Prejudice and Intergroup Attributions - Archive ouverte HAL

HAL Id: hal-00571614https://hal.archives-ouvertes.fr/hal-00571614

Submitted on 1 Mar 2011

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Prejudice and Intergroup Attributions: The Role ofPersonalization and Performance Feedback

Nurcan Ensari, Norman Miller

To cite this version:Nurcan Ensari, Norman Miller. Prejudice and Intergroup Attributions: The Role of Personalizationand Performance Feedback. Group Processes and Intergroup Relations, SAGE Publications, 2005, 8(4), pp.391-410. �10.1177/1368430205056467�. �hal-00571614�

https://hal.archives-ouvertes.fr/hal-00571614

https://hal.archives-ouvertes.fr

Prejudice and IntergroupAttributions: The Role ofPersonalization andPerformance Feedback

Nurcan EnsariAlliant International University

Norman MillerUniversity of Southern California

We manipulated personalization and group performance feedback to examine their effects onintergroup attributions and prejudice. Following high or low levels of personalized contact witha typical out-group member, participants learned either that the out-group had generallysucceeded or that the in-group had failed at the participant’s task. Under high personalizationand out-group success, participants exhibited less attributional bias in explaining the success ofnew out-group job applicants and less prejudice toward them than those under lowpersonalization. By contrast, when one’s in-group had failed, we found similar favorabilitytoward in-group and out-group job applicants. Importantly, when ability attributions andfriendliness were separately combined with subjective personalization, both combinationsmediated the effects of manipulated personalization in reducing prejudice toward new out-group persons.

keywords attribution, feedback, personalization, prejudice

Group Processes &Intergroup Relations

2005 Vol 8(4) 391–410

ME TA-A N A L Y S E S confirm that cooperativecontact improves attitudes and behavior towardout-group members (Johnson, Johnson, &Maruyama, 1984; Miller & Davidson-Podgorny,1987; Pettigrew & Tropp, 2000). They alsosupport the view that personalized contactaugments the beneficial effect of cooperativeinteraction (Johnson et al., 1984; Miller &Davidson-Podgorny, 1987). By personalization,we mean interpersonal interactions that elicitself-disclosure (Miller, 2002), promote self/other comparisons (Brewer & Miller, 1984),and induce empathy (Batson et al., 1997;

Galinsky & Moskowitz, 2000). Thus, we concep-tualize personalization as more than merereceipt of individuating information, which canbe provided by a third party in the absence of

GPIR

Copyright © 2005 SAGE Publications(London, Thousand Oaks, CA and New Delhi)8:4; 391–410; DOI: 10.1177/1368430205056467

Authors’ noteAddress correspondence to Nurcan Ensari,Alliant International University, 1000 S.Fremont Ave., Alhambra, CA 91803, USA or toNorman Miller, Dept. of Psychology, SGM 501,University of Southern California, Los Angeles,CA 90089, USA [email: [email protected] or [email protected]]

04 Ensari 056467 (bc-s) 30/9/05 1:52 pm Page 391

any direct social interaction. Here, however, wefocus only on one of its components: theprejudice-reducing effects of self-disclosure bythe out-group person with whom one is inter-acting. More important, we examined how itaffects attributional bias and how such biasmediates prejudice under two key conditionsthat are likely to elicit differential attributionaldistortion: out-group task success and in-grouptask failure.

Personalization and the relationbetween the decategorization andmutual intergroup differentiationmodels

There are a number of reasons for suspectingthat self-disclosure during contact with an out-group member will reduce prejudice. Self-disclosure induces trust, which in turn reducesthe anxiety and discomfort that often character-ize intergroup settings (Ensari & Miller, 2002;Stephan & Stephan, 1985). In addition, it hasmotivational effects, eliciting an inclination toreciprocate the observed self-disclosure (Miller,2002). Finally, it can promote perceptions ofsimilarity and familiarity, and thereby permitbetter processing of individuating informationabout persons irrespective of their socialcategory (Rothbart & John, 1985; Wilder, 1986).These processes simultaneously provide anopportunity to disconfirm negative stereotypesabout the out-group and break down the mono-lithic perception of them as a homogeneousunit (Cook, 1978; Wilder, 1978). Consequently,personalized contact can augment the benefi-cial effect of cooperative contact (Bettencourt,Brewer, Croak, & Miller, 1992; Miller, Brewer, &Edwards, 1985).

In accord, Brewer and Miller’s Decategoriza-tion Model argues that out-group contact willmost effectively reduce prejudice when inter-actions are person-based rather than category-based (Brewer & Miller, 1984; Miller &Harrington, 1992). However, an alternativeperspective on contact, the Mutual IntergroupDifferentiation Model, asserts that ‘favorablecontact with an out-group member must be

defined as an intergroup encounter in orderfor out-group evaluations to be successfullychanged’ (Brown & Turner, 1981; Hewstone &Brown, 1986, p.18). In this view, social categorycues must remain salient and individuals in thecontact setting must see each other as repre-sentative of their group in order for positivecontact effects to extend to other members ofthat social category (Brown & Turner, 1981).Hewstone and Brown (1986) further argued,however, that personalized interaction duringintergroup contact gives the out-groupmember an individual identity. This creates aperception of atypicality. That out-groupmember is no longer seen as representing thecategory. Consequently, in their view, a goodexperience with that out-group person cannotimprove attitudes toward other members ofthat out-group.

Recently, we integrated these two seeminglyopposing models. We proposed that a greaterreduction in out-group prejudice is achievedunder the interactive effects of both personal-ization and high out-group salience (Ensari &Miller, 2002). In confirmation, reduced preju-dice toward new out-group members wasgreater when personalization experimentallyco-occurred with high typicality or salience ofgroup membership, by comparison with theirindependent effects.

Here we argue that personalized contact hasimportant implications for intergroup attribu-tions, as well as prejudice. Consequently, weextend the Ensari and Miller model to examineattributions, as well as prejudice. As indicated,a personalized interaction with an out-groupmember may elicit a subtyping process(Hewstone, 2000). Such subtyping may extendto attributional processes. That is, rather thanaltering attributional stereotypes about theout-group as a whole, personalized contact mayinduce causal attributions based on the uniqueattributes of that particular out-group person. Ifthat out-group person, however, is highly typicalof her group, it is more difficult to explain hercounterstereotypical behavior in terms ofsituational demands or individual exceptions tothe rule.

Group Processes & Intergroup Relations 8(4)

392


Intergroup attributional bias

The ultimate attribution error reflects theexpression of ethnocentric bias in the attribu-tions made to explain outcomes obtained byin-group and out-group members (Pettigrew,1979). Negative in-group and positive out-group outcomes are attributed to situationalfactors, whereas positive in-group and negativeout-group outcomes are attributed to causesseen as internal. These dynamics protect theesteem of the in-group and cast the out-groupin a bad light (Hewstone & Jaspars, 1984;Pettigrew, 1979) thereby augmenting a positiveself-identity and insulating negatively stereo-typed views of the out-group (Hewstone, 1990;Pettigrew, 1979).

Two interesting cases: Out-group success andin-group failureSome behaviors, such as a successful act by anout-group member or an unsuccessful act byan in-group member, are inconsistent withexpectations. Conceptually, these behaviors arenot only more interesting, but they involvemore complex cognitive and motivational pro-cesses. Therefore, in exploring the attributionalconsequences of manipulating performancefeedback, we examined only the reactions tothese two types of outcomes (i.e. out-groupsuccess and in-group failure) rather than usinga more complete factorial design.

The psychological processes involved in theperception of out-group success and in-groupfailure differ. Out-group success not only sur-prises people, but it can disconfirm negativestereotypes of the out-group. The impact ofout-group behaviors that violate such negativestereotypes can be diluted, however, by attribut-ing them to situational influences (such astemporary high motivation, luck) rather thanto dispositional attributes (Hewstone, 1989;Pettigrew, 1979). When an out-group is success-ful, we predicted an interaction. That is, underconditions wherein the out-group generally wassuccessful, we expected those liberals whoexperienced a more personalized contact witha typical conservative to exhibit less prejudicetoward other conservatives, and make higher

ability attributions for their success, thanthose in the low personalization condition. Bycontrast, we expected attitudes and attributionstoward other liberals (in-group members) toremain uniformly positive and undifferentiatedin the high and low personalization conditions,thereby yielding our predicted interaction.

On the other hand, a negative evaluation ofone’s in-group can be perceived as a threat thatcalls for some coping response (Ellemers,Spears, & Doosje, 2002). One way to protectone’s individual self from such threat is todistance one’s self from the in-group (e.g.exhibit higher perceptions of in-group hetero-geneity, Doosje, Spears, Ellemers, & Koomen,1999, less in-group favoritism, Ellemers, VanRijswijk, Roefs, & Simons, 1997, and less in-group identification, Ellemers, 1993) and asso-ciate instead with the out-group (Mullen,Brown, & Smith, 1992). Consequently, membersof a group that receives in-group failurefeedback will typically constrain expression ofdifferential prejudice toward in-group and out-group members. Accordingly, we expected nodifferential prejudice on either evaluative orattributional measures in this condition. Wethought this lack of differentiation to be par-ticularly likely because the within-subjectsportion of our design allowed participants tocomparatively evaluate in-group and out-groupjob applicants. Under these conditions, weexpected the protective motivation generatedby strong general evidence of in-group failureto wipe out any potential benefit from person-alization.

Mediators of prejudice

Causal attribution has long been thought toguide decision and action (e.g. Kelley, 1973).Most research has focused on the first connec-tion in the mediational chain that sees attribu-tions as the causal link between antecedentstimuli and subsequent action. Thus, numerousstudies examine attributions as dependentmeasures. Further supporting this mediationalchain, some studies manipulate attributions andshow predicted effects on affective reactions(Weiner, Russell, & Lerman, 1979; Russell, &

Ensari & Miller prejudice and intergroup attributions

393


McAuley, 1986). A few studies correlationallyhave examined both links within a single exper-iment (e.g. Islam & Hewstone, 1993; Town &Harvey, 1981; Yarkin, Harvey, & Bloxom, 1981).Nevertheless, despite the fact that attributionshave frequently been invoked in explanationsof prejudice and bias (Hewstone, 1989) wecould not find a single instance of such media-tional evidence.

Although many attributional factors mayplay a role in mediating prejudice, we focusspecifically on two potential consequences ofpersonalization: attributions of friendliness andattributions of ability. Substantial evidencesupports the view that interpersonal percep-tions of friendliness are conceptually indepen-dent from perceptions of ability. The five-factordescriptive model of personality (Costa &McCrae, 1992) is now widely accepted (McCrae& John, 1992). Important here is that theresearch supporting this structural organizationis not merely based on self-reports on personal-ity scales. Much of it rests on factor analyses ofthe meaning of words, or the language of per-sonality descriptors (Goldberg, 1990; Norman,1967). Another very substantial supportingcomponent emerges from interpersonal per-ceptions or judges’ ratings of others (Botwin &Buss, 1989). Viewed most broadly, then, thefive-factor model provides an organization ofhuman experience—perceptions of inter- andintrapersonal attributes. Two of its factors(Extraversion and Agreeableness) are moreclosely related to interpersonal friendliness,whereas another pair (Conscientiousness andIntellect) is more closely associated with per-ceptions of ability. Correspondingly, based onAsch’s (1946) research on person perception,Fiske, Xu, Cuddy, and Glick (1999) argue thatwarmth (sociable and agreeable traits) andcompetence (task performance and intellectualtraits) are the two independent dimensions thatunderlie many outgroup stereotypes.

Friendliness as a mediatorOne important aspect of personalized contactis its link to friendliness—perceiving the out-group member with whom one has had per-sonalized contact as friendly. Contact at the

personal level augments perceived friendlinessat the intergroup level (Cook, 1984; Wilder,1984). Receipt of intimate information, whichis usually shared only with friends, is rewardingand valued more (Lynn, 1978; Petty & Mirels,1981). It implies being trusted by the discloser,thereby creating a positive attitude toward her(Altman & Haythorn, 1965; Fitzgerald, 1963).

Such friendship with an out-group memberreduces prejudice, and creates more positiveattitudes toward other out-group members(Pettigrew, 1997). Work on the extendedcontact hypothesis further elaborates theimportance of friendliness, proposing thatmere ‘knowledge that an in-group member(other than self) has a close relationship withan out-group member can lead to morepositive intergroup attitudes’ (Wright, Aron,McLaughlin-Volpe, & Ropp, 1997, p. 74). Con-sequently, we propose that personalized contactwith an out-group member not only yields sub-jective experiences of uniqueness and intimacy,but also, induces perceptions of friendliness.Taken together, we expect these subjective statesto promote favorable attitudes toward other out-group members. In sum, then, we predicted thatin the high personalization condition partici-pants would perceive the self-information dis-closed by the out-group confederate as indeedmore personal and unique, and simultaneouslyperceive her as friendlier than those in the lowpersonalization condition. In turn, we expectedthe combination of these effects to mediatereduced prejudice toward new members of theout-group category.

Internal ability attributions as a mediator The internal attribution of ability as a mediatorof prejudice is important because its invocationas an explanation for out-group success can beinterpreted as a sign of reduced intergroupprejudice (Pettigrew, 1979). One of the effectsof personalization is that it invokes better pro-cessing of individuating information aboutpersons irrespective of their social category(Rothbart & John, 1985; Wilder, 1986), therebyincreasing the potential impact of informationabout out-group success. Personalization canalso decrease the accessibility of stereotypes


394


that interfere with internal ability attributionsfor successful out-group performance (Galinsky& Moskowitz, 2000). Consequently, weexpected that subjective perceptions of abilitywould combine with subjective personalizationeffects to mediate reduced prejudice towardnew out-group members.

Overview of the study

This experiment is the first to explore theattributional consequences of personalizedinteraction. Within a single design, it examinesthe impacts of (a) personalization and (b)information about the general success orfailure of out-group or in-group members on aproblem-solving task that the participant couldnot solve without the out-group confederate’shelp. It assesses their effects on both prejudicetoward and, more important, attributions madeto explain the success of new out-group andin-group members in having obtained a uni-versity job.

During an interview task, a confederate posedas an out-group member (a conservative). Inthe high personalization condition, she volun-tarily self-disclosed personal information thatwas absent in the low personalization condition.To preclude the possibility that she could beseen as an exception and thereby undercut thebearing of our manipulations on attitudes andattributions toward new out-group members,we ensured that she was seen as typical of hergroup while keeping her category membershipsalient. Then, after participants had worked ona problem-solving task that required the out-group confederate’s help, they learned eitherthat the out-group (conservatives) had gener-ally succeeded at that task or that their in-group(liberals) had generally failed. Key dependentmeasures were: (a) participants’ evaluationsboth of new individual in-group and out-groupjob applicants who had successfully obtaineduniversity employment, but more importanthere, (b) the attributions for having successfullyobtained that job. By giving success/failurefeedback about the generally positive perform-ance of the out-group or the generally poor per-formance of the in-group rather than individual

out-group or in-group performance, we reducedthe relevance of subtyping as an explanationfor the unexpectedness of these respectiveoutcomes.

Method

Participants and designForty-three politically liberal undergraduates,36 females and 5 males, in the Introductory Psy-chology course at the University of SouthernCalifornia participated in partial fulfillment oftheir course requirement. Two were laterremoved from analyses due to suspicion.

The design was a 2 (high vs. low personaliza-tion) � 2 (performance feedback: in-groupfailure vs. out-group success) � 2 (group mem-bership of the job applicants: in-group vs. out-group) mixed design with the first two factorsmanipulated between, and the third within-subjects.

Procedure After the participant had arrived, a female con-federate signed in as a second participant. Thefemale experimenter seated them oppositeeach other at a table and asked them tocomplete a brief background questionnaire inwhich they self-categorized themselves alongthe dimensions of political orientation (liberalvs. conservative), age, gender, major, year inschool, and socioeconomic status. Only thosewho identified themselves as liberals wereallowed to continue. Conservatives or indepen-dents were debriefed and released.

We imposed several contextual features in allconditions. First, the confederate always cate-gorized herself as a conservative. Second, theexperimenter always left the confederate’s com-pleted questionnaire on top in a readableposition to ensure throughout that the partici-pant could easily see that the confederate con-sidered herself to be a conservative. Third,participants were told that they would engagein three separate studies that respectively con-sisted of a communication task, a cooperativeproblem-solving task, and an evaluation task.In the communication task (the interview),though their roles were ostensibly random, all


395


participants were made the interviewer; theconfederate was always the interviewee. Fourth,after their role assignment but prior to the startof the communication task (wherein personal-ization was manipulated), the experimenterasked the participant and the confederate tointroduce themselves to each other, suggestingthat the confederate go first. The confederatealways established herself as a typical conserva-tive by stating the pretested characteristics usedin Ensari and Miller (2002) (e.g. ‘I am a con-servative person’, ‘My father is a lawyer’, ‘I workfor the Student Community of Young Republi-cans in my spare time’). After the confederate’sintroduction, the participant then providedinformation about herself. Note that whateverthe variation in the information given by theparticipants, it could not co-vary systematicallywith the experimental conditions because theyhad not yet been introduced.

In addition to establishing that the confeder-ate was a typical conservative, we also sought tomaintain the salience of her out-group categorymembership as another constant feature of theexperiment. In all conditions we then told theparticipants that we were especially interestedin their political orientation and therefore, inorder to easily identify them, they were asked towear name badges throughout the experiment.The 3 in. � 4 in. badges contained the wordliberal (in blue) for the participant, and conserv-ative (in red) for the confederate. In addition,in the interview task (described below) that wasused to implement the personalization con-ditions, the confederate always made theStudent Association of Young Republicanssalient. In the high personalization condition,she reminded the participant of her politicalidentity by disclosing that one of her hobbieswas part-time work in the Student Associationof Young Republicans. In the low personaliza-tion condition she made this category salient bynaming it in response to the interview questionabout prominent student organizations oncampus.

Manipulation of personalization After the par-ticipant and confederate had introduced them-selves and had exchanged the individuating

information described above, we reiterated thatthe research was concerned with how peoplefrom different perspectives work together on acommunication task. The experimenter thengave the real participant a list of five interviewquestions, the last of which had four subparts.She was instructed to use them as an interview-ing guide and to engage in as natural conver-sation as possible for 10 min. As taken fromEnsari and Miller (2002), this form includedeither five personal (high personalization con-dition), or five impersonal questions (low per-sonalization condition). The distinct interviewschedules for personalization conditions clearlyimposed differential constraint on the levels ofself-disclosure provided by the confederate. Tofurther augment the comparative impact of thepersonalization conditions, however, beforestarting the interview the experimenter publiclytold the confederate, ‘anything you wish to tellabout yourself is okay—it is completely up you’.We thereby encouraged the participant to viewthe act of disclosure (or lack of it) as voluntary(despite the differential constraints implicit inthe interview questions) and hence, feelentrusted with personal information in thehigh personalization condition. The experi-menter then left the room.

In the high personalization condition, theconfederate disclosed personal, and uniqueinformation about herself. Specifically, inresponse to the last question, requesting adescription of two good and two bad thingsabout herself, she stated: 1. I work part-time ina company so that I can give my savings to myyounger brother because he doesn’t haveenough money to go to school; 2. The happiestnews I had this year was that my father decidedto become engaged because after my mother’sdeath he wasn’t happy with his life and feltlonely all the time; 3. The most embarrassingmoment in my life was when a professor caughtme cheating during an exam; 4. I have madeonly a few good friends during the past fewyears, and therefore, I feel lonely most of thetime. These four pieces of information werebalanced for level of intimacy and valence. Aspreviously reported (Ensari & Miller, 2002)both the comparative extremity and the


396


intimacy of the positively and negativelyvalenced responses did not reliably differ intheir magnitude of deviation from the scalemidpoints.

In the low personalization condition, theconfederate responded to the interview ques-tions with neutral impersonal statements thatincluded no additional unique, individuatinginformation beyond that which had previouslybeen provided in all conditions as part of theparticipant’s and confederate’s respective intro-ductions of self to one another. Specifically,the questions and responses only concernedgeneral aspects of campus life.

Manipulation check of personalization Afterthe communication task, we checked the per-sonalization manipulation. Participants indi-cated the extent to which the informationrevealed by the confederate was personallyunique, personal, and unexpected. They alsorated her friendliness, using 7-point scales forall items. Higher scores indicated greater per-sonalization and friendliness. We used thesescores in the mediational analyses.

Manipulation of out-group success vs. in-groupfailure The cooperative problem-solving taskwas introduced as allegedly assessing the con-ditions under which cooperation produces ben-eficial outcomes. Actually, it provided a basis forthe manipulation of out-group success versusin-group failure. The participant and confeder-ate were told that they would work coopera-tively and train each other to solve two cognitiveability problems. The participant received thefirst problem, along with instructions thatexplained the solution and asked her to teachthe confederate how to solve it. Next, the con-federate taught the participant how to solve thesecond problem. To ensure that the participantand the confederate were mutually dependenton each other and worked cooperatively, bothproblems were chosen to be too difficult tosolve without help from the other. After com-pleting the cooperation task, participants indi-cated on 7-point scales how well they hadworked together and the degree to which theyhad worked cooperatively.

To manipulate performance feedback theparticipant and confederate were given asummary of the results obtained by previousparticipants who allegedly had worked onanother version of the problem-solving taskscompleted by the participant and confederate.They were asked to review it and to complete anevaluation form that would be helpful in inter-preting research results from these experi-ments. The questions in this evaluation form, asdescribed in the next section, were used notonly as a manipulation check of feedback, butalso as attributional measures for subsequentanalyses of the mediating role of attributions inreducing prejudice.

The two conditions of the performancefeedback were out-group success and in-groupfailure. The performance feedback manipu-lation in this experiment only included group-based feedback. In the out-group successcondition, the result summary page includedtwo pieces of information: the first piece ofinformation was that these previous partici-pants were conservatives (i.e. out-groupmembers). The second was that the averageperformance score of these conservative par-ticipants was 80%, which indicated that theyhad in general succeeded on a highly similartask. In this condition, this summary page didnot include any information with respect to in-group performance.

In the in-group failure condition, on theother hand, this result summary which alsoincluded two pieces of information indicatedthat these previous participants were liberals(i.e. in-group members), and that their averageperformance score was 20% which indicatedthat these in-group participants had in generalfailed. Again, this page did not give any infor-mation about out-group performance.

Manipulation check on performance feedbackWe used a conceptual check on the feedbackmanipulation, assessing whether it had elicitedthe attributional processes implicit in theultimate attribution error. Using 7-point scaleswith higher scores indicating more of theattribute, participants indicated the degree towhich the group’s performance was a result of


397


(a) their ability; (b) their effort; (c) luck; (d) aconsequence of the task. Separate from the pre-viously described measure of the confederate’suniqueness and friendliness, three other itemsevaluated the group’s friendliness (‘howfriendly are these participants?’), trust (‘to whatdegree are they trustworthy?’), and similarity(‘how similar are you to these participants?’)We told the participants that because they didnot know these people their best guess wasappropriate.

Dependent measures The participant wastold that the purpose of the third and finalstudy was to examine how non-faculty membersevaluate the outcome of freshman applicantswho had applied for university employment(Jackson, Sullivan, & Hodge, 1993). We gavethe participant job applications of a liberal (in-group) and a conservative (out-group) appli-cant (both female), both of whom had beenaccepted for employment. We told her toexamine the materials carefully before evaluat-ing them. The political orientation of theapplicants appeared under the skills andextracurricular activities section of the form. Tomanipulate in-group/ out-group membershipof the applicants, one was depicted as havingserved five years at the Student Liberals Com-munity whereas the other had similarly volun-teered for the Organization of YoungRepublicans Conference. We constrained allother information to be sparse so as to providelittle basis for subtyping the individual appli-cants. We counterbalanced the order of thein-group and out-group application forms andthe descriptions of all other information aboutthe applicants. This latter constraint meantthat (between-subjects) there were actually twoin-group and two out-group applicants. Asdependent measures, participants first sepa-rately indicated the importance of ability,effort, luck, and task characteristics (attribu-tion measures) in explaining the applicant’semployment success. Then, as a prejudicemeasure, they used 7-point scales (with higherscores indicating more of the attribute) to ratethe degree to which each applicant was quali-fied, friendly, and trusting.

Results

Check of contextual levels of typicality,salience, and cooperationAlthough typicality and salience were intention-ally made constant across the conditions, wesought to induce high levels of each. After thecommunication task, participants responded tothree items that assessed the degree to whichthey saw the confederate as a typical conserva-tive: typical; representative of her politicalgroup; behaved like a typical member of hergroup on 7-point scales (1 = ‘not at all’, 7 =‘extremely’). When averaged (alpha = .72), themean typicality (M = 4.97) exceeded the scalemidpoint (4.0) (t(40) = 6.55, p < .0001), indi-cating that the out-group confederate wasindeed seen as a typical conservative.

To assess the level of social category salience,participants indicated at the end of the experi-ment the degree to which their political orien-tation had influenced their evaluations of theconfederate on a 7-point scale (1 = ‘not at all’,7 = ‘extremely’) (Lalonde & Silverman, 1994).If participants directly acknowledged that theirown political affiliation had influenced theirevaluation of the confederate, it implies asalient awareness of the confederate’s allegedpolitical position. Confirming the salience ofpolitical orientation, participants’ indication ofthe degree to which their own political affilia-tion had intruded into their evaluations (M =5.48) exceeded the scale midpoint (4.0) (t(26)= 5.41, p < .000).

After the cooperative problem-solving task,the participants indicated on 7-point scales thatthey had worked well and cooperatively on it.When averaged (r = .57), the mean of 5.73exceeded the scale midpoint (4.0) (t(38) =7.88, p < .000). Analyses of the combined typi-cality, salience, and cooperative problem-solving measures in a 2 (personalization) � 2(performance feedback) analysis of variance(ANOVA) yielded neither main effects nor aninteraction (ps > .05).

Manipulation checksPersonalization Participants indicated howpersonally unique, intimate, and unexpected


398


the information revealed by the confederatewas, using 7-point scales ranging from ‘not atall’ (1) to ‘very’ (7). We combined the threemeasures to create a subjective personalizationmeasure (alpha = .91). A 2 (personalization) �2 (performance feedback) ANOVA revealedonly a main effect of personalization (F(1, 37)= 90.59, p < .000, �2 = .71). As expected, therevealed information was seen as more person-alized (M = 5.85) in the high personalizationcase than in the low personalization condition(M = 2.05).

Participants evaluated the degree to whichthey would like to be friends with the confed-erate. A (2 � 2) ANOVA revealed a trendtoward a main effect (F(1, 37) = 3.07, p = .07,�2 = .08), suggesting that the confederate wasviewed as more desirable as a friend in the high(M = 5.05) as compared with the low personal-ization condition (M = 4.42).

Out-group success and in-group failure Toform a composite external attribution measure,luck and task attributions were combined (r =.36, p < .05). The other attribution measureswere not correlated, and therefore analyzedseparately.

Three parallel 2 (personalization) � 2 (per-formance feedback) ANOVAs revealed onlymain effects of performance feedback: forability (F(1, 37) = 12.20, p < .01, �2 = .25),for effort (F(1, 37) = 9.80, p < .01, �2 = .21), andfor composite external attributions (F(1, 37) =7.38, p < .01, �2 = .17). Participants in theout-group success condition more stronglyattributed that performance to ability andeffort (M = 5.25 and M = 5.45, respectively)than did those attempting to explain unsuc-cessful in-group performance (M = 4.19, M =4.38, respectively). Further, the latter groupmore strongly invoked external factors toaccount for their in-group’s failure (M = 4.62)than did those attempting to explain out-groupsuccess (M = 3.70). These comparative effectsconfirm a successful manipulation of out-groupsuccess versus in-group failure. Participantsmade out-group favoring attributions—byattributing out-group success to ability, andin-group favoring attributions—by attributing

in-group failure to luck. By contrast, and asanticipated, the group evaluative measures ofthe friendliness, trust, and similarity of the in-group (in the in-group failure condition) or theout-group (in the out-group success condition)yielded no effects (p > .05). This discriminativevalidity makes sense in that these latter measureswere unrelated to the group performancefeedback and are not further discussed.

Prejudice toward new out-group and in-group membersThe degree to which each applicant was quali-fied, friendly, and trusting were combined toform composite scores for the in-group (alpha= .72) and the out-group applicants (alpha =.74). A 2 (personalization) � 2 (feedback) � 2(in-group vs. out-group applicant) ANOVA withrepeated measures on the last factor yielded athree-way interaction (F(1, 36) = 5.66, p < .05,�2 = .14), and a two-way interaction between theapplicant’s group membership and personaliza-tion (F(1, 36) = 4.36, p < .05, �2 = .11). To fullyunderstand these interactions, we separatelyanalyzed the out-group success and in-groupfailure feedback conditions.

Out-group success feedback condition The 2(personalization) � 2 (in-group vs. out-groupapplicant) ANOVA yielded an interaction (F(1,18) = 6.52, p < .05, �2 = .27), and a main effectof personalization (F(1, 36) = 4.68, p < .05, �2 =.21) (see Figure 1, upper panel). Supportingour predictions and the Personalization Model(Brewer & Miller, 1984; Ensari & Miller, 2002),participants who examined the results summaryof successful out-group members exhibited lessprejudice toward new out-group applicantsunder high (M = 5.23) versus low personalization(M = 4.10) (F(1, 18) = 16.09, p < .01, �2 = .47).Evaluations of in-group applicants, however,were equally favorable in both the high (M =4.88) and the low personalization conditions(M = 4.90; F(1, 18) = .95, p > .05, �2 = .00). Thisresult is consistent with our predictions and withthe results of Ensari and Miller (2002), showingthat under out-group success the participantsevaluated the in-group applicants similarly irre-spective of the personalization manipulation.


399



400

Figure 1. Mean prejudice toward in-group and out-group applicants in the out-group success feedbackcondition as a function of personalization (upper panel). Mean prejudice toward in-group and out-groupapplicants in the in-group failure feedback condition as a function of personalization (lower panel). Higherscores indicate less prejudice.

Out-group success feedback

In-group failure feedback

in-group applicant out-group applicant

pre

judic

e to

war

d i

n-g

roup a

nd o

ut-

gro

up a

ppli

cants

personalization no personalization

pre

judic

e to

war

d i

n-g

roup a

nd o

ut-

gro

up a

ppli

cants




Thus, the personalization main effect is attrib-utable to the out-group applicants. Finally,there was no main effect of applicant (F(1,18)= 1.00, p > .05; �2 = .05; M = 4.66 for the in-group applicant, and M = 4.89 for the out-group applicant).

In-group failure feedback condition The 2(personalization) � 2 (in-group vs. out-groupapplicant) ANOVA yielded neither an inter-action nor main effects in the in-group failurefeedback condition (p > .05). As expected,under the in-group failure condition partici-pants exhibited the same level of favorability intheir evaluations of the in-group and out-groupapplicants (see Figure 1, lower panel). Thosewho had received in-group failure feedbackfailed to exhibit differential prejudice towardout-group and in-group applicants.

Attributional bias toward new in-group andout-group membersWe examined participants’ attributions for thesuccess of the new in-group and out-groupapplicants in obtaining jobs. Luck and taskattributions were combined to create a com-posite external attribution measure (r s = .63and .73 respectively for the in-group and out-group applicants). Ability and effort attribu-tions were uncorrelated, and thus, analyzedseparately.

We expected ability attributions to exhibitthe same pattern as the prejudice measures.The (2 � 2 � 2) ANOVA of attributions ofability revealed a three-way interaction (F(1, 36)= 5.61, p < .05, �2 = .14), and a main effect ofpersonalization (F(1, 36) = 5.45, p < .05, �2 =.13). Again, we analyzed responses in the out-group success and in-group failure feedbackconditions separately.

Out-group success feedback condition Despitethe absence of a two-way interaction betweenpersonalization and the applicants’ groupmembership (F(1, 18) = 2.37, p =.14, �2 = .12),we performed additional analyses on the abilityattributions because theory suggested a specifica priori hypothesis (Rosenthal & Rosnow,1985). Participants who had previously received

feedback indicating out-group success morestrongly attributed the out-group applicants’success to ability in the high (M = 5.60) than inthe low personalization condition (M = 4.30)(F(1, 18) = 14.49, p < .01, �2 = .45) (see Figure2, upper panel). That is, when out-groupmembers had previously succeeded on a taskthat the participant could not perform cor-rectly without the out-group confederate’s aid,personalization increased the tendency to givecredit for the out-group applicants’ success inattaining a job (i.e. greater ability attributions).By contrast, in the ability attributions made forthe in-group applicants, this difference disap-peared (M = 5.60 and M = 5.30; F(1, 18) = .40,p > .05, �2 = .02). Participants made similarlyfavorable attributions for the job success of newin-group applicants. These results parallel thosefor the prejudice measure.

In-group failure feedback condition A 2 � 2ANOVA of the ability attributions made in thein-group failure feedback condition yielded noreliable effects (p > .05, see Figure 2, lowerpanel). When the in-group had failed, partici-pants gave credit (i.e. high ability attributions)for both out-group and in-group applicants’ jobsuccess.

Other attributional measures The analyses ofthe external attribution and effort measures didnot reveal any effects. We consider these nulleffects in the Discussion section.

Mediational analysesEarlier we noted that perceptions of friendli-ness or sociability are conceptually distinct fromperceptions of ability. We expected that per-sonalization would elicit both types of subjec-tive responses. In line with their conceptualindependence, however, we anticipated distinctmediational roles for subjective perceptions offriendliness and ability. Therefore, to accountfor the prejudice-reducing effect of personal-ization we examined the mediational roles ofsubjective personalization when separatelycombined with (a) perceived friendliness of theconfederate and (b) with ability attributions forthe out-group’s success. Note that our concern


401



402

Figure 2. Mean ability attributions made for in-group and out-group applicant’s job success in the out-groupsuccess feedback condition as a function of personalization (upper panel). Mean ability attributions made forin-group and out-group applicant’s job success in the in-group failure feedback condition as a function ofpersonalization (lower panel). Higher scores indicate less attributional bias.



abil

ity a

ttri

buti

ons

for

the

appli

cant’

s su

cces

sab

ilit

y a

ttri

buti

ons

for

the

appli

cant’

s su

cces

sOut-group success feedback

In-group failure feedback




here is with the evaluations of new out-groupmembers—the applicant files.

Mediational role of personal information andfriendliness The friendliness measure was thedegree to which the out-group confederate wasperceived as friendly after the communicationtask. Then, the subjective personalization andfriendliness measures were transformed toz-scores, and combined by taking their average(alpha = .48). This combination, calledpersonal/friendliness, was treated as a mediator.

We then performed a series of mediationalanalyses of prejudice toward the out-groupapplicants by applying Judd and Kenny’s (1981)four-step regression equations. Confirming ourpreviously presented manipulation checkanalyses, the regression of the compositepersonal/friendliness measure (the mediator)on personalization (the independent variable)showed that participants perceived the dis-closed information as more personal, and theconfederate as more friendly in the high thanthe low personalization condition (F(1, 39) =36.19, p < .001, R2 = .48, standardized beta =–.69). Next, regression of the out-group evalu-ation on the independent variable confirmedour previously reported ANOVA effect ofmanipulated personalization on prejudice,indicating less prejudice against the out-groupapplicants in the high personalization conditionthan in the low personalization condition(F(1, 38) = 5.36, p < .05, R2 = .12, standardizedbeta = –.35). The third step of the mediationalanalysis showed that with the independentvariable entered simultaneously in the equation,the personal/friendly measure influenced theparticipants’ evaluations of the out-groupapplicants (F(2, 38) = 5.02, p < .05, R2 = .12,standardized beta = .34). After the insertionof the personal/friendly measure into theequation, the direct effect of the independentvariable on prejudice was no longer significant(p > .05). Sobel’s (1982) procedure for testingthe significance of the indirect mediationalrelationship showed a reliable change in thebeta (z(38) = –2.17, p < .05). Thus, theseanalyses suggest that the combined subjectiveexperience of the confederate’s personal

information and friendliness mediated theeffect of manipulated personalization on preju-dice toward new out-group members, the out-group job applicants.

Additional analyses show that when takenalone, neither subjective personalization norfriendliness by itself mediated the effect of per-sonalization on prejudice. Separately regressingprejudice on subjective personalization andon friendliness showed neither a direct linkbetween subjective personalization and preju-dice (F(1, 38) = 3.72, p > .05), nor betweenfriendliness and prejudice (F(1, 38) = 2.90,p > .05). Thus, these results further supportthe argument that it is the combination ofboth subjective personalization and friendlinessthat mediates the effect of personalization onprejudice.

Mediational role of personal information andability attributions We next examined themediational role of ability attributions inreducing prejudice. We sought to ensure thatthe subjective perceptions of ability were bothmethodologically and conceptually distinctfrom the measures of prejudice. Therefore,although the attributions that participants hadmade about the success of the job applicantsserved as one of our two key dependentmeasures (evaluations of them being the other)we did not use these attributions in our media-tional analyses. As previously argued, theselatter attribution measures can be viewed, alongwith adjective evaluative measures, as co-varyingsymptoms of prejudice. Moreover, they werespatially adjacent in the dependent measurespacket. To avoid this conceptual and methodo-logical contamination, we used instead the attri-butions that participants had made to explainthe feedback indicating general success of theout-group or failure of the in-group respectivelyin solving the types of problems that the partic-ipant could not solve without the confederate’shelp.

We expected subjective personalization(separate from its link with friendliness, but incombination with ability attributions) tomediate the link between manipulated person-alization and reduced prejudice. To examine


403


this model we first separately transformed toz-scores both the subjective personalization andthe ability attribution measures regarding theout-group’s general problem-solving success.We then combined them into two compositesand averaged them (alpha = .41). This combi-nation, called personal/ability, was treated as amediator.

Following Judd and Kenny’s (1981) media-tion analyses, regression of the personal/abilitymeasure (the mediator) on personalization(the independent variable), showed thisrelationship to be highly reliable (F(1, 38) =47.43, p < .00, R2 = .56, standardized beta =–.75). Regression of prejudice toward theout-group applicants on the personal/abilitymeasure, with the independent variable simul-taneously entered in the equation, showed thatthe personal/ability measure influenced preju-dice (F(2, 37) = 18.16, p < .00, R2 = .50, stan-dardized beta = .33). Finally, controlling for thepersonal/ability measure, the direct effect ofthe independent variable no longer remainedsignificant (p > .05). Supporting this media-tional effect, Sobel’s test of the change in betawas reliable (z(38) = –4.16, p < .001).

We also examined whether ability attributionalone mediates the effect of manipulated per-sonalization on prejudice. Regressing abilityattributions for out-group success on manipu-lated personalization showed no direct linkbetween personalization and ability attributions(F(1, 38) = .67, p > .05). Therefore, as was thecase for subjective personalization and friendli-ness, we conclude that it is the combination ofboth subjective personalization and ability attri-bution that (partially) mediates personaliza-tion’s effect on prejudice.1

Discussion

In this first study examining the attributionalconsequences of personalized interaction, weassessed the impact of (a) personalization and(b) information about the success of the out-group or failure of the in-group not only on theevaluations of new out-group and in-groupapplicants, but more important, the attribu-tions made to explain their success. Our

findings provide further confirmation of theameliorative effects of personalization on prej-udice toward new out-group members (e.g.Ensari & Miller, 2002) but more critically, theyextend prior work by examining intergroupattributions. When an out-group was successful,personalization was an effective tool forreducing out-group derogating attributions.Thus, we add further support to the view thatthe Mutual Intergroup Differentiation (Brown& Turner, 1981; Hewstone & Brown, 1986) andPersonalization (Brewer & Miller, 1984) modelscontain complementary ingredients for improv-ing intergroup relations.

In-group failureBy comparison with the differential effects seenunder out-group success, we found similarfavorability toward in-group and out-groupmembers in the in-group failure conditions. Weargue that negative feedback on in-group per-formance poses a threat, which leads in-groupmembers to reduce their prejudice towardout-group members. Whereas our results areconsistent with some previous findings (e.g.Ellemers et al., 1997; Mullen et al., 1992), theycontradict others (e.g. Fein & Spencer, 1997;Jetten, Spears, & Manstead, 1999). Forexample, Mullen et al.’s (1992) meta-analyticalreview of 42 studies assessing ethnocentrismfound a weak in-group bias effect when the in-group was judged to be of lower status. On theother hand, Fein and Spencer (1997) foundthat threat increases prejudice toward out-group members in an attempt to maintain self-image. This inconsistency can be explainedwith a number of moderating factors, such asstability, legitimacy, permeability, group identifi-cation (Mummendey, Kessler, Klink, & Mielke,1999), group status (Ellemers et al., 1997;Jetten et al., 1999), and commitment (Ellemerset al., 2002). Another important moderatingfactor is whether threat is individual-based(negative feedback to one’s self) or group-based (negative feedback to one’s in-group).Individual-based threat, as in the case of Feinand Spencer’s studies (1997), is perceived asdirected toward one’s self, and thus provokesself-image restoration and increases self-esteem.


404


Group-based threat, on the other hand, as inthe case of the present study, leads to socialmobility responses when the in-group’s lowstatus is seen as legitimate (Branscombe,Spears, Ellemers, & Doosje, 2002).

Additionally, however, one might wonder whyunder the in-group failure condition there wereno personalization effects with respect to evalu-ations or attributions about the out-group appli-cants. In the in-group failure condition theparticipants faced direct evidence of general in-group failure. This evidence was likely to arousein-group protective motivations. In turn, suchmotivation may have contributed to thetendency of those in the in-group failure con-dition to evaluate the in-group no less favorablythan the out-group. One way in which such pro-tective motivation may have produced equiva-lent evaluations of the in-group and out-groupapplicants was by leading the in-group failureparticipants to suppress the beneficial effect ofpersonalization that was seen in evaluations ofthe out-group applicants under conditions ofout-group success. Unfortunately, an assessmentof the different processes that personalizationmight induce under out-group success by com-parison with in-group failure was compromisedin this experiment by the absence of a no-feedback control condition. In retrospect, thiswas an important omission. Had we had a no-feedback condition we would have expected tofind beneficial personalization effects in theevaluations of out-group applicants.

Ability versus luck and task attributions Ouroutcomes showed that when personalized inter-action occurs with typical out-group memberswhose category identity is salient, it will have astrong impact on internal (ability) attributions.Yet external (luck and task) attributions failedto be affected under these conditions. Thereare two plausible explanations for this finding.First, people prefer dispositional attributions toexplain behaviors of the actor because suchexplanations are available and are simple (Fiske& Taylor, 1984; Tversky & Kahneman, 1973). Ittakes more effort and imagination to think ofthe situational factors that might function asexplanations (Tversky & Kahneman, 1973). By

contrast, it is easier to invoke the dispositionsthat are generally associated with the out-groupin the past (Fiske & Taylor, 1984; Tversky &Kahneman, 1973). Second, people tend toperceive dispositions rather than environ-mental causes as the explanation of groupdifferences (Jones & Davis, 1965; Tajfel, 1978).This occurs in part, because environmentalcauses are complex and diffuse (LeVine &Campell, 1972). Consequently, dispositions areseen as the main cause of hostility from theout-group (Allport, 1954; LeVine & Campell,1972). For these reasons, our key manipulatedvariables may have affected the dispositionalexplanations of the acts performed by out-group members, but not external reasons fortheir outcomes. To eliminate reliance on cog-nitively simple heuristics, and to activate the useof situational attributions, it may be necessarythat the experimental manipulations encour-age thoughtful analysis (Tetlock, 1985). That is,participants may need to be given incentives forcareful reasoning about the attributionalproblems posed to them. It may be necessary tomake them feel accountable or responsible forthe judgments they express (Tetlock, 1985) inorder to promote reliance on situational ratherthan dispositional explanations for causes ofothers’ behavior.

Effort attributions Our manipulations had noeffects on effort attributions. This may reflectthe ambiguity inherent in the interpretation ofeffort attributions, as illustrated by both meta-analytic (Ensari & Miller, 1998) and empiricalstudies (Feldman-Summers & Kiesler, 1974;Hewstone, Gale, & Purkhardt, 1990). Effortattributions can be internal, but both unstableand controllable (Weiner, 1974). Therefore,people can explain away the success of adisliked out-group member by claiming that heor she had to try very hard (Hewstone, 1989),or favor the in-group by attributing its failure totemporary lack of motivation (Taylor & Jaggi,1974). Consider, for example, attributionsmade by male and female students for a highlysuccessful male or female physician whosespeciality was either pediatrics or surgery andwho did or did not take on his or her father’s


405


practice. Male participants attributed thefemale’s success more to her greater motivationor to her having had an easier task. Femalesperceived the male physician as having had aneasier task than the female and they alsoattributed greater motivation to the femalethan the male physician (Feldman-Summers &Kiesler, 1974). From this perspective, the failureto find any effects with effort attributions mayreflect their interpretational ambiguity.

Mediational analysesMost important, perhaps, was the outcome ofour mediational analyses. We had previouslyargued that manipulated personalization doesnot merely induce subjective perceptions thatthe out-group partner had disclosed unique,personal, and unexpected attributes. It alsoinduces simultaneously two conceptuallydistinct effects: a sociability attribution (friend-liness) and a competency attribution (ability).

Although attributions made to explain anout-group’s outcomes have long been concep-tually invoked to account for (mediate) preju-dice toward that out-group, no prior work hasprovided direct evidence that supports thisprocess. In our mediational analyses weexamined the effect of subjectively experiencedpersonalization in combination with the abilityattributions that participants made to explainthe general success of the out-group in solvingproblems similar to those comprising the co-operative problem solving task. Recall thatparticipants were unable to solve these puzzleswithout aid from the confederate. As part ofthe manipulation of out-group success versusin-group failure, however, those in the out-group success condition were told thatmembers of the out-group had typically solved80% of such puzzles correctly. Our mediationalanalyses showed that the ability attributionsthat participants made to account for thisout-group success, when combined with thesubjective effects of personalization, partiallymediated the reduced prejudice towardsuccessful job applicants who were out-groupmembers. Note further that by using the attri-butional measure based on the manipulationof out-group success/in-group failure and not

the attributions specifically made to account forthe out-group job applicants’ success in obtain-ing their university job, we thereby made theattributions in our mediational analysis con-ceptually and methodologically independent ofthe prejudice measure. Noteworthy is that attri-butions alone did not mediate reduced preju-dice. Only when they were combined with thesubjective effects of personalization was preju-dice reduced. Thus, these results extend theconceptual importance of personalization inreducing prejudice.

The effectiveness of personalization in medi-ating reduced out-group prejudice rests in parton the subjective perceptions of friendlinessthat were elicited in combination with the per-sonalization-induced perception that the self-disclosed information was unique, personal,and unexpected. Thus both mediational pathsin the model received support. These resultsare consistent with other work showing out-group friendship to have an ameliorative effecton intergroup prejudice (Levin, van Laar, &Sidanius, 2002; Pettigrew, 1997; Pettigrew &Tropp, 2000). At the same time, it is alsopossible that the reliable mediational effectsfound for the composite personal/friendlinessand personal/ability variables, but not in theseparate analyses of personalization, friendli-ness, and ability attributions, merely reflect thegreater reliability obtained by adding the lattertwo measures to the personalization composite.

Conclusion

The present study adds further evidence for amodel proposed in Ensari and Miller (2002)that integrates the category-based (Hewstone &Brown, 1986; Hewstone & Lord, 1998) and thepersonalization models (Brewer & Miller,1984). In this integration, personalization, typi-cality, and salience were shown to be usefulcomponents for augmenting the ordinarybenefit of cooperative settings in reducing out-group bias and increasing the benefits of inter-group contact. The present study extends thismodel to intergroup attributions. Though wedid not experimentally manipulate typicalityand salience in the present study, the results are


406


consistent with the view that high levels of thesethree factors (i.e. personalization, typicality,and salience) are critical for reducing out-group derogating attributions.

One of the objectives of our investigation wasto go beyond the mere documentation of ben-eficial effects of personalization on bias andattributions, and to identify the mechanismsthat may help explain these effects. The media-tional model we proposed suggests that therelationship between personalization and out-group bias is due to the subjective effects ofpersonalization in combination with both sub-jective friendliness and subjective ability attri-butions. This model not only facilitates ourunderstanding of the psychological processesthat are responsible for the beneficial effects ofpersonalization, but also throws light on thelong noted, yet not fully explored relationshipbetween attributions and bias. To our knowl-edge, we provide the first mediational evidenceshowing that the subjective effects of personal-ization, in combination with subjective abilityattributions, partially explain reduced out-group prejudice. Although other mediatorsmay also warrant study, we believe that we haveisolated some of the major processes thatunderlie the prejudice-reducing role of person-alization.

Note1. In an additional set of analyses we examined a

model that reverses the causal path by viewingprejudice as mediating the perceptions ofsubjective personalization and friendliness, andthe outcome on the subjective personal/abilitymeasure. The outcomes showed no support forthese reverse causal models. Thus, the outcomesof these backward mediational analyses counterthe viability of a model that sees prejudice as themediator of subjective personalization effects andability attributions.

AcknowledgmentsThis work is based on a doctoral dissertation byNurcan Ensari under Norman Miller’s supervision.Financial support was provided by the HaynesDoctoral Fellowship. We thank the members of the

dissertation committee Steve Read, Shelley Duval,Vicki Pollock, David Schwartz, and JeffreySanchez-Burks for their helpful comments. We alsothank Fredy Aviles and Lily Praisan for theirassistance in data collection.

ReferencesAltman, I., & Haythorn, W. W. (1965). Interpersonal

exchange in isolation. Sociometry, 28, 411–426.Batson, C. D., Sager, K., Garst, E., Kang, M.,

Rubchinsky, K., & Dawson, K. (2001). Isempathy-induced helping due to self-othermerging? Journal of Personality and Social Psychology,73, 495–509.

Asch, S. E. (1946). Forming impression ofpersonality. Journal of Abnormal and SocialPsychology, 41, 1230–1240.

Bettencourt, B. A., Brewer, M. B., Croak, M. R., &Miller, N. (1992). Cooperation and the reductionof intergroup bias: The role of reward structureand social orientation. Journal of Experimental SocialPsychology, 28, 301–319.

Botwin, M. D., & Buss, D. M. (1989). The structureof act report data: Is the five-factor model ofpersonality recaptured? Journal of Personality andSocial Psychology, 56, 988–1001.

Branscombe, N. R., Spears, R., Ellemers, N., &Doosje, B. (2002). Intragroup and intergroupevaluation effects on group behavior. Personalityand Social Psychology Bulletin, 28, 744–753.

Brewer, M. B., & Miller, N. (1984). Beyond thecontact hypothesis: Theoretical perspectives ondesegregation. In N. Miller & M. B. Brewer (Eds.),Groups in contact: The psychology of desegregation(pp. 281–302). New York: Academic Press.

Brown, R., & Turner, J. C. (1981). Interpersonal andintergroup behaviour. In J. C. Turner & H. Giles(Eds.), Intergroup behaviour (pp. 33–65). Oxford,UK: Basil Blackwell.

Cook, S. W. (1978). Interpersonal and attitudinaloutcome in cooperating interracial groups. Journalof Research and Development in Education, 12,97–113.

Cook, S. W. (1984). Cooperative interaction inmultiethnic contexts. In N. Miller & M. B. Brewer(Eds.), Groups in contact: The psychology ofdesegregation (pp. 155–185). Orlando, FL:Academic Press.

Costa, P. T., & McCrae, R. R. (1992). Revised NEOPersonality Inventory (NEO-PI-R) and NEO Five-FactorInventory.

Doosje, B., Spears, R., Ellemers, N., & Koomen, W.


407


(1999). Perceived group variability in intergrouprelations: The distinctive role of social identity.European Review of Social Psychology, 10, 41–74.

Ellemers, N. (1993). Influence of socio-structuralvariables on identity enhancement strategies.European Review of Social Psychology, 4, 27–57.

Ellemers, N., Spears, R., & Doosje, B. (2002). Selfand social identity. Annual Review of Psychology, 53,161–186.

Ellemers, N., Van Rijswijk, W., Roefs, M., & Simons,C. (1997). Bias in intergroup perceptions:Balancing group identity with social reality.Personality and Social Psychology Bulletin, 23, 186–198.

Ensari, N., & Miller, N. (1998). The ‘ultimateattribution error’: A meta-analytic integration ofintergroup attributional bias. Unpublishedmanuscript.

Ensari, N., & Miller, N. (2002). Out-group must notbe so bad after all: The effects of personalization,typicality and salience on intergroup bias. Journalof Personality and Social Psychology, 83, 313–329.

Fein, S., & Spencer, S. J. (1997). Prejudice asself-image maintenance: Affirming the selfthrough derogating others. Journal of Personalityand Social Psychology, 73, 31–44.

Feldman-Summers, S., & Kiesler, S. B. (1974). Thosewho are number two try harder: The effect of sexon attributions of causality. Journal of Personalityand Social Psychology, 30, 846–855.

Fiske, S. T., & Taylor, S. E. (1984). Social cognition.New York: Random House.

Fiske, S. T., Xu, J., Cuddy, A., & Glick, P. (1999).(Dis)respecting versus (dis)liking: Status andinterdependence predict ambivalent stereotypes ofcompetence and warmth. Journal of Social Issues,55, 473–491.

Fitzgerald, M. P. (1963). Self-disclosure andexpressed self-esteem, social distance and areas ofself-revealed. Journal of Psychology, 56, 405–412.

Galinsky, A. D., & Moskowitz, G. B. (2000).Perspective-taking: Decreasing stereotypeexpression, stereotype accessibility, and in-groupfavoritism. Journal of Personality and SocialPsychology, 78, 708–724.

Goldberg, L. R. (1990). An alternative ‘descriptionof personality’: The big five factor structure.Journal of Personality and Social Psychology, 59,1216–1229.

Hewstone, M. (1989). Causal attribution: Fromcognitive processes to collective beliefs. Oxford, UK:Basil Blackwell.

Hewstone, M. (1990). The ‘ultimate attributionerror’? A review of the literature on intergroup

causal attribution. European Journal of SocialPsychology, 20, 311–335.

Hewstone, M. (2000). Contact and categorization:Social psychological interventions to changeintergroup relations. In C. Stangor (Ed.),Stereotypes and prejudice: Essential readings. Keyreadings in social psychology (pp. 394–418).Philadelphia: Psychology Press.

Hewstone, M., & Brown, R. J. (1986). Contact is notenough: An intergroup perspective on the‘contact hypothesis’. In M. Hewstone & R. J.Brown (Eds.), Contact and conflict in intergroupencounters (pp. 1–44). Oxford, UK: Basil Blackwell.

Hewstone, M., Gale, L., & Purkhardt, N. (1990).Intergroup attributions for success and failure:Group-serving bias and group-serving causalschemata. European Bulletin of Cognitive Psychology,10, 23–44.

Hewstone, M., & Jaspars, J. M. F. (1984). Socialdimensions of attribution. In H. Tajfel (Ed.), Thesocial dimension: European developments in socialpsychology (pp. 379–404). Cambridge, UK:Cambridge University Press.

Hewstone, M., & Lord, C. G. (1998). Changingintergroup cognitions and intergroup behavior:The role of typicality. In C. Sedikides, J. Schopler,& C. Insko (Eds.), Intergroup cognition and intergroupbehavior (pp. 367–392). Hillsdale, NJ: Erlbaum.

Islam, M. R., & Hewstone, M. (1993). Intergroupattributions and affective consequences in majorityand minority groups. Journal of Personality andSocial Psychology, 64, 936–950.

Jackson, L. A., Sullivan, L. A., & Hodge, C. N.(1993). Stereotype effects on attributions,predictions, and evaluations: No two socialjudgments are quite alike. Journal of Personality andSocial Psychology, 65, 69–84.

Jetten, J., Spears, R., & Manstead, A. S. R. (1999).Group distinctiveness and intergroupdiscrimination. In N. Ellemers & R. Spears(Eds.), Social identity: Context, commitment, content(pp. 107–126). Oxford, UK: Blackwell Science.

Johnson, D. W., Johnson, R. T., & Maruyama, G.(1984). Goal interdependence and interpersonalattraction in heterogeneous classrooms: A meta-analysis. In N. Miller & M. B. Brewer (Eds.),Groups in contact: The psychology of desegregation(pp. 187–212). Orlando, FL: Academic Press.

Jones, E. E., & Davis, K. E. (1965). From acts todispositions: The attribution process in personperception. In L. Berkowitz (Ed.), Advances inexperimental social psychology (Vol. 2). New York:Academic Press.


408


Judd, C. M., & Kenny, D. A. (1981). Process analysis:Estimating mediation in treatment evaluations.Evaluation Review, 5, 602–619.

Kelley, H. H. (1973). The process of causalattribution. American Psychologist, 28, 107–128.

Lalonde, R. N., & Silverman, R. A. (1994).Behavioral preferences in response to socialinjustice: The effects of group permeability andsocial identity salience. Journal of Personality andSocial Psychology, 66, 78–85.

Levin, S., van Laar, C., & Sidanius, J. (2003). Theeffects of ingroup and outgroup friendships onethnic attitudes in college: A longitudinal study.Group Processes & Intergroup Relations, 6, 76–92.

LeVine, R. A., & Campbell, D. T. (1972).Ethnocentrism: Theories of conflict, ethnic attitudes andgroup behavior. New York: Wiley.

Lynn, S. J. (1978). Three theories of self-disclosureexchange. Journal of Experimental Social Psychology,14, 466–479.

McCrae, R. R., & John, O. P. (1992). An introductionto the five-factor model and its applications.Journal of Personality, 60, 175–215.

Miller, N. (2002). Personalization and the promise of contact theory. Journal of Social Issues, 58,387–410.

Miller, N., Brewer, M. B., & Edwards, K. (1985).Cooperative interaction in desegregated settings:A laboratory analogue. Journal of Social Issues, 41,63–79.

Miller, N., & Davidson-Podgorny, G. (1987).Theoretical models of intergroup relations andthe use of cooperative teams as an intervention fordesegregated settings. In Review of personality andsocial psychology: Group processes and intergrouprelations (Vol. 9, pp. 23–39). Newbury Park, CA.:Sage.

Miller, N., & Harrington, H. J. (1992). Socialcategorization and intergroup acceptance. InR. Hertz-Lazarowitz & N. Miller (Eds.), Interactionin cooperative groups: The theoretical anatomy of grouplearning (pp. 203–227). New York: CambridgeUniversity Press.

Mullen, B., Brown, R., & Smith, C. (1992). In-groupbias as a function of salience, relevance and status:An integration. European Journal of Social Psychology,22, 103–122.

Mummendey, A., Kessler, T., Klink, A., & Mielke, R.(1999). Strategies to cope with negative socialidentity: Predictions by social identity theory andrelative deprivation theory. Journal of Personalityand Social Psychology, 76, 229–245.

Norman, W. T. (1967). 2,800 personallity traitdescriptors: Normative operating characteristics

for a university population. Res. Rept. 08310-1-T,University of Michigan, Ann Arbor, MI.

Pettigrew, T. F. (1979). The ultimate attributionerror: Extending Allport’s cognitive analysis ofprejudice. Personality and Social Psychology Bulletin,5, 461–476.

Pettigrew, T. F. (1997). Generalized intergroupcontact effects on prejudice. Personality and SocialPsychology Bulletin, 23, 173–185.

Pettigrew, T. F., & Tropp, L. R. (2000). Doesintergroup contact reduce prejudice? Recentmeta-analytic findings. In S. Oskamp (Ed.),Reducing prejudice and discrimination (pp. 93–114).Mahwah, NJ: Erlbaum.

Petty, R. E., & Mirels, H. L. (1981). Intimacy andscarcity of self-disclosure: Effects on interpersonalattraction for males and females. Personality andSocial Psychology Bulletin, 7, 493–503.

Rosenthal, R., & Rosnow, R. L. (1985). Contrastanalysis: Focused comparisons in the analysis ofvariance. New York: Cambridge University Press.

Rothbart, M., & John, O. P. (1985). Socialcategorization and behavioral episodes: Acognitive analysis of the effects of intergroupcontact. Journal of Social Issues, 41, 81–104.

Russell, D., & McAuley, E. (1986). Causalattributions, causal dimensions, and affectivereactions to success and failure. Journal ofPersonality and Social Psychology, 50, 1174–1185.

Sobel, M. (1982). Asympototic confidence intervalsfor indirect effects in structural equations models.In S. Leinhart (Ed.), Sociological methodology(pp. 290–312). San Fransisco: Jossey-Bass.

Stephan, W. G., & Stephan, C. W. (1985). Intergroupanxiety. Journal of Social Issues, 41, 157–175.

Tajfel, H. (1978). Differentiation between social groups:Studies in the social psychology of intergroup relations.New York: Academic Press.

Taylor, D. M., & Jaggi, V. (1974). Ethnocentrism andcausal attribution in a South Indian context.Journal of Cross-Cultural Psychology, 5, 162–171.

Tetlock, P. E. (1985). Accountability: A social checkon the fundamental attribution error. SocialPsychology Quarterly, 48, 227–236.

Town, J. P., & Harvey, J. H. (1981). Self-disclosure,attribution, and social interaction. Social PsychologyQuarterly, 44, 291–300.

Tversky, A., & Kahneman, D. (1973). Availability: Aheuristic for judging frequency and probability.Cognitive Psychology, 5, 207–232.

Weiner, B. (1974). Achievement motivation andattribution theory. Morristown, NJ: GeneralLearning Press.

Weiner, B., Russell, D., & Lerman, D. (1979). The


409


cognition–emotion process in achievement-relatedcontexts. Journal of Personality and Social Psychology,37, 1211–1220.

Wilder, D. A. (1978). Perceiving persons as a group:Effects on attribution of causality and beliefs.Social Psychology, 41, 13–23.

Wilder, D. A. (1984). Predictions of beliefhomogeneity and similarity following socialcategorization. British Journal of Social Psychology,23, 323–333.

Wilder, D. A. (1986). Cognitive factors affecting thesuccess of intergroup contact. In S. Worchel &W. Austin (Eds.), Psychology of intergroup relations(pp. 49–66). Chicago: Nelson-Hall.

Wright, S. C., Aron, A., McLaughlin-Volpe, T., &Ropp, S. A. (1997). The extended contact effect:Knowledge of cross-group friendships andprejudice. Journal of Personality and Social Psychology,73, 73–90.

Yarkin, K. L., Harvey, J. H., & Bloxom, B. M. (1981).Cognitive sets, attribution, and social interaction.Journal of Personality and Social Psychology, 41,243–252.

Paper received 22 April 2004; revised version accepted 20October 2004.

Biographical notesNURCAN ENSARI is an assistant professor at the

California School of Organizational Studies,Alliant International University. She received herPhD at the University of Southern California, andcompleted postdoctoral work at the KravisLeadership Institute, Claremont McKennaCollege. Her primary research interests are inintergroup relations, leadership, and diversitymanagement.

NORMAN MILLER is the Silberberg Professor ofPsychology at the University of SouthernCalifornia. His research focuses on intergrouprelations; aggression; and social projection.


410


Prejudice and Intergroup Attributions - Archive ouverte HAL

Documents