categorization) accounts, which predict an exacerbation of or nochange in the happy face advantage, respectively.As some previous studies have found differential error ratesacross different target expressions during categorization (e.g., Leppa¨nen & Hietanen, 2003), of secondary interest was the pattern oferrors in categorization. To investigate the extent to which targetrace and target expression modulated patterns of errors, error rateswere aggregated across participants into percentage error scoresfor happy Black, happy White, angry Black, and angry White facesseparately for each participant. These data were also submitted toa 2 2 repeated-measures ANOVA identical to that used for theresponse latency data. This analysis yielded a marginally significant main effect of target expression, F(1, 19) 3.61, p .07,which was qualified by a marginally significant interaction oftarget race and target expression, F(1, 19) 3.77, p .07. Thismarginally significant interaction displayed a pattern similar tothat found in the response latency data, in that the tendency tomake more errors for negative than for positive expressions onWhite faces, F(1, 19) 34.42, p .001, d 1.26, was reducedon Black faces, F(1, 19) 9.71, p .01, d 0.67. Notably, themain effect for target race failed to approach significance, F(1,19) 0.02, p .85. This lack of a main effect of target raceindicates that the observed differences in response latencies werelikely not due to differences in the visual quality of the Black andWhite faces. Were any substantive differences in visual quality ofBlack and White faces in evidence, this should lead to overalldifferences in the patterns of errors for Black and White faces.Experiment 2The results of Experiment 1 indicate that the response latencyadvantage for happy faces is not robust across different targetraces. Although happy expressions were recognized more quicklythan angry expressions on White faces, replicating the commonlyfound response latency advantage for happy faces, EuropeanAmericans showed a reversal of the effect for Black faces, suchthat angry expressions were recognized more quickly than happyfaces. Of the three hypotheses tendered to explain the responselatency advantage for happy faces, the data from Experiment 1 fitbest with the evaluative context hypothesis. Both the accountbased on negativity inhibiting responses and the account based onthe differential ease of processing single features versus configurations of features in facial expressions seem inconsistent with theFigure 2. Experiment 1 categorization response latencies as a function oftarget race and target expression. Error bars represent standard errors.TARGET RACE MODERATES HAPPY FACE ADVANTAGE 271observed pattern of data. Moreover, given that Experiment 1 usedBlack and White computer-generated faces matched for facialphysiognomy and facial expression, this eliminates the possibilitythat the reversal of the observed pattern was due to any real-worlddifferences in the ways in which members of different racesexpress facial affect or to differences in the evaluation of particularrace-linked facial features that were not explicitly manipulated todifferentiate the races of the targets. Finally, as the overall errorrates for Black and White faces closely approximate one another( p .85), it is unlikely that the current effect could be easilyexplained by differences in the perceptual quality or ease ofparsing the Black and White faces.Although these results from Experiment 1 are in line withpredictions made by the evaluative context hypothesis, there is atleast one potential alternate explanation for the observed pattern ofdata. Past research indicates that, despite their endorsement ofegalitarian principles, many European Americans continue to experience anti-Black affect and evaluate African Americans negatively (e.g., Dovidio & Gaertner, 1988, 2000). In addition to beingsubject to these globally negative evaluations by many EuropeanAmericans, the specific content of the consensual cultural stereotype of African Americans includes, among other things, aggressiveness (Devine, 1989). As anger is the facial expression that ismost congruent with the African American stereotypic trait ofaggression, perhaps it is not surprising that anger seems to be seenmore readily on Black as compared with White faces by EuropeanAmerican participants (Hugenberg & Bodenhausen, 2003). Thus,rather than Black faces serving as a globally negative context inwhich to categorize expressions, it may be that facilitated perceptions of anger on Black faces may be driven by many EuropeanAmericans’ stereotypes of angry African Americans. Perhaps thecongruency between the cultural stereotype of African Americansand the specific emotion of anger on Black faces, rather thancongruence between a globally negative evaluation and negativeexpressions in general, leads to a facilitation of responses amongEuropean Americans for such targets. Thus, a stereotype congruence hypothesis would predict that facilitation for responsesshould occur only for stereotypic expressions (i.e., anger) and notfor other negatively valenced but nonstereotypic expressions (e.g.,sadness).Conversely, the findings of Experiment 1 may be more universalin that they apply generally to negative expressions across theboard. While the culturally pervasive stereotype (i.e., specificbeliefs) about African Americans includes hostility/anger, lingering anti-Black prejudice (i.e., global evaluations) among manyEuropean Americans is of negative valence. It may be that anoverall negative evaluation of Black (as compared with White)targets would lead to a general activation of negativity, causing areversal of the happiness advantage for all negatively valencedexpressions, not just for stereotype congruent expressions. Thus,the broader evaluative context hypothesis would predict that theeffects found in Experiment 1 should hold true for different negatively valenced expressions, regardless of whether they are stereotype congruent or nonstereotypic.Experiment 2 was intended to be a replication and extension ofExperiment 1, specifically designed to provide a competitive testof the stereotype congruence and the evaluative context hypotheses. Although both hypotheses predict identical patterns of responses for happy versus angry face decisions, replicating thefindings of Experiment 1, they differ in the predictions regardingother negatively valenced expressions. As such, the current experiment used a procedure quite similar to that of Experiment 1, butadded a between-subjects manipulation of whether the expressionswere happy versus angry or happy versus sad. The stereotypecongruence hypothesis predicts that the pattern of results seen inExperiment 1 should occur only (or at least most strongly) forhappy versus angry decisions, whereas the evaluative contexthypothesis predicts that categorization of negatively valenced expressions, regardless of the particular expression displayed, shouldbe facilitated on Black faces as compared with White faces.MethodParticipants and design. Forty European American undergraduates(20 women) participated in this experiment for partial course credit. Targetrace (2: Black vs. White) and expression valence (2: positive vs. negative)were manipulated on a within-subject basis, and target expressions (2:happy vs. angry vs. happy vs. sadness) were manipulated on a betweensubjects basis.Materials and procedure. The materials and procedure were identicalto Experiment 1 except as noted. The eight core faces used in Experiment1 (four Black and four White) were manipulated using the Poser 4 softwareto create three versions of each core face: one happy, one angry, and onesad (see Figure 1). The happy and angry faces were identical to those usedin Experiment 1. Additional pretesting was conducted on the sad faces toensure that they were equally perceived as such on Black and Whitetargets. Forty European American participants were shown the four Black(n 20) or the four White (n 20) sad target faces and were asked toanswer the following question for each face in a free response format:“What emotion is this person displaying?”. Participants responded withequal levels of either “sad” or sad-related words (e.g., “grieving”) for bothBlack (80%) and White (86%) target faces, 2(1, N 40) 1.11, ns.Participants arrived at the laboratory in groups of up to four and wererandomly assigned to either the happy versus angry condition (identical toExperiment 1; n 20) or the happy versus sad condition (n 20). Thehappy versus sad condition was identical to the happy versus angrycondition except that, as noted, participants were asked to make speededdecisions about happy and sad faces rather than happy and angry faces.Results and DiscussionSimilar to Experiment 1, the primary dependent measure ofinterest in this study was the mean time taken by participants tocorrectly categorize the happy versus angry and happy versus sadBlack and White faces. Incorrect responses were infrequent (8%)and were removed from analyses of response latencies. Additionally, responses slower than two standard deviations above themean of correct responses ( 3%) were eliminated. The remainingresponse latencies were aggregated into mean response times forpositive Black, positive White, negative Black, and negative Whitefaces separately for each participant. Although all participants sawthe same positive expressions, the negative expressions seen byparticipants differed on a between-subjects basis, which accountsfor the different terminology used in Experiment 1 and Experiment2. As initial analyses yielded no significant effects for participantsex, this variable is not discussed further.These response latencies were then subjected to a 2 (target race:Black vs. White) 2 (expression valence: positive vs. negative) 2 (target expressions: happy versus angry vs. happy versus sad)mixed-model ANOVA with repeated measures on the first two272 HUGENBERGfactors. Similar to the results of Experiment 1, the ANOVAyielded a main effect for target race such that responses to Blacktargets were made more quickly than were responses to Whitetargets, F(1, 39) 8.14, p .01. This main effect of target racewas qualified by a Target Race Expression Valence interaction,F(1, 38) 45.07, p .001, an interaction predicted by theevaluative context account. Notably, this two-way interaction wasnot qualified by a three-way interaction of target race, expressionvalence, and target expressions, as was predicted by the stereotypecongruence account, F(1, 38) 0.66, p .4. The lack of aninteraction of target race, expression valence, and target expressions suggests against the stereotype congruence hypothesis,which predicts for Black targets an elimination or reversal of theresponse latency advantage only for happy as compared with angryfaces, but not for happy as compared with other negative expressions (e.g., sadness). No other effects approached statistical significance (all ps .3).To examine the nature of the Target Race Expression Valenceinteraction, a series of planned contrasts was conducted. Thesecontrasts were designed to further investigate the influence oftarget race on the speed of recognition for positive versus negativeexpressions. As predicted by the evaluative context hypothesis,positive expressions (i.e., happiness: M 546.9, SE 11.8) wererecognized more quickly than were negative expressions (i.e.,anger or sadness: M 569.4, SE 12.5) on White faces, F(1,38) 34.56, p .001, d 0.91, replicating the findings ofExperiment 1. This pattern showed a significant reversal on Blackfaces, however, such that negative expressions (M 542.0, SE 12.8) were recognized more quickly than positive expressions(M 555.7, SE 11.2), F(1, 38) 12.81, p .001, d 0.55.Thus, averaging across specific expression, negative expressionswere recognized more quickly than happiness on Black faces.Further replicating the pattern of data from Experiment 1, responses to negative expressions on Black faces were made substantially faster than responses to matched White faces, F(1, 38) 51.23, p .001, d 1.11, whereas responses to positive expressions on Black faces were made significantly slower than responses to matched White faces, F(1, 38) 5.29, p .03, d 0.36.Although the Target Race Expression Valence interactionwas not qualified by the three-way interaction of target race,expression valence, and target expressions, it may be premature toeliminate the stereotype congruence hypothesis based on this nulleffect without examining the nature of the happy versus angry andthe happy versus sad conditions separately. To draw the strongestconclusion in favor of the evaluative context hypothesis, both thehappy versus angry and the happy versus sad conditions shouldseparately show the happy over negative advantage for Whitefaces and an elimination or a reversal of this pattern for Blackfaces. As such, two separate 2 (target race: Black vs. White) 2(expression valence: positive vs. negative) ANOVAs were conducted, one at each level of target expression (see Figure 3). The2 2 ANOVA for the happy versus angry condition yielded theinteraction of target race and expression valence predicted by theevaluative context hypothesis, F(1, 19) 17.46, p .001. OnWhite faces, happy expressions (M 545.7, SE 18.3) enjoyeda significant recognition speed advantage as compared with angryexpressions (M 561.6, SE 18.7), F(1, 19) 8.66, p .01,d 0.63. This pattern showed a significant reversal on Blackfaces, such that responses to angry expressions (M 536.1, SE 20.3) were faster than to happy expressions (M 552.2, SE 18.1), F(1, 19) 8.88, p .01, d 0.64. The separate 2 2ANOVA for the happy versus sad condition indicated a verysimilar pattern of results, yielding an interaction of target race andexpression valence, F(1, 19) 28.23, p .001. Again, on Whitefaces, happy expressions (M 548.0, SE 15.1) enjoyed areliable recognition speed advantage over sad expressions (M 577.3, SE 16.7), F(1, 19) 29.19, p .001, d 1.16. Thispattern showed a significant reversal for Black faces, with sadBlack faces (M 547.8, SE 15.8) being categorized morequickly than happy Black faces (M 559.2, SE 13.5), F(1,19) 4.42, p .05, d 0.45. As predicted by the evaluativecontext hypothesis, when analyzed separately, both the happyversus angry and happy versus sad conditions yielded a TargetRace Expression Valence interaction. In both cases, happyexpressions were recognized significantly more quickly than negative expressions on White faces. The patterns showed significantreversals for Black faces, with negative expressions being recognized more quickly than happiness. The stereotype congruencehypothesis predicts that no interaction of target race and expression valence would occur in the happy versus sad condition. Giventhe clear evidence of such an interaction, and that both the happyversus angry and the happy versus sad conditions yielded reversalsof the happy face advantage for Black targets, it seems reasonableto conclude that the evaluative context hypothesis has receivedmore support than the stereotype congruence hypothesis. Takentogether, these data indicate that the relatively negative evaluativeFigure 3. Experiment 2 categorization response latencies as a function oftarget race and target expression for both happy versus angry (top panel)and happy versus sad (bottom panel) conditions. Error bars representstandard errors.TARGET RACE MODERATES HAPPY FACE ADVANTAGE 273context created by Black targets facilitates the processing of evaluatively congruent negative expressions, while inhibiting the speedof processing of evaluatively incongruent positive expressions, ascompared with the speed of processing matched expressions displayed by an evaluatively positive White target.Error rates were also aggregated across participants and submitted to 2 2 2 mixed-model ANOVA identical to that used forthe response latency data. As with Experiment 1, this analysisyielded a marginally significant main effect of expression valence,F(1, 38) 3.72, p .06. This main effect was qualified by aninteraction of target race and expression valence, F(1, 38) 10.11,p .01, displaying a pattern similar to that found in the responselatency data, which indicated that the tendency to make moreerrors for negative than for positive expressions on White faces,F(1, 38) 2.76, p .10, d 0.26, was eliminated among Blackfaces, F(1, 38) 0.00, p 1, d 0.00. Notably, this two-wayinteraction of target race and expression valence was not qualifiedby the three-way interaction of target race, expression valence, andtarget expressions, F(1, 38) 0.33, p .55. No other effectsachieved statistical significance.General DiscussionAcross two experiments, the results indicate that the responselatency advantage for happy faces is moderated by the race of thetarget displaying the expressions. Experiment 1 replicated theresponse latency advantage for happy faces (as compared withangry faces) among White targets, but found that the recognitionadvantage in response latencies reversed for matched Black targets, clearly supporting an evaluative context hypothesis. Experiment 2 replicated and extended the results of Experiment 1, providing a context in which the stereotype congruence and evaluativecontext hypotheses were empirically opposed, with the evidenceagain strongly supporting the latter. Not only does the currentresearch competitively test multiple mechanisms for the responselatency advantage for recognizing happy expressions, but it extends the small but growing body of previous research regardingthe evaluative context of affect categorization. Instead of manipulating evaluative context by a manipulation of a subjective experience that is irrelevant to the act of categorization (e.g., bymanipulating odors), the current research manipulates evaluativecontext within the stimuli themselves in a highly ecologically validmanner.Although the current research does suggest that differences inevaluative contexts drive the response latency advantage for happyexpressions, the particular mechanisms underlying this effect stillrequire further research. As Leppa¨nen and Hietanen (2003) suggested, baseline levels of activation of the positive affective systemmay facilitate the processing of valence-congruent positive stimuliwhile inhibiting valence-incongruent negative stimuli. Indeed, anegative judgmental context (i.e., unpleasant odors) eliminates thehappy face advantage. Although positivity offset is certainly oneplausible explanation for the robust response latency advantage forhappy faces, another possible reason may be that participants arecommonly judging faces that may be perceived as members oftheir own racial or ethnic in group. There is an extensive literatureon in-group bias, or the tendency both to evaluate one’s in groupsmore positively than out groups and to discriminate in favor ofone’s in group over out groups (see Brewer & Brown, 1998, for areview), which could lead to facilitated evaluations of positivityamong one’s racial in group. In fact, these two explanations are notmutually exclusive, and may account for why the reversal of theadvantage for happy faces among Black targets is not as strong asthe happy face advantage for White targets. Again, although notmutually exclusive, neither of these explanations is sufficient toprovide a specific mechanistic account of the cognitive processesinvolved in how an evaluative context might influence categorization speed.One plausible mechanistic explanation for the current pattern ofdata is evaluative priming. For example, in one of the initialdemonstrations of evaluative priming, Fazio and colleagues (e.g.,Fazio, Sanbonmatsu, Powell, & Kardes, 1986) found that thepresentation of an attitude object facilitated the processing of asubsequent attitude object of the same valence. Thus, a positivelyvalenced stimulus facilitates the evaluation of a subsequent positively evaluated stimulus, an effect that seems quite robust acrossnumerous different manipulations of task parameters. As race isamong the most easily processed of social categories (Hugenberget al., 2005; Stangor, Lynch, Duan, & Glas, 1992), participantsmay make race categorizations before affect categorizations,which would permit the evaluative tone of the category to activatepositivity or negativity, facilitating the subsequent processing ofevaluatively congruent facial affect.Given the speed and accuracy with which facial affect is typically processed (see Adolphs, 2002), however, it is more likely thatsocial categorization and affect categorization are occurring inparallel via separate, but interconnected, systems (Haxby et al.,2002). Thus, the typical understanding of sequential evaluativepriming may be insufficient to account for the current data. Instead, the current results may be due to evaluative fluency anddysfluency experienced during the categorization process. If participants feel negativity because of race (i.e., from a Black face)and positivity because of expression (i.e., from a happy expression), such evaluative dysfluency may lead to response inhibition,as indicated by increased response latencies. Alternatively, whenboth race and expression are evaluatively congruent, this mayfacilitate responding, leading to a quicker and more accuratecategorization decision.Alternatively, it may be that the current data may also beexplained by small and fleeting shifts in participants’ affectiveresponses to the different stimuli. At this point, it is unclearwhether participants’ affective state was manipulated by the stimuli themselves; however, it may be possible that participants experienced more negative affect while perceiving Black stimuli andmore positive affect when perceiving White stimuli. According tothe work of Niedenthal and colleagues (e.g., Niedenthal, Halberstadt, Margolin, & Innes-Ker, 2000), individuals who are experiencing a particular emotion tend to perceive more readily thatsame emotion in the faces of others. For example, in a morphmovies task in which faces shift in expression from happiness tosadness, happy participants see happiness lingering longer on facesthan do participants in other affective states. If participants arehaving specific affective reactions to the races of the target faces,the pattern of data seems quite sensible for White faces. Happiness(momentarily induced by White faces) would lead to facilitatedperceptions of happiness. However, the most common affectiveresponse of European Americans to African Americans is anxietyand fear (e.g., Rickett & Cacioppo, 2003). If anxiety or fear (or274 HUGENBERGboth) is induced in participants, this sensibly facilitates the perception of anger; however, it is unclear how it might facilitate theperception of sadness, an evaluatively congruent but clearly different emotion. Thus, although participants may experience different temporary affective states to Black and White targets, theexperience of this emotion seems insufficient to account for all ofthe current data.Given the relatively substantial literature on facial expressions,there is a surprising dearth of research investigating how the socialcategories of perceptual targets may moderate how facial expressions are perceived, attended to, and decoded. Affect researchershave recently begun to take seriously the issue of the race of atarget, and of social category in general, and are finding thattargets of different races can commonly elicit very different patterns of responses, even when displaying similar patterns of facialaffect.More specifically, the current findings that target race modulatesthe speed at which facial expressions are categorized may be ofbroader interest when one considers the particular neural structuresby which face and emotion perception are typically mediated.Haxby and colleagues (2002) claimed that the core face perceptionsystem feeds into a more extended system, including in the case ofemotion perception, the amygdala, insula, and the limbic system atlarge. These connections are particularly important, as a wealth ofrecent research has shown the amygdala to be a center for socialcognition, implicated in attentional modulation (LeDoux, 1996), inperception of facial expression (e.g., O¨ hman, 2002), as well as inthe perception of out-group members (e.g., Hart et al., 2000), andin implicit prejudice (Phelps et al., 2000). For example, Hart andcolleagues (2000) found that the amygdala tends to activate whenprocessing novel faces, both of ethnic in groups and out groups;however, this activation decreases relatively quickly to in-groupfaces. Conversely, the amygdala does not easily habituate to repeated presentations of the faces of ethnic out-group members.Although Hart and colleagues used only nonemotional faces asstimuli, the implications are that faces of out-group members mayappear unfamiliar, threatening, or both for much longer than thoseof in-group members. Related work by Phelps et al. found that notonly did White participants show greater amygdala activation tounfamiliar Black faces, but that White participants’ Implicit Association Test (Greenwald, McGhee, & Schwartz, 1998) scorescorrelated reliably with the magnitude of amygdala activation toBlack-versus-White faces. The neural systems that underlie bothracial attitudes and face perception are certainly more extensivethan the amygdala (Phelps et al., 2000); however, it is interestingthat the amygdala seems to play an important role therein, especially considering that it also seems central to decoding facialexpressions.In fact, although the current research did not measure individualdifferences in prejudice, past work has found a clear link betweenimplicit measures of prejudice, social categorization, and perceptions of facial expressions. For example, Hugenberg and Bodenhausen (2003) found that as implicit prejudice increased, so did thetendency to perceive anger both as lingering on Black as comparedwith White faces (Study 1), and as appearing more quickly onBlack as compared with White faces (Study 2). In a related set offindings, Hugenberg and Bodenhausen (2004) found that as perceivers’ implicit prejudice increases, so does the tendency tocategorize angry racially ambiguous targets, as compared withmatched happy targets, as Black. The current work may shed newlight on this previous work. All of this previous work has focusedon the distinction between happy and angry facial expressions,under the assumption that perceivers would assimilate visual targets to specific stereotypes rather than to broadly negative evaluations. Thus, although anger seemed to linger on Black as compared with White targets, I would not have hypothesized that thiseffect would also occur for sadness. In the current work, however,Study 2 clearly showed that valence congruency rather than stereotype congruency is the mechanism behind the current effects.Although the current research used substantially different methodology and measures than my previous work on race, prejudice, andcategorization, Study 2 suggests that perhaps valence may play amore potent role in explaining the results of the past work thanpreviously thought.The current research hopefully serves to emphasize the importance of understanding how social categories such as race andethnicity moderate important psychological effects in theoreticallymeaningful ways. Perhaps more interestingly, the current researchsuggests that the manipulation of important social variables, suchas social categories, can also yield insights into the mechanisms bywhich such effects occur. Although more research is certainlyneeded to understand how social categorization and the perceptionof facial affect intersect, the current investigation suggests thathow individuals perceive and categorize facial expressions candepend quite critically on who it is that is displaying theexpression.ReferencesAdolphs, R. (2002). Recognizing emotion from facial expressions: Psychological and neurological mechanisms. Behavioral and CognitiveNeuroscience Reviews, 1, 21–62.Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001).Bad is stronger than good. Review of General Psychology, 5, 323–370.Billings, L. S., Harrison, D. W., & Alden, J. D. (1993). Age differencesamong women in the functional asymmetry for bias in facial affectperception. Bulletin of the Psychonomic Society, 31, 317–320.Brewer, M. B., & Brown, R. J. (1988). Intergroup relations. In D. T.Gilbert, S. T. Fiske, & G. Lindzey (Eds.), Handbook of social psychology (pp. 554–594). Boston: McGraw-Hill.Cacioppo, J. T., Gardner, W. L., & Bernston, G. G. (1999). The affectsystem has parallel and integrative processing components: Form follows function. Journal of Personality and Social Psychology, 76, 839–855.Devine, P. G. (1989). Stereotypes and prejudice: Their automatic andcontrolled components. Journal of Personality and Social Psychology,56, 5–18.Dovidio, J. F., & Gaertner, S. L. (1988). On the nature of contemporaryprejudice: The causes, consequences, and challenges of aversive racism.In J. Eberhardt & S. T. Fiske (Eds.), Prejudice, discrimination, andracism (pp. 61–89). Orlando, FL: Academic Press.Dovidio, J. F., & Gaertner, S. L. (2000). Aversive racism and selectiondecisions: 1989 and 1999. Psychological Science, 11, 315–319.Ducci, L. (1981). Reaction times in the recognition of facial expressions ofemotion. Italian Journal of Psychology, 8, 183–193.Ekman, P. (2003). Emotions revealed. New York: Henry Holt.Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986).On the automatic activation of attitudes. Journal of Personality andSocial Psychology, 50, 229–238.Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001). Do threateningTARGET RACE MODERATES HAPPY FACE ADVANTAGE 275stimuli draw or hold visual attention in subclinical anxiety? Journal ofExperimental Psychology: General, 130, 681–700.Fox, E., Russo, R., & Dutton, K. (2002). Attentional bias for threat:Evidence for delayed disengagement from emotional faces. Cognitionand Emotion, 16, 355–379.Fridlund, A. (1994). Human facial expression: An evolutionary view. SanDiego, CA: Academic Press.Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuringindividual differences in implicit cognition: The implicit association test.Journal of Personality and Social Psychology, 74, 1464–1480.Hansen, C. H., & Hansen, R. D. (1988). Finding the face in the crowd: Ananger superiority effect. Journal of Personality and Social Psychology,54, 917–924.Hart, A. J., Whalen, P. J., Shin, L. M., McInerney, S. C., Fischer, H., &Rauch, S. L. (2000). Differential response in the human amygdala toracial outgroup vs. ingroup face stimuli. Neuroreport, 11, 2351–2355.Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. (2002). Human neuralsystems for face recognition and social communication. Biological Psychiatry, 51, 59–67.Hugdahl, K., Iversen, P., & Johnsen, B. (1993). Laterality for facialexpressions: Does the sex of the subjects interact with the sex of thestimulus face? Cortex, 29, 325–331.Hugenberg, K., & Bodenhausen, G. V. (2003). Facing prejudice: Implicitprejudice and the perception of facial threat. Psychological Science, 14,640–643.Hugenberg, K., & Bodenhausen, G. V. (2004). Ambiguity in social categorization: The role of prejudice and facial affect in face categorization.Psychological Science, 15, 342–345.Hugenberg, K., Bodenhausen, G. V., Perrott, D., & Becker, A. (2005).Affective states influence category selection in person perception. Manuscript in preparation.Kirita, T., & Endo, M. (1995). Happy face advantage in recognizing facialexpressions. Acta Psychologica, 89, 149–163.LeDoux, J. E. (1996). The emotional brain: The mysterious underpinningsof emotional life. New York: Simon & Schuster.Leppa¨nen, J. M., & Hietanen, J. K. (2003). Affect and face perception:Odors modulate the recognition advantage of happy faces. Emotion, 3,315–326.Leppa¨nen, J. M., Tenhunen, M., & Hietanen, J. K. (2003). Faster choicereaction times to positive than to negative facial expressions: The role ofcognitive and motor processes. Journal of Psychophysiology, 17, 113–123.Livingston, R. W., & Brewer, M. B. (2002). What are we really priming?:Cue-based versus category-based processing of facial stimuli. Journal ofPersonality and Social Psychology, 82, 5–18.Niedenthal, P. M., Halberstadt, J. B., Margolin, J., & Innes-Ker, A. H.(2000). Emotional state and the detection of change in facial expressionof emotion. European Journal of Social Psychology, 30, 211–222.O¨ hman, A. (1992). Orienting and attention: Preferred preattentive processing of potentially phobic stimuli. In B. A. Campbell, H. Hayne, & R.Richardson (Eds.), Attention and information processing in infants andadults: Perspectives from human and animal research (pp. 263–295).Hillsdale, NJ: Erlbaum.O¨ hman, A. (2002). Automaticity and the amygdala: Nonconscious responses to emotional faces. Current Directions in Psychological Science, 11, 62–66.O¨ hman, A., Lundqvist, D., & Esteves, F. (2001). The face in the crowdrevisited: A threat advantage with schematic stimuli. Journal of Personality and Social Psychology, 80, 381–396.Phelps, E. A., O’Connor, K. J., Cunningham, W. A., Funayma, E. S.,Gatenby, J. C., Gore, J. C., & Banaji, M. R. (2000). Performance onindirect measures of race evaluation predicts amygdala activity. Journalof Cognitive Neuroscience, 12, 1–10.Poser 4 (Version 4.0.3) [computer software]. (2000). Santa Cruz, CA:Curious Labs.Posner, M. I., Inhoff, A. W., Friedrich, F. J., & Cohen, A. (1987). Isolatingattentional systems: A cognitive-anatomical analysis. Psychobiology,15, 107–121.Rickett, E., & Cacioppo, J. T. (2003, February). Differential emotions inprejudice. Poster presented at the Conference of the Society of Personality and Social Psychology, Los Angeles, CA.Rolls, E. T. (1992). Neurophysiological mechanisms underlying face processing within and beyond the temporal cortical visual areas. Philosophical Transactions of the Royal Society (London) B, 335, 11–21.Sommers, S. R., & Ellsworth, P. C. (2001). White juror bias: An investigation of prejudice against Black defendants in the American courtroom.Psychology, Public Policy, and Law, 7, 201–229.Stangor, C., Lynch, L., Duan, C., & Glas, B. (1992). Categorization ofindividuals on the basis of multiple social features. Journal of Personality and Social Psychology, 62, 207–218.Vuilleumier, P., & Schwartz, S. (2001). Emotional facial expressionscapture attention. Neurology, 56, 153–158.Word, C. O., Zanna, M. P., & Cooper, J. (1974). The nonverbal mediationof self-fulfilling prophecies in interracial interaction. Journal of Experimental Social Psychology, 10, 109–120.Received May 12, 2004Revision received November 3, 2004Accepted December 13, 2004 276 HUGENBERGView publication stats