Implicit Ingroup Metafavoritism: Subtle Preference for Ingroup Members Displaying Ingroup Bias

http://psp.sagepub.com

Personality and Social Psychology Bulletin

DOI: 10.1177/0146167208315210 2008; 34; 807 originally published online Mar 22, 2008; Pers Soc Psychol Bull

Luigi Castelli, Silvia Tomelleri and Cristina Zogmaister Implicit Ingroup Metafavoritism: Subtle Preference for Ingroup Members Displaying Ingroup Bias

http://psp.sagepub.com/cgi/content/abstract/34/6/807 The online version of this article can be found at:

Published by:

http://www.sagepublications.com

On behalf of:

Society for Personality and Social Psychology, Inc.

can be found at:Personality and Social Psychology Bulletin Additional services and information for

http://psp.sagepub.com/cgi/alerts Email Alerts:

http://psp.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://psp.sagepub.com/cgi/content/refs/34/6/807 Citations

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

807

Implicit Ingroup Metafavoritism:Subtle Preference for Ingroup MembersDisplaying Ingroup Bias

Luigi CastelliSilvia TomelleriCristina ZogmaisterUniversity of Padova

seem largely reduced (Crosby, Bromley, & Saxe, 1980;Dovidio & Gaertner, 2000). For instance, increasinglymore people report willingness to accept a Black individualas a superior at work or as a close relative. This encourag-ing picture, however, is often in sharp contrast to the still-present differential treatment accorded to members ofmajority and minority groups (e.g., Bodenhausen et al.,1998; Dovidio, Kawakami, & Gaertner, 2002). In par-ticular, spontaneous negative affective responses areautomatically triggered in response to members of stig-matized groups, such as Blacks, even by individuals whodefine themselves as egalitarian (Fazio, Jackson, Dunton,& Williams, 1995; Greenwald, McGhee, & Schwartz,1998). In short, there is clear evidence that intergroupattitudes are often organized into two different levelsand that, within the same individual, controlled egalitar-ian attitudes may coexist with negative automaticresponses (Wilson, Lindsey, & Schooler, 2000).

The goal of the current article is to verify whethersimilar dual attitudes may be quickly developed towardmembers of one’s own social group who perform dif-ferent types of intergroup behaviors. We predict thatingroup members who systematically favor the ingroupare spontaneously preferred as compared to ingroupmembers who behave in a more egalitarian way. In con-trast, we expect verbal and controlled responses towardingroup members who consistently try to be fair towardoutgroup members and display egalitarian behaviors to

Authors’ Note: Correspondence should be addressed to Luigi Castelli,Dipartimento di Psicologia dello Sviluppo e della Socializzazione, ViaVenezia, 8, 35131 Padova, Italy; e-mail: [email protected].

PSPB, Vol. 34 No. 6, June 2008 807-818DOI: 10.1177/0146167208315210© 2008 by the Society for Personality and Social Psychology, Inc.

The current article investigated how individuals evaluateingroup members displaying either ingroup bias or egal-itarian intergroup behaviors. The hypotheses predictedthat on explicit responses a preference for the egalitarianingroup member would emerge; in contrast, on morespontaneous and uncontrolled responses, a preferencefor the ingroup favoritist would result. Across four stud-ies these hypotheses were confirmed for both minimalgroups (Studies 1 and 2) and ethnic groups (Studies 3and 4). Despite a verbal preference for those whobehaved in an egalitarian way, an implicit ingroupmetafavoritism was found. Overall, results indicated thepresence of dual attitudes in the perception of ingroupmembers and the strict interconnection between inter-group behaviors and intragroup perception.

Keywords: ingroup bias; implicit processes; intragroup dynamics

Individuals have a passion for their own social groups.Both in artificially created groups (Brewer, 1979;

Otten & Wentura, 1999; Tajfel, Billig, Bundy, &Flament, 1971) and in meaningful groups (Bodenhausen,Macrae, & Garst, 1998; Maass, Salvi, Arcuri, & Semin,1989), there is a strong tendency to favor the ingroup overthe outgroup. Money, jobs, and more positive inner feel-ings are more likely to be devoted to ingroup membersthan to outgroup members, and there is little debate aboutthe strong consistency of ingroup bias, at least in the caseof majority and relatively high-status groups (see Jost,2001; Rudman, Feinberg, & Fairchild, 2002).

What is interesting to note, however, is the apparentchange of ingroup bias during the last decades. Indeed,especially in the case of normatively protected socialgroups, overt forms of ingroup bias and discrimination

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

be more positive. In other words, we test the hypothesisof an implicit ingroup metafavoritism in the form of asubtle preference for ingroup members displayingingroup bias.

Not All Ingroup Members Are Alike

The overall preference for ingroup members is mod-erated by several factors such as the identity relevance ofgroup membership (Leonardelli & Brewer, 2001),increased mortality salience (Castano, Yzerbyt, Paladino,& Sacchi, 2002), or the presence of specific threats(Stephan et al., 2002). The evaluation of ingroupmembers is also affected by other factors that pertain tothe characteristics of the specific ingroup members. Forinstance, prototypical exemplars of the ingroup are moreliked as compared to more peripheral exemplars (Hogg,Hardie, & Reynolds, 1995; Turner, Hogg, Oakes,Reicher, & Wetherell, 1987). Moreover, ingroupmembers who exhibit negative behaviors are evaluated ina particularly negative way (Branscombe, Wann, Noel,& Coleman, 1993; Marques & Yzerbyt, 1988; Marques,Yzerbyt, & Leyens, 1988). People tend to polarize judg-ments against ingroup members who engage in negativebehaviors so that such “black sheep” can be psycholog-ically placed outside the boundaries of the group,thereby leaving unaffected its general positive image. Ingeneral, social perceivers are also extremely cautiouswhen deciding whether someone can become a memberof the ingroup. Only when strict criteria are met is thetarget accepted in the ingroup (i.e., ingroup overexclu-sion effect; Leyens & Yzerbyt, 1992; Yzerbyt, Leyens,& Bellour, 1995). Again, individuals who may diminishthe positivity of the ingroup are simply left outside ofthe group or expelled from the group. This is especiallytrue when the ingroup member may be construed as adeviant and is a threat to the norms of the ingroup(Marques, Abrams, Paez, & Taboada, 1998; Scheepers,Branscombe, Spears, & Doosje, 2002). All these differ-ent lines of research show that intragroup evaluationsmay serve important functions for the management ofthe group image as a whole.

The idea has recently been advanced that the types ofsocial judgments provided in intergroup contexts mayalso affect intragroup processes (Castelli, Arcuri, &Zogmaister, 2003; Castelli, Vanzetto, Sherman, &Arcuri, 2001; Castelli, Zecchini, Sherman, & De Amicis,2005). In a series of studies, participants were presentedwith ingroup members who used either stereotypical orcounterstereotypical information to describe an out-group member as, for instance, a North African or anelderly person. Results showed that whereas partici-pants’ verbally reported judgments clearly condemnedand disapproved the use of stereotypes (see also

Greenberg & Pyszczynski, 1985; Mae & Carlston,2005), on more subtle measures there was indeed a pref-erence for the ingroup member who relied on stereotyp-ical knowledge. This further suggests that in intergroupsettings not all ingroup members are alike. Judgmentstoward ingroup members were affected by the intergroupjudgments these members had provided. Importantly,explicit responses toward stereotypers were largely neg-ative, but spontaneous conformity arose.

In the aforementioned work, the focus was centeredon a specific social behavior, namely, the use of stereo-types. The reliance on stereotypical information to com-municate with other ingroup members is a way toreaffirm commonly shared knowledge (Ruscher, 1998).Indeed, stereotypes are widespread within the groupand most group members are aware of what stereotyp-ical contents are associated with relevant outgroups(Devine, 1989; Moskowitz, Salomon, & Taylor, 2000).As such, stereotype use may signal to other groupmembers the sharing of a common background, whichmay increase the positivity of the evaluation of thesource (Castelli, Pavan, Ferrari, & Kashima, 2007).Besides the specific domain of stereotyping, it remainsunclear how intergroup behaviors may affect the per-ception of ingroup members who exhibit these behav-iors. In particular, the expression of discriminatorybehaviors, such as behaviors that favor the ingroup overthe outgroup, or the expression of egalitarian behaviorsmay have a strong impact on the evaluation of theactor. It is evident that an ingroup member who bringsresources and prestige to the ingroup, as in the case of askillful football player, will be liked more than aningroup member who is less useful to the cause of thegroup, such as a player used to make own goals.However, what happens when an individual favors theingroup by blatantly damaging the outgroup, as inthe case of a player who scores while a member of theopposing team is injured on the ground? Will this playerbe preferred as compared to a fair player who stops theaction and waits for the opponent to be recovered? Inthe current article, we focus on the specific cases ofresource allocations among minimal groups and jobselection decisions to verify how ingroup members whoeither favor the ingroup or not are evaluated on bothimplicit and explicit levels.

The Evaluation and Treatment ofIngroup Favoritists

Marques et al. (1998) recently proposed a subjectivegroup dynamics model according to which those who vio-late group norms—the deviants—are negatively evalu-ated. In contrast, ingroup members who follow thenorms are more positively evaluated. In intergroup

808 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

settings, however, who is more likely to be identified asdeviant? This relevant question arises from the difficul-ties in determining whether fairness or a discriminatorynorm will be prevailing. In many cases, discriminationis expected to be the rule. For instance, groups such asrapists or drug pushers must be discriminated against,and so they are (Crandall, Eshleman, & O’Brien, 2002;see also Franco & Maass, 1999). Processes, however,might be different for socially protected groups, as wellas for unspecified groups, such as minimal groups(Crandall et al., 2002; Jetten, Spears, & Manstead,1996). Here, social judgments must often reconcile dif-ferent needs and motivational requirements. On onehand, as stated previously, emerging social norms pre-scribe fairness in most intergroup settings. This is thecase for several minority groups, such as Blacks orhomosexuals, but it also affects more trivial group dis-tinctions. For instance, Jetten et al. (1996, Study 1)employed a minimal group paradigm and found that theingroup as a whole was evaluated more positively whenparticipants were led to believe that most ingroupmembers distributed money equally to the ingroup andoutgroup, as compared to a condition in which ingroupprofit was the norm. On the other hand, despite thisappreciation for the fair and egalitarian ingroup, whenparticipants were personally required to distributemoney to ingroup and outgroup members, they system-atically generated asymmetrical distributions so as tofavor the ingroup (Jetten et al., 1996). This suggeststhat when evaluations are abstract and (possibly) incon-sequential, such as verbally reporting a generic prefer-ence for a fair or discriminatory ingroup norm, peopleprefer fairness. In contrast, when resources have to bedistributed, people are forcefully constrained by theirgroup membership and abandon abstract principles infavor of more pragmatic reward allocation strategiesthat might benefit the ingroup.

In addition, it has been shown that people have clearexpectations about how other members of their owngroup will behave in intergroup contexts (Jetten et al.,1996). Indeed, the majority of respondents consideredingroup favoritism to be the most likely behavior per-formed by other ingroup members. In short, at least inminimal group settings, a fairness norm is positivelyevaluated but it is neither personally followed norexpected to be followed by other members of the group.This overall scenario is paradoxical and casts doubts onthe actual positive evaluation of intergroup fairness and thecorresponding derogation of those who discriminatethe outgroup in favor of the ingroup. The main goal ofthe present article is to shed light on this issue and toassess both the implicit and explicit perceptions ofingroup members who either favor the ingroup whiledamaging the outgroup or opt for more egalitarian

behaviors. This goal is pursued across four studies inwhich both minimal (Studies 1 and 2) and real (Studies3 and 4) groups are used. In the various studies, differentdependent measures will be used in order to assess howingroup members are perceived (i.e., conformity mea-sures, go/ no-go association task, memory perfor-mances). We hypothesize that participants will quicklydevelop positive implicit attitudes toward those whofavor the ingroup whereas explicit attitudes will bemore positive toward egalitarian ingroup members.

STUDY 1

Method

Participants. Sixty-four students at the University ofPadova (50 females, 14 males) participated in the studyon a voluntary basis.

Procedure. The experiment was presented as a studyon perceptual information processing. Participants wereinitially required to perform a perceptual discriminationtask. It was explained that people can be divided intotwo types—Type A and Type B—according to the waythey process briefly presented information. Hence, par-ticipants were told that two polygons would be brieflypresented on a computer screen and that the partici-pant’s task was to judge as fast and accurately as possi-ble whether these two polygons were identical ordifferent. It was stressed that speed and accuracy wereimportant because these were the basic criteria needed tocategorize individuals into Type A or Type B. Overall,participants were presented with 18 pairs of polygons.Each pair was shown for 30 ms and participants had torespond using two labeled keys on the computer key-board. Once a response was provided, the next pair ofstimuli appeared after 500 ms. At the end of the task thesoftware indicated that the processing of the responseswas ongoing, and after 2,500 ms the following sentenceappeared on the screen: “You are a Type A person.”Participants were required to write their category mem-bership on a piece of paper that was kept beside themthroughout the rest of the experiment.

Next, we told participants that we wanted them tofamiliarize themselves with some distribution matrixes(see Tajfel et al., 1971) because they would have to usethem in a subsequent phase of the study. We explainedthat the matrixes could be used to distribute points toother individuals. We further explained that to help par-ticipants become familiar with the matrixes we weregoing to show them how two previous participants in theexperiment had used these matrixes. These two previousparticipants were identified by their names—Marco and

Castelli et al. / IMPLICIT INGROUP METAFAVORITISM 809

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

Paolo—and they were both Type A persons (i.e., ingroupmembers). Marco and Paolo used the matrixes to distrib-ute points to two other unknown individuals, one ofwhom was a Type A person and the other a Type Bperson. Hence, participants saw on the computer screena series of six pages. On each page they could simultane-ously see the choices of Marco and Paolo on the samematrix. This phase was self-paced and we instructed par-ticipants to move to the next page only once they hadcarefully inspected the matrixes and the choices made bythe two previous participants. One of the two Type Apersons systematically favored the ingroup memberwhereas the other one was much more egalitarian, givingon average an equal number of points to the two targets.Whether the egalitarian Type A person was namedMarco or Paolo was counterbalanced across participants.

Next, a new supposedly perceptual task was intro-duced. Participants were told that pages with a series ofletter As would appear on the computer screen and thattheir task was to look at the letters and to say aloudhow many of them they thought there were on thescreen (see Castelli et al., 2001, for details; see alsoVaes, Paladino, Castelli, Leyens, & Giovanazzi, 2003).Each page remained visible for 5 s and participants hadto provide a response as soon as the page disappeared.Importantly, at the top of the screen, participants couldsee the estimate provided by either Marco or Paolowhen they did the same task (see Figure 1 for anexample). Overall, 36 pages were presented: On 18 tri-als the estimates of Marco were shown, and on theremaining 18 trials the estimates of Paolo were shown.Whether it was Marco or Paolo that provided the esti-mate on a specific page was counterbalanced betweenparticipants.

Finally, participants were presented with a question-naire in which they were asked a series of questions aboutthe two ingroup members. First, they were asked whetherthey would have preferred Marco’s or Paolo’s estimates asanchors in a subsequent estimation task. Next, they wereasked whether they would have preferred to meet Marcoor Paolo. The following two questions concerned howaccurate the estimates provided by Marco and Paolo wereperceived to be. Responses to these items were providedon 7-point Likert-type scales ranging from extremely inac-curate to extremely accurate. Next, we asked the extent towhich participants believed they were influenced by theestimates provided by Marco and Paolo. Again, responseswere provided on 7-point Likert-type scales. At the end ofthe questionnaire a manipulation check was introduced.Participants were asked to report which of the twoingroup members tended to favor other Type A personsinstead of being egalitarian. Finally, participants werethanked and debriefed.

Results

Data from 3 participants were not included in theanalyses because at the manipulation check they did notcorrectly remember the respective behaviors performedby the two ingroup members. Data from an additional4 participants were discarded because they did notproperly follow the instructions (e.g., they did not saythe estimates aloud). Therefore, data from 57 partici-pants were available for analyses.

Conformity measure. The conformity measure wasoperationalized as the absolute distance between theprovided anchors and the participants’ actual judg-ments. In this way, lower values correspond to greaterconformity. Hence, two scores were calculated: Thefirst score indicated conformity toward the discrimina-tory ingroup member and the other indicated confor-mity toward the egalitarian ingroup member. A pairedsample t test showed that conformity toward the dis-criminatory ingroup member was indeed stronger thanconformity toward the egalitarian ingroup member (Ms =59.4 and 62.3, SDs = 33.12 and 35.39, respectively),t(56) = 2.05, p < .05.

Explicit measures. The majority of the respondents(65%) indicated they would have preferred the egalitarian

810 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

Figure 1 Example of the pages with the As (Study 1).

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

ingroup member’s estimates rather than the discrimina-tory ingroup member’s estimates as anchors in a fol-lowing estimation task (35%; analysis of the binomialdistribution, p < .05). In addition, the majority of therespondents were more willing to interact with the egal-itarian ingroup member (67%) than with the discrimi-natory ingroup member (33%; analysis of the binomialdistribution, p < .02). No significant difference betweenthe two ingroup members emerged in relation to theperception of the accuracy of their estimates, t(55) =1.40, p = .165, and in relation to how much theresponses had been affected by their estimates, t(56) =1.59, p = .117. However, inspection of the means indi-cated that the egalitarian ingroup member was per-ceived as slightly more accurate than the discriminatoryingroup member (Ms = 3.73 and 3.50, SDs = 0.98 and1.02, respectively), as well as slightly more influential(Ms = 2.91 and 2.71, SDs = 1.19 and 1.25, respectively).Finally, we computed differential scores of spontaneousconformity toward the two targets and explicit percep-tion of accuracy and influence. The correlation betweenthe differential perception of accuracy and influence ofthe two targets was positive and significant, r(56) = .661,p < .001, indicating that the more a target was perceivedas accurate the more he was also considered to be influ-ential. Neither perceived accuracy, r(56) = –.159, p =.24, nor perceived influence, r(56) = –.097, p = .472, wassignificantly correlated with differences in actual confor-mity. Differences in actual conformity were also unre-lated to responses about the preferred ingroup memberin a subsequent estimation task, t(55) = 0.79, p = .42,and as person to meet, t(55) = 0.75, p = .45.

Discussion

Very different pictures emerged from the conformitymeasure and participants’ controlled responses. Indeed,on the questionnaire participants reported a preferencefor meeting the egalitarian ingroup member rather thanthe discriminating ingroup member. In general, theingroup member who was fair toward the outgroup wasbetter evaluated and was perceived as at least as accu-rate as the discriminating ingroup member (see Jettenet al., 1996). Besides this positive verbally reported per-ception, responses on the indirect measure of confor-mity showed a tendency to rely more on the estimates ofthe discriminating ingroup member. Indeed, in thisstudy, the ingroup member displaying ingroup bias wasactually more influential on the unrelated estimationtask. This finding suggests that blatant pro-ingroupbehaviors may be verbally disapproved, but theyenabled the actor to receive preferential treatment on anindirect measure, such as the employed conformity task.This stronger conformity toward the discriminatory

ingroup member is assumed to reflect an implicit prefer-ence for such an ingroup member. Of course, responsesin the conformity task could be at least potentially con-trolled; therefore, in the following study we directly testour main hypothesis adopting an implicit attitude mea-sure based on speeded responses, namely, a go/no-go task(see Nosek & Banaji, 2001). In accordance with theresults of the present study, we predict that discrimina-tory ingroup members will be associated with a morepositive implicit measure whereas egalitarian ingroupmembers will be preferred on the explicit measures.

STUDY 2

Method

Participants. Fifty-six students at the University ofPadova (45 females, 11 males) participated in the studyon a voluntary basis.

Procedure. Participants were initially categorized asGroup A members following the same procedureadopted in Study 1. Also, as in Study 1, participants sawTajfel et al.’s (1971) matrixes and the distributionstrategies of two ingroup members (i.e., Marco andPaolo). The main dependent measure, however, waschanged. Participants were required to perform a go/no-go association task. During this task, participants werepresented on the computer screen with four types ofstimuli: (a) the name Marco, (b) the name Paolo, (c)positive words (i.e., friendly, likable, pleasant, splendid,good, smiling, fabulous, kind, cheerful, wonderful), and(d) negative words (i.e., angry, brutal, terrible, tragic,hateful, unpleasant, dirty, ugly, nasty, dishonest). Foreach participant we assessed either the attitude towardthe egalitarian or the attitude toward the favoritistingroup member. To this end, each participant wentthrough two phases. Each phase comprised 40 trials. Inone phase, participants were presented with all fourtypes of stimuli in random order and were asked topress a key on the computer keyboard any time a giventarget name—either Marco or Paolo—or a positiveword appeared. No response was required to the otherstimuli. In the other phase, the participant was asked topress the same response key any time the target name ora negative word appeared. The relative order of thesephases, as well as the specific target name, was counter-balanced across participants. Importantly, participantswere allowed a very restricted time window to respond(i.e., 500 ms). d-prime scores (d′) were calculatedaccording to the indications provided by Nosek andBanaji (2001). Participants are expected to be more sen-sitive (i.e., easier discrimination of signal from noise)

Castelli et al. / IMPLICIT INGROUP METAFAVORITISM 811

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

when the two components of the signal are positivelyassociated than when they are not.

Finally, we investigated the verbally reported percep-tion about the two ingroup members. Participants hadto evaluate each of the two targets along a series of 9-point Likert-type scales assessing how pleasant, intelli-gent, nice, reliable, sincere, and honest the target wasperceived to be. In addition, participants reported howmuch they would have liked to be friends with each ofthe two targets and how similar they perceived them-selves to be to the two targets. Again, responses wereprovided along 9-point Likert-type scales. As in Study 1,a manipulation check measure was included to verifywhether participants actually remembered who favoredthe ingroup and who was egalitarian.

Results

Data from 2 participants were not included in theanalyses because they failed to correctly remember theactual behavior of the two ingroup members. Two addi-tional participants were removed from the analyses ofthe go/no-go association task data because they had sen-sitivity scores of 0 or below.

Go/no-go association task. For each participant, wecalculated a d-prime score both when they were asked topair the ingroup member with positive words and whenthey were asked to pair the ingroup member with nega-tive words. These two scores were then submitted to a 2(valence of the words: positive vs. negative) × 2 (target’sbehavior: egalitarian vs. ingroup bias) × 2 (specific targetname: Marco vs. Paolo) × 2 (order of the two phases:positive first vs. negative first) mixed-model ANOVA.The first factor was within participant whereas the otherswere manipulated between participants. The only signifi-cant effect that emerged was the predicted two-way inter-action between the valence of the words and the target’sbehavior, F(1, 44) = 7.88, p < .01, ηp

2 = .15. Positive

words were more easily matched with the name of theingroup favoritist rather than with the egalitarian target,t(50) = 2.75, p < .01 (see Table 1). No effect was foundfor negative words, t(50) = 0.75, p > .46.

Explicit measures. The responses to the Likert-typescales assessing how pleasant, intelligent, nice, reliable,sincere, and honest the egalitarian target was perceivedto be were combined to create a single index (α = .89).The same computation was carried out on responsesrelated to the discriminatory ingroup member (α = .90).The two means were almost identical, t(53) < 1, ns, indi-cating no preference for either of the two targets (seeTable 1). No significant effect emerged for the differentialdesire to become friends with the two targets, t(53) =1.60, p = .11, and for the difference in perceived similar-ity, t(53) < 1, ns. Responses to the go/no-go associationtask were marginally related to the single index of eval-uation, r(52) = .24, p = .076, but unrelated to both thedesire to become friends, r(52) = –.06, p = .66, and per-ceived similarity, r(52) = .12, p = .36.

Discussion

Results from the go/no-go association task indicatedthat the ingroup member favoring the ingroup elicitedmore positive spontaneous responses than the egalitarianingroup member. A strategy of systematic favoritism forthe ingroup was thus shown to increase both conformity(Study 1) and implicit preference (Study 2). Overall, thesefindings provide strong support for the implicit ingroupmetafavoritism hypothesis. Results from the explicit mea-sures were partially at odds with the preference for theegalitarian target that emerged in Study 1. Indeed, in thecurrent study, the two targets were evaluated in a similarway. A likely explanation for this discrepancy lies in the response format. Whereas participants in Study 1were required to make forced choices between the twotargets, in Study 2 each target was separately evaluatedon Likert-type scales. This means that in the currentstudy, participants could provide similar positive judg-ments toward both ingroup members without committingthemselves to favor one over the other. In contrast, whena preference is required, as in Study 1, the egalitarianingroup member seems to be the preferred target.

STUDY 3

In Studies 1 and 2 we focused on minimal groups sothat favoritism and equality were shown toward noveland undefined groups. It thus remains unclear whether theimplicit ingroup metafavoritism is confined to the veryspecific context of minimal groups. In these cases, thetwo observed ingroup members distributed unspecified

812 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

TABLE 1: Means and Standard Deviations (in Parentheses) ofImplicit and Explicit Responses in Study 2

Egalitarian Favoritist

M (SD) M (SD)

d-prime positive words 1.34 2.15(0.78) (1.24)

d-prime negative words 1.93 1.69(1.21) (1.09)

Explicit evaluation 5.33 5.37(0.96) (1.49)

Desire to become friends 5.29 4.98(0.90) (1.23)

Perceived similarity 4.77 4.70(1.47) (1.42)

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

“points,” and participants had no clue about the realconsequences of the discriminatory behavior. It isarguable that participants perceived both the discrimi-natory and egalitarian behaviors as relatively inconse-quential and trivial. Quite a different scenario mightemerge when highly relevant resources, such as jobs ormoney, are differentially distributed. In addition, mini-mal groups are, by definition, constituted by anony-mous group members and there is no a priori specificnormative protection of the outgroup. In contrast, dis-crimination toward most relevant outgroups is morallyand sometimes legally sanctioned by shared socialnorms (Crandall et al., 2002). Accordingly, the implicitingroup metafavoritism hypothesis needs to be tested inmore socially relevant scenarios. Therefore, in the fol-lowing study, we tested whether ingroup membersshowing ingroup bias are preferred in meaningful socialgroups such as one’s own national group and an ethnicstigmatized outgroup. To this end, participants werepresented with an Italian (i.e., ingroup member) whogives a job either to an Italian candidate or to a NorthAfrican candidate. Again, we predicted that at animplicit and spontaneous level the ingroup member whofavors the ingroup would be preferred whereas at theexplicit and controlled level the ingroup member who isfair toward the outgroup would be preferred or at leastevaluated equally well.

Method

Participants. Sixty-three Italian students (53 females,10 males) at the University of Padova participated in thestudy as a partial course requirement.

Procedure. The experiment was presented as a studyon decision-making processes in the domain of job selec-tion. In particular, it was said that two candidates—anItalian and a North African—applied for the same job and they had similar curricula and skills. There weretwo conditions, manipulated between participants, so that half the participants were told that the job selector—an Italian young man named Marco—chosethe Italian candidate because of his nationality. In con-trast, in the other experimental condition participantswere told the job selector preferred the North Africancandidate because, as it was explained, he wanted to befair and egalitarian so as to provide equal opportunitiesto everyone.

After this initial presentation, all participants wererequired to perform the go/no-go association task. As inthe previous study, participants were presented with fourdifferent types of stimuli on the computer screen: thename Marco, the name Paolo, positive words, and nega-tive words. Again, each participant went through twophases. Each phase comprised 40 trials. In the first phase

participants were required to press a key on the computerkeyboard any time the name Marco (i.e., the job selector)or a positive word appeared. No response was requiredto all other types of stimuli (i.e., the name Paolo and neg-ative words). In the second phase, participants wererequired to press the key every time the name Marco or anegative word appeared. As in Study 2, participants wereallowed a very restricted time window of 500 ms torespond. The order of the two phases was counterbal-anced across participants. Sensitivity scores, indexed byd-prime, were calculated as in Study 2.

Participants were then asked to fill in a paper-and-pencil questionnaire concerning their perception of thepresented Italian job selector. Four items addressed theevaluation of the target (e.g., “How much do you likeMarco?” “Are you proud of him?”), one item addressedthe willingness to interact with him, and one item wasrelated to his perceived typicality (i.e., “How much doyou think Marco is similar to the typical Italian?”). Allresponses were provided on 7-point Likert-type scalesranging from –3 (not at all) to +3 (very much). Next,participants had to report the percentage of Italians thatthey thought would have behaved in the same way as thepresented job selector. Finally, participants were askedto report the emotions they felt in response to the behav-ior performed by the ingroup job selector. Overall, 10items were presented (i.e., surprise, discomfort, regret,guilt, embarrassment, blame, shame, happiness, satis-faction, ease) and responses were provided on 7-pointLikert-type scales ranging from –3 to +3.

Results

Go/no-go association task. Data from 1 partici-pant were removed because he did not properly fol-low the instructions. Therefore, data from 62participants were used in the analyses. d-prime scoreswere calculated for the two tasks where the targetname had to be paired with positive words and withnegative words. These two indices were submitted toa 2 (valence of the target words: positive vs. negative)× 2 (target’s behavior: egalitarian vs. favoritist) × 2(order of the two phases) ANOVA with the first fac-tor within participant and the last two between par-ticipants. No main effect was significant, and theonly significant effect was a two-way interactionbetween the valence of the target words and the typeof behavior performed by the target, F(1, 58) = 5.85,p < .05, ηp

2 = .09 (see Table 2). Positive words weremore easily paired to the ingroup member whofavored the ingroup than to the egalitarian ingroupmember, t(60) = 1.68, p < .05, one-tailed. In contrast,no effect emerged for negative words, t(60) = –1.16,p > .24.

Castelli et al. / IMPLICIT INGROUP METAFAVORITISM 813

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

Explicit measures. Responses to the four itemsassessing the evaluation of the target showed good con-sistency (α = .86) and were thus averaged. An indepen-dent sample t test showed that the target was evaluatedmore positively when he behaved in an egalitarian wayrather than when he favored the ingroup, t(61) = 2.33,p < .05 (see Table 2). No difference emerged in partici-pants’ willingness to interact with the egalitarian target.In contrast, the ingroup favoritist was perceived as moretypical of the group, t(61) = 4.01, p < .001, and partic-ipants reported that a higher percentage of other groupmembers would behave like him rather than like theegalitarian ingroup member, t(61) = 6.87, p < .001.These latter results indicate that consistent with thefindings of Jetten et al. (1996), ingroup favoritism is theexpected behavior.

Next, the emotions elicited by the choice of the jobselector were analyzed. The responses related to nega-tive emotions were averaged (α = .81) as were theresponses to positive emotions (α = .85). The surprisereaction was uncorrelated to responses to the otheritems and it was thus considered separately. As for neg-ative emotions, they were more accentuated when theingroup member favored the ingroup to the detrimentof the outgroup member, t(61) = 2.60, p < .05. In

contrast, positive emotions were more accentuatedwhen the ingroup members behaved in an egalitarianway, t(61) = 4.04, p < .001. Finally, a significant effectwas found for surprise, t(61) = 2.21, p < .05. Indeed,participants were less surprised when the ingroupmember selected another member of the ingroup thanwhen he selected an outgroup member. This furtherdemonstrates that ingroup favoritism is perceived as themost likely and predictable behavior. As shown in Table2, in no cases were explicit responses correlated withimplicit responses (all ps > .33).

Discussion

The pattern of results confirms that even in mean-ingful groups, such as national groups, the displacementof egalitarian or ingroup-favoring behaviors producesopposite effects on the perceiver’s implicit and explicitreactions. Therefore, implicit ingroup metafavoritismdoes not seem to be confined to minimal group settings.

STUDY 4

The studies presented so far compared reactionstoward two ingroup members—favoring the ingroup orbeing egalitarian—by using a conformity measure and thego/no-go association task. In the current study, we aimedat generalizing the effect to still another implicit attitudemeasure, and to this end we employed a different researchstrategy based on an opposition memory paradigm (seeCastelli, Zogmaister, Smith, & Arcuri, 2004, Study 4).This strategy allows us to assess the role of consciousmemory. Participants had to initially learn whether aseries of ingroup members hired an Italian candidate or aNorth African candidate. Later, half the participants wereshown members of a social group that were very positive(i.e., child counselors) and half were shown members thatwere extremely negative (i.e., usurers). Finally, partici-pants were shown all the pictures and asked to identifyonly the members of the positive or negative group. Wehypothesized that if participants had a spontaneous affec-tive reaction toward the ingroup members on the basis oftheir behaviors, there would be a higher number of intru-sions into an evaluatively congruent group than into anevaluatively incongruent group. More specifically, we pre-dicted that ingroup members who favored the ingroupwould later be more easily confounded with the membersof the positive group rather than with the members of thenegative group.

Method

Participants. Ninety-two Italian students (71females, 21 males) at the University of Padova took partin the study on a voluntary basis.

814 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

TABLE 2: Means and Standard Deviations (in Parentheses) of Implicitand Explicit Responses in Study 3, and Correlations BetweenExplicit Responses and a Single Index of Implicit PerceptionBased on d-Prime Scores [(d-prime positive words) –(d-prime negative words)]

Egalitarian Favoritist CorrelationWith the

M M Implicit(SD) (SD) Measure (r)

d-prime positive words 2.04 2.57(1.07) (1.36)

d-prime negative words 1.89 2.20(0.85) (1.22)

Evaluation 0.15 –0.47 –0.102(0.93) (1.18)

Willingness to interact 0.77 0.97 –0.056(1.43) (1.22)

Typicality 0.06 1.69 –0.031(1.44) (1.77)

Percentage of Italians 45.97 78.19 0.125who would behavelike the target

(13.72) (22.53)Negative emotions –0.47 0.23 0.008

(0.90) (1.23)Positive emotions –0.03 –1.36 –0.037

(1.27) (1.33)Surprise –0.35 –1.19 –0.113

(1.62) (1.35)

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

Procedure. As in Study 3, the experiment was intro-duced as a study on decision-making processes in thedomain of job selection. Again, it was said that two can-didates—an Italian and a North African—applied forthe same job and that they had similar curricula andskills. It was further explained that some Italian selec-tors choose the Italian candidate (i.e., an ingroupmember), named Giovanni, whereas other selectors pre-ferred the North African candidate (i.e., an outgroupmember), named Mohammad. Thus, we presented on acomputer screen a series of 40 pictures portraying head-and-shoulders White men who were said to be the Italianselectors who evaluated the curricula of Giovanni andMohammad. Pictures were presented one after theother, in random order, for 5 s (interstimulus interval[ISI] = 1 s), and below each picture participants couldread the name of the candidate chosen by that selector(i.e., either Giovanni or Mohammad). Twenty selectorschose the Italian candidate and the other 20 chose theNorth African candidate. The target chosen by eachselector was balanced across participants. Next, partic-ipants were shown an additional 20 pictures of Whitemen. Half the participants were told that the picturesportrayed “child counselors” whereas the other halfwere told that the pictures portrayed loan sharks andwere labeled as “usurers.” Again, each picture wasshown for 5 s (ISI = 1 s) in random order. After an inter-val of about 3 min in which participants performed anintervening task, they were shown again, in randomorder, all 60 pictures seen thus far and the task was toidentify the members of the last group they were pre-sented (i.e., child counselors or usurers). Of course, nogroup label was shown during this phase. Participantshad to press one of two labeled keys on the computerkeyboard (member vs. nonmember) to provide theirresponses. Participants were allowed as much time asthey needed, and accuracy rather than speed was stressed.

The main dependent variable of this study is the pro-portion of members of the two groups of selectors seenin the first experimental phase that are later erroneouslyincluded in the third group (e.g., number of egalitarianselectors erroneously included in the target groupdivided by the total number of egalitarian selectors pre-sented). In other words, we looked separately at theproportion of selectors who chose the Italian candidateand at the proportion who chose the North African can-didate who were erroneously included in the targetgroup. We predicted that selectors who chose aningroup member would more likely be included in thepositive target group whereas selectors who chose anoutgroup member would more likely be included in thenegative target group. The idea is that whereas a correctrecollection should avoid the inclusion, when recollec-tion is at least partially impaired, the implicit effects of

the affective components should bias responses. Thevirtue of this procedure is that the effects of explicit andimplicit memory are thus placed in opposition (Castelliet al., 2004; Jacoby, Kelley, Brown, & Jasechko, 1989;Jacoby, Yonelinas, & Jennings, 1997).

Finally, participants were asked to indicate on a con-tinuum (i.e., a 10-cm line) their preference either for theItalian or the North African candidate (from a strongpreference toward Giovanni to a strong preferencetoward Mohammad), and their attitude toward NorthAfricans (from extremely negative to extremely positive).

Results

A 2 (chosen candidate: Italian vs. North African) × 2(target group valence) mixed-model ANOVA was car-ried out on the portion of exemplars erroneouslyincluded in the target group. No significant main effectemerged (both Fs < 1). Importantly, errors were equallylikely for the two types of ingroup members, suggestingthat there was no better memory for those who favoredthe ingroup than for those who chose the North Africancandidate. The predicted interaction effect was signifi-cant, F(1, 90) = 4.03, p < .05, η2

p = .04. In particular,when the target group was positive, errors were morelikely for selectors who favored the ingroup by givingthe job to one of its members (Ms = 0.31 and 0.25,SDs = 0.13 and 0.13, respectively), t(46) = 2.76, p < .01.In contrast, when the target group was negative, errorswere equally likely regardless of whether the selectorchose the Italian or the North African candidate (Ms =0.27 and 0.28, both SDs = 0.13 and 0.13, respectively),t(44) = –0.23, p > .81.

Data from the explicit questions showed no significantpreference for either of the two job applicants and alargely positive attitude toward North Africans in general;indeed, mean response on the continuum was significantlydifferent from the scale midpoint in a positive direction,t(91) = 3.71, p < .001, indicating that North Africans werepositively evaluated. A single index of relative preferencefor the ingroup member who favored the ingroup asopposed to the one who favored the outgroup was com-puted based on errors in the memory task. This index wasnot correlated with responses to the two explicit scales,r(92) = –.087, p > .41, and r(92) = –.073, p > .49.

Discussion

The results from the present study support the existenceof an implicit ingroup metafavoritism. Most important,current findings demonstrated that a conscious memory ofthe initial behavior is not necessary for the ingroup metafa-voritism to emerge. Indeed, even when memory wasimpaired participants showed a preference for ingroupmembers who gave jobs to other ingroup members.

Castelli et al. / IMPLICIT INGROUP METAFAVORITISM 815

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

GENERAL DISCUSSION

The present research clearly demonstrates that indi-viduals might have differentiated spontaneous intra-group responses based on the behavior of ingroupmembers in intergroup contexts. In the case of bothminimal groups and real groups, ingroup members whofavored the ingroup were spontaneously preferred, eventhough verbal responses were not consistent with thispreference. This finding indicates the relevance of theimplicit–explicit distinction not just in relation to directintergroup perception but also in relation to intragroupperception. What emerges is a clear implicit ingroupmetafavoritism. Therefore, the present findings confirmcurrent models that underline the crucial role played byintragroup dynamics (Abrams, Marques, Brown, &Henson, 2000; Levine, 1989; Marques et al., 1998;Turner et al., 1987) and extend them in two significantways. First, we show that intragroup perception needs tobe addressed on two separate levels, namely, on con-trolled verbal responses as well as on a more implicit level.In several cases, these two levels are likely to go hand inhand and to correlate with each other. For instance, groupmembers who excel in sports will likely be admired onboth controlled and spontaneous responses. However,when behaviors are performed in sensitive domains, suchas intergroup contexts, responses may appear to becomedissociated. Second, we examine what occurs when theperceiver observes ingroup members involved in inter-group situations and show that intergroup fairness is val-ued on a controlled level but spontaneous responsessignal a preference for ingroup members who displayingroup favoritism. Accordingly, egalitarian normswould become the standard for the explicit evaluationof ingroup members in intergroup contexts so that dis-criminatory members are likely to be considered asdeviants and be negatively evaluated (Marques et al.,1998). In line with the subjective group dynamic modelby Marques et al. (1998), members who reinforce a pos-itively valued egalitarian norm are valued, whereas thosewho undermine such a norm are judged in a negativeway. Importantly, however, the intergroup behaviorsperformed by ingroup members seem to be sponta-neously evaluated against different norms at an implicitlevel. Indeed, as shown in the present studies, ingroupfavoritists were markedly preferred as compared to egal-itarian ingroup members on spontaneous responses.This further suggests that just as people often hold dualattitudes toward social targets (Wilson et al., 2000), theymay also have dual standards for the evaluation ofingroup members in intergroup contexts.

The overall pattern suggests that intergroup fairnessmay be associated with social costs in terms of intra-group reputation. Whereas efforts aimed at providing

resources to outgroup members may receive publicapproval, an egalitarian ingroup member may also riskbeing automatically devalued. This implies that supportof integration policies—such as affirmative action—may invite long-run detrimental effects for those whosupport them in terms of intragroup regard, at least ascompared to ingroup favoritists.

In the current work, we examined the perception ofrelatively high-status-group members toward discrimi-natory or egalitarian ingroup members. However, itremains to be explored whether similar effects wouldemerge for low-status groups (see Jost, 2001). Indeed,low-status minorities often show automatic ingroupdevaluation (Rudman et al., 2002); as a consequence,they might also show implicit outgroup metafavoritism,which is a preference for ingroup members who favorthe outgroup rather than the ingroup. In general, therelative status of the involved groups might be particu-larly relevant, and the strongest implicit ingroupmetafavoritism is expected to emerge when the statussuperiority of the ingroup is maximized.

Developmental Origin of an Ingroup Metafavoritism

We recently analyzed the emergence of behaviors thatare consistent with ingroup metafavoritism among veryyoung children (i.e., 4-7 years old; Castelli, De Amicis, &Sherman, 2007). The studies conducted with youngchildren may prove to be particularly useful becausebefore 6-7 years of age verbal and spontaneous responsestend to be similar and not yet differentiated (see Baron &Banaji, 2006). Therefore, verbal responses become partic-ularly informative. We found that children displayed astrong preference for ingroup peers who consistentlyfavored the ingroup in their playmate choices as com-pared to ingroup peers who were more open toward aracial outgroup (i.e., Blacks) and played with its members.These findings indicate that even very young children paycareful attention to the intergroup behaviors performedby ingroup members and that behaviors that support theingroup are valued and rewarded. Therefore, beginning inchildhood we observe a preference both for the ingroup(Aboud, 1988; Nesdale & Flesser, 2001) and for ingroupfavoritists. This does not necessarily imply that attitudestoward the outgroup or toward egalitarian ingroupmembers will be negative at an absolute level. In contrast,the focus is set on the ingroup. As suggested by Brewer’s(1999; see also Aboud, 2003) “primacy of the ingrouphypothesis,” individuals seem to be primarily concernedabout building or maintaining good intragroup relationsand a positive image of the ingroup. Mutual trust, coop-eration, and relative ingroup favoritism may serve thiscrucial function. The ingroup metafavoritism may thus beinstrumental to achieve this goal.

816 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

Interestingly, the findings obtained for preschool-agedchildren (Castelli, De Amicis, et al., 2007) resemble whatwas observed here from adults’ spontaneous responses.One intriguing possibility is that early intergroup experi-ences shape later implicit attitudes, as has been recently sug-gested (Greenwald & Banaji, 1995; Rudman, 2004; Wilsonet al., 2000) and empirically demonstrated (Rudman &Goodwin, 2004; Rudman, Phelan, & Heppen, 2007),whereas explicit attitudes are more likely to reflect recentand accessible knowledge. As such, the early tendency tofavor the ingroup would continue to emerge from adults’automatic responses, whereas explicit responses would bemore sensitive to the pressure of egalitarian normativeconcerns. This would explain the observed inconsistencybetween explicit and implicit responses by assuming thatthey stem from different sources of information. As thechild grows up, his or her explicit evaluations wouldquickly start to incorporate the pressures toward intergroupfairness, but implicit attitudes would change more slowlyand, in most cases, would continue to signal a preferencefor ingroup favoritism (see Gregg, Seibt, & Banaji, 2006).

Positive–Negative Asymmetry in theEvaluation of Ingroup Members

Across the four studies, we found that ingroup favoritistsand egalitarian ingroup members were differentiated in apositive domain but not in a negative domain. For instance,ingroup members who favored the ingroup were more eas-ily paired to positive words than was the egalitarian target,but no difference was found for negative words (Studies 2and 3). Similarly, ingroup favoritists were more likely includedin a positive group than were egalitarian ingroup members,but erroneous inclusions in a negative group did not differ(Study 4). This finding closely resembles the positive–negativeasymmetry reported by Mummendey, Otten, Berger, andKessler (2000). They found that it is primarily in a positivedomain that the differentiation between groups emerges.Along the same lines, we found evidence that the differenti-ation between members of one’s own group who behavedifferently in intergroup settings is more likely to be foundin the attribution of positive rather than negative features.As discussed previously, this suggests that egalitarianingroup members are not actively discriminated against.However, positive features are spontaneously more likelyreserved for ingroup members who prove with their behav-ior to have the best interests of the ingroup at the top oftheir minds. This is indeed implicit ingroup metafavoritism.

REFERENCES

Aboud, F. E. (1988). Children and prejudice. Oxford, UK: BasilBlackwell.

Aboud, F. E. (2003). The formation of ingroup attitudes and out-group prejudice in young children: Are they distinct attitudes?Developmental Psychology, 39, 48-60.

Abrams, D., Marques, J. M., Brown, N., & Henson, M. (2000). Pro-norm and anti-norm deviance within and between groups. Journalof Personality and Social Psychology, 78, 906-912.

Baron, A. S., & Banaji, M. R. (2006). The development of implicitattitudes: Evidence of race evaluations from ages 6 and 10 andadulthood. Psychological Science, 17, 53-58.

Bodenhausen, G. V., Macrae, C. N., & Garst, J. (1998). Stereotypesin thought and deed: The social-cognitive origins of intergroupdiscrimination. In C. Sedikides, J. Schopler, & C. A. Insko (Eds.),Intergroup cognition and intergroup behavior (pp. 311-335).Mahwah, NJ: Lawrence Erlbaum.

Branscombe, N. R., Wann, D. L., Noel, J. G., & Coleman, J. (1993). In-group or out-group extremity: Importance of the threatened socialidentity. Personality and Social Psychology Bulletin, 19, 381-388.

Brewer, M. B. (1979). In-group bias in the minimal intergroup situa-tion: A cognitive-motivational analysis. Psychological Bulletin, 86,307-324.

Brewer, M. B. (1999). The psychology of prejudice: In-group love orout-group hate? Journal of Social Issues, 55, 429-444.

Castano, E., Yzerbyt, V., Paladino, M. P., & Sacchi, S. (2002). Ibelong, therefore, I exist: Ingroup identification, ingroup entitativ-ity, and ingroup bias. Personality and Social Psychology Bulletin,28, 135-143.

Castelli, L., Arcuri, L., & Zogmaister, C. (2003). Perceiving ingroupmembers who use stereotypes: Implicit conformity and similarity.European Journal of Social Psychology, 33, 163-175.

Castelli, L., De Amicis, L., & Sherman, S. J. (2007). The loyal membereffect: On the preference for ingroup members who engage inexclusive relations with the ingroup. Developmental Psychology,43, 1347-1359.

Castelli, L., Pavan, G., Ferrari, E., & Kashima, Y. (2007). The stereo-type and the chameleon: On the subtle processes that reinforcestereotype use. Manuscript submitted for publication.

Castelli, L., Vanzetto, K., Sherman, S. J., & Arcuri, L. (2001). Theexplicit and implicit perception of ingroup members who usestereotypes: Blatant rejection but subtle conformity. Journal ofExperimental Social Psychology, 37, 419-426.

Castelli, L., Zecchini, A., Sherman, S. J., & De Amicis, L. (2005). Theimpact of implicit prejudice about the elderly on the reaction tostereotype confirmation and disconfirmation. Current Psychology,24, 134-146.

Castelli, L., Zogmaister, C., Smith, E. R., & Arcuri, L. (2004). On theautomatic evaluation of social exemplars. Journal of Personalityand Social Psychology, 86, 373-387.

Crandall, C. S., Eshleman, A., & O’Brien, L. (2002). Social norms andthe expression and suppression of prejudice: The struggle forinternalization. Journal of Personality and Social Psychology, 82,359-378.

Crosby, F., Bromley, S., & Saxe, L. (1980). Recent unobtrusive stud-ies of Black and White discrimination and prejudice: A literaturereview. Psychological Bulletin, 87, 546-563.

Devine, P. (1989). Stereotypes and prejudice: Their automatic andcontrolled components. Journal of Personality and SocialPsychology, 56, 5-18.

Dovidio, J. F., & Gaertner, S. L. (2000). Aversive racism and selectiondecisions: 1989 and 1999. Psychological Science, 11, 315-319.

Dovidio, J. F., Kawakami, K., & Gaertner, S. L. (2002). Implicit andexplicit prejudice and interracial interaction. Journal of Personalityand Social Psychology, 82, 62-68.

Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995).Variability in automatic activation as an unobtrusive measure ofracial attitudes: A bona fide pipeline? Journal of Personality andSocial Psychology, 69, 1013-1027.

Franco, F. M., & Maass, A. (1999). Intentional control over preju-dice: When the choice of the measure matters. European Journalof Social Psychology, 29, 469-477.

Greenberg, J., & Pyszczynski, T. (1985). The effects of an overheardracial slur on evaluations of the target: How to spread a social dis-ease. Journal of Experimental Social Psychology, 21, 61-72.

Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition:Attitudes, self-esteem, and stereotypes. Psychological Review,102, 4-27.

Castelli et al. / IMPLICIT INGROUP METAFAVORITISM 817

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from

Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998).Measuring individual differences in implicit cognition: TheImplicit Association Test. Journal of Personality and SocialPsychology, 74, 1464-1480.

Gregg, A. P., Seibt, B., & Banaji, M. R. (2006). Easier done thanundone: Asymmetry in the malleability of implicit preferences.Journal of Personality and Social Psychology, 90, 1-20.

Hogg, M. A., Hardie, E. A., & Reynolds, K. J. (1995). Prototypicalsimilarity, self-categorization, and depersonalized attraction: Aperspective on group cohesiveness. European Journal of SocialPsychology, 25, 159-177.

Jacoby, L. L., Kelley, C. M., Brown, J., & Jasechko, J. (1989).Becoming famous overnight: Limits on the ability to avoid uncon-scious influences of the past. Journal of Personality and SocialPsychology, 56, 326-338.

Jacoby, L. L., Yonelinas, A. P., & Jennings, J. M. (1997). The relationbetween conscious and unconscious (automatic) influences: A dec-laration of independence. In J. D. Cohen & J. W. Schooler (Eds.),Scientific approaches to consciousness (pp. 13-48). Mahwah, NJ:Lawrence Erlbaum.

Jetten, J., Spears, R., & Manstead, A. R. (1996). Intergroup normsand intergroup discrimination: Distinctive self-categorization andsocial identity effects. Journal of Personality and SocialPsychology, 71, 1222-1233.

Jost, J. T. (2001). Outgroup favoritism and the theory of system jus-tification: A paradigm for investigating the effects of socioeco-nomic success on stereotype content. In G. B. Moskowitz (Ed.),Cognitive social psychology: The Princeton Symposium on theLegacy and Future of Social Cognition (pp. 89-102). Mahwah,NJ: Lawrence Erlbaum.

Leonardelli, G. J., & Brewer, M. B. (2001). Minority and majoritydiscrimination: When and why. Journal of Experimental SocialPsychology, 37, 468-485.

Levine, J. M. (1989). Reaction to opinion deviance in small groups. InP. B. Paulus (Ed.), Psychology of group influence (pp. 187-231).Hillsdale, NJ: Lawrence Erlbaum

Leyens, J.-P., & Yzerbyt, V. Y. (1992). The ingroup overexclusion effect:Impact of valence and confirmation on stereotypical informationsearch. European Journal of Social Psychology, 22, 549-569.

Maass, A., Salvi, D., Arcuri, L., & Semin, G. R. (1989). Language usein intergroup context: The linguistic intergroup bias. Journal ofPersonality and Social Psychology, 57, 981-993.

Mae, L., & Carlston, D. E. (2005). Hoist your own petard: Whenprejudice remarks are recognized and backfire on speakers.Journal of Experimental Social Psychology, 41, 240-255.

Marques, J. M., Abrams, D., Paez, D., & Taboada, C. (1998). Therole of categorization and in-group norms in judgments of groupsand their members. Journal of Personality and Social Psychology,75, 976-988.

Marques, J. M., & Yzerbyt, V. Y. (1988). The black sheep effect:Judgmental extremity toward ingroup members in inter- and intra-group situations. European Journal of Social Psychology, 18, 287-292.

Marques, J. M., Yzerbyt, V. Y., & Leyens, J.-P. (1988). The “black-sheep effect”: Extremity of judgments toward ingroup members asa function of group identification. European Journal of SocialPsychology, 18, 1-16.

Moskowitz, G. B., Salomon, A. R., & Taylor, C. M. (2000).Preconsciously controlling stereotyping: Implicitly activated egali-tarian goals prevent the activation of stereotypes. SocialCognition, 18, 151-177.

Mummendey, A., Otten, S., Berger, U., & Kessler, T. (2000). Positive-negative asymmetry in social discrimination: Valence of evaluationand salience of categorization. Personality and Social PsychologyBulletin, 26, 1258-1270.

Nesdale, D., & Flesser, D. (2001). Social identity and the developmentof children’s group attitudes. Child Development, 72, 506-517.

Nosek, B., & Banaji, M. R. (2001). The go/no-go association task.Social Cognition, 19, 625-664.

Otten, S., & Wentura, D. (1999). About the impact of automaticity inthe minimal group paradigm: Evidence from affective primingtasks. European Journal of Social Psychology, 29, 1049-1071.

Rudman, L. A. (2004). Sources of implicit attitudes. CurrentDirections in Psychological Science, 13, 79-82.

Rudman, L. A., Feinberg, J., & Fairchild, K. (2002). Minoritymembers’ implicit attitudes: Automatic ingroup bias as a functionof group status. Social Cognition, 20, 294-320.

Rudman, L. A., & Goodwin, S. A. (2004). Gender differences in auto-matic in-group bias: Why do women like women more than men likemen? Journal of Personality and Social Psychology, 87, 494-509.

Rudman, L. A., Phelan, J. E., & Heppen, J. B. (2007). Developmentalsources of implicit attitudes. Personality and Social PsychologyBulletin, 33, 1700-1713.

Ruscher, J. B. (1998). Prejudice and stereotyping in everyday com-munication. In M. P. Zanna (Ed.), Advances in experimentalsocial psychology (Vol. 30, pp. 241-307). New York: AcademicPress.

Scheepers, D., Branscombe, N. R., Spears, R., & Doosje, B.(2002). The emergence and effects of deviants in low and highstatus groups. Journal of Experimental Social Psychology, 38,611-617.

Stephan, W. G., Boniecki, K. A., Ybarra, O., Bettencourt, A., Ervin,K. S., Jackson, L. A., et al. (2002). The role of threats in the racialattitudes of Blacks and Whites. Personality and Social PsychologyBulletin, 28, 1242-1254.

Tajfel, H., Billig, M. G., Bundy, H. P., & Flament, C. I. (1971). Socialcategorization and intergroup behavior. European Journal ofSocial Psychology, 1, 149-178.

Turner, J. C., Hogg, M., Oakes, P. J., Reicher, S. D., & Wetherell, M. S.(1987). Rediscovering the social group: A self-categorizationtheory. Oxford, UK: Basil Blackwell.

Vaes, J., Paladino, M. P., Castelli, L., Leyens, J.-P., & Giovanazzi, A.(2003). On the behavioral consequences of infra-humanization: Theimplicit role of uniquely human emotions in intergroup relations.Journal of Personality and Social Psychology, 85, 1016-1034.

Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dualattitudes. Psychological Review, 107, 101-126.

Yzerbyt, V. Y., Leyens, J.-P., & Bellour, F. (1995). The ingroupoverexclusion effect: Identity concerns in decisions about groupmembership. European Journal of Social Psychology, 25, 1-16.

Received May 15, 2007Revision accepted November 26, 2007

818 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

at UNIV DEGLI STUDI MILANO - BICOCCA on October 23, 2008 http://psp.sagepub.comDownloaded from