Messick Und Mackie

7/28/2019 Messick Und Mackie

1/39

Ann. Rev. Psychol. 1989. 40:45-81Copyright1989by AnnualReviews nc. All rights reserved

INTERGROUP RELATIONSDavid M. Messick and Diane M. MackieDepartmentf Psychology, niversity f California,SantaBarbara, alifornia 3106

CONTENTSINTRODUCTION.................................................................................... 45COGNITIVE EPRESENTATIONSF SOCIALCATEGORIES........................ 46Models f Category epresentation.......................................................... 46Content f Groupepresentations............................................................ 49Superordinate ndSubordinate ategories................................................. 51THEOUT-GROUPOMOGENEITYFFECT............................................... 55In-group s Out-group epresentations..................................................... 55Motivationalactors............................................................................ 57OtherProcessing ffects ....................................................................... 58INTERGROUPIAS................................................................................ 59Theoreticalssues ................................................................................ 59Methodologicalssues ........................................................................... 63IMPROVINGNTERGROUPELATIONS.................................................... 66Intergroup ontact.............................................................................. 66OtherApproaches................................................................................ 69Institutional and egislative change...................................................... 69Conflict resolution, bargaining, ndnegotiation ...................................... 69Categorization................................................................................ 70CONCLUDINGOMMENTS..................................................................... 71INTRODUCTIONWhen eople are judged, either singly or together, on the basis of groupmemberships,ntergroupprocesses are involved. This review, which ollowsthose of Tajfel (1982) andBrewer&Kramer1985), is structured o highlightfour research oci currentlyreceiving ntense attention.

Thestudy of intergroup relations, like many ther areas of research insocial psychology, has acquireda distinctly cognitive tone. We ccentuatethis cognitive atmosphere oth to portray current thinking about intergroupprocessesand o signal ouroptimismhat the cognitive approach ill fruitful-450066-4308/89/0201-0045502.00

www.annualreviews.org/aronlineAnnual Reviews

Annu.R

ev.

Psychol.1989.4

0:45-81.

Dow

nloadedfroma

rjournals.annualreviews.org

by

DeutscheForschungsgemeinschafton09/07/06.

Forpersonaluse

only.
http://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronline


2/39

46 MESSICK & MACKIEly augment raditional approaches. The backbone f the chapter is the study ofthe way information about groups (categories) and their members s repre-sented mentally. This approach promises fresh ideas about improving rela-tions between groups and mayelucidate underlying processes. We eview theeffect of categorization on the perceptionof the variability or heterogeneityofgroup members.Also important is the effect of categorization on tendenciesto differentiate behaviorally betweenmembers f different categories, partic-ularly in-group and out-group members.We ddress the difficult but crucialissue of extracting principles from this research that can be applied to improvethe relations amonggroups. Weconclude by noting some inroads that in-tergroup theory has made in other research domains, and by listing someresearch questions that appear especially timely.COGNITIVE REPRESENTATIONS OF SOCIALCATEGORIESCognitive representations of groups are assumed to be multiple-element(typically a category label, attributes, and/or exemplars) tructures with bothhorizontal links, connecting related concepts at a similar level of generality,and vertical ones, reflecting the hierarchical progression from more o lessinclusive categories. The encoding and retrieval of these representations isassumed to underlie judgments about groups and group members, which inturn guide intergroup behavior. In this section we review recent modelsof theinformation stored in category representations, progress in uncovering heircognitive and affective content, and the implications for intergroup perceptionof classification at subordinate compared o superordinate categories.Models of Category RepresentationCategory representations have traditionally been conceptualized as a categorylabel linked to an abstracted prototypeor list of the features assumedo be trueof the group as a whole the group stereotype). Suchgroup-level informationis assumed o derive from social learning or from multiple experiences withindividual category members,about whomnformation is not stored (Posner& Keele 1968; Reed 1972).Prototype models of social stereotypes have been widely used in socialpsychology, and the storage of knowledge bout a group as central-tendencyinformation has been assumed o play a major role in stereotyped judgments.However, people can make estimates about how variable a group is, soprototype models must be complementedor supplanted by models that in-clude variability information (E. E. Smith & Medin1981; Posner & Keele1968). Mixedmodels (Estes 1986; Hayes-Roth & Hayes-Roth 1977; FriedHolyoak 1984; Flannagan et al 1986), in which both central-tendency and


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma


byDeutscheForschungsgemeinsch

afton09/07/06.

Forpersonaluseonly.


3/39

INTERGROUP ELATIONS 47frequency information are stored, and pure exemplar models (Hintzman1986;Medin& Schaffer 1978; Elio & Anderson 1981), in which only informationabout individual category memberss stored, have recently been proposed asbetter representations of social categories (Linville et al 1986, 1988; ParkHastie 1987; Rothbart & John 1985; Smith & Zarate 1988).Linvilles (Linville et al 1986, 1988) multiple-exemplar model includesboth specific instances of the category and abstracted subtypes. In contrast tothe use of prototypes, in whicha single abstracted set of features is storedabout a category, this approach suggests that individual exemplarsas well asabstracted subtypes might be stored. Judgmentsabout the group as a whole(including variability judgments) are made by retrieving and integratinginformation about exemplars. Rothbart (1988; Rothbart & John 1985) alsoproposed that judgmentsabout a groups attributes maybe madeby integrat-ing those "episodes" from memorymost strongly associated with the categorylabel (Wilder & Shapiro 1984; Lord et al 1984).Although pure exemplar models have proven useful for studying groupvariability, mixedmodels seem more appropriate for several reasons (Lingleet al 1984; Smith & Zarate 1988). First, strict prototype and strict exemplarmodelsboth predict a substantial relation between udgmentsof the group as awhole and judgments of individual group members.However,Park & Hastie(1987) found judgmentsabout the variability of individual (out)groupbers to be unrelated to judgments about the group as a whole (AllisonMessick 1985; Judd & Park 1988; although there are problems in comparingjudgments t the different levels).Second, whereas variability judgmentspose problemsfor strict prototypemodels, strong versions of exemplarmodelsare inconsistent with the possibil-ity that at least some udgments bout groups are made nd stored as informa-tion is received (on-line), rather than on the basis of retrieval alone (ParkHastie 1987). Whenperceivers form coherent impressions of targets asinformation is received, incongruent information is likely to receive extraprocessing and thus be differentially recalled (Hastie 1980; Srull 1981). Sucheffects have been found when nformation about cohesive, close-knit groups(such as political caucusesand fraternities) is received, but not when nforma-tion about loosely related groups or aggregates is processed (Srull et al 1985;Wyeret al 1984; see also Stem et al 1984).These findings suggest that on-line processing of coherent group-levelimpressions might occur for cohesive groups perceived as a unit, but not foraggregates. Group-level judgmentsabout aggregates, in which consistency isnot expected, are therefore more likely to be memory ased and subject toretrieval biases. For example,distinctiveness-based illusory correlations [theperception of a relationship between, for example, category membership nda particular feature, whenno such relationship exists (Hamilton & Sherman


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by

DeutscheForschungsgemeinsch

afton09/07/06.

Forpersonaluse

only.


4/39

48 MESSICK & MACKIE1988)] appear to be based on the overrecall of cooccurrences of distinctivetargets and distinctive events [such as minority group members erforminginfrequent behaviors (Acorn et al 1988; Arkes & Rothbart 1985; Hamilton etal 1985; Regan & Crawley 1984)]. These findings suggest that coherentimpressions of laboratory groups labelled merely GroupA or GroupB are notformed on-line. However, llusory correlations do not develop whensubjectsare explicitly told to form an impression of the group (Pryor 1986; Sanbon-matsu et al 1987). These findings suggest that group judgments re not alwaysbased on exemplarretrieval, but can occur via on-line processing.Third, pure exemplar models cannot explain the social learning of group-level information ("big boys dont cry," "boys will be boys") in contrastlearning from direct experience. Smith & Zarate (1988, Exp. 2) showed hatlearning a group prototype before encountering individual exemplars in-creased use of prototype information in later judgments, while subjects whoreceived information in the opposite order were more ikely to use exemplars.Similarly, Park & Hastie (1987, Exp. 2) found lower estimates of groupvariability when subjects received prototype before exemplar informationthan when he same information was presented in the opposite order.Someevidence regarding the factors that determine storage and use ofexemplar compared to group-level information has appeared. In general,categorization cues, which highlight the group as a unit, appear to reduceincorporation of individual exemplar nformation into representations (Tayloret al 1978; Nesdale et al 1987; Miller 1986), whereas drawing attention toindividual membersncreases their inclusion. McCannt al (1985) found thatheterogeneous timulus groups (differing on g~nderand race and thus interfer-ing with perception of the group as a unit) producedmoreclustering of recallby person (suggesting the presence of exemplarstructures in the representa-tion) than homogeneous groups. Smith & Zarate (1988) found thatstructions to form an impression of each individual memberncreased use ofexemplar information (compared o memorynstructions), although all sub-jects showed qually good learning of the groups prototypical attributes.Familiarity with particular targets might also be expected o increase inclusionof exemplar information in representations (Hampson 983; Pryor & Ostrom1981; Pyror et al 1982). On the other hand, processing constraints such astime pressure or information overload are likely to increase formation ofcategory-level representations (Medinet al 1984; Rothbart et al 1978).As noted above, exemplar-based representations produce memory-basedjudgmentsof greater variability, are moreeasily elaborated and differentiatedthan prototype representations, and are thus more ikely to produce veridicalsocial perception. However ecause exceptions or inconsistent exemplarsareeasily dealt with by subclassification in exemplar-based epresentations, ex-ceptional features maybe less likely to become losely associated with the


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


5/39

INTERGROUP ELATIONS 49category label (Rothbart & Lewis 1988). In addition, on-line estimatesgroup variability may be more accurate than judgments based on exemplarretrieval; if this is the case, a reduction in on-line group-level judgmentscaused by the increased processing of exemplars necessary to produce ex-emplar-based representations mayreduce accurate perception.Content of Group RepresentationsInvestigation of the content of group representations has always been hinderedby the possibility of response biases, particularly in post-Civil Rights Move-ment Americansociety. For example, recent evidence from survey studiessuggests that white Americans stereotypes of Blacks are becoming essprejudiced in content and less negative in affect (see Dovidio & Gaertner1986). These results have been challenged by the argument that, althoughwhite Americans comply with egalitarian norms in rejecting blatantstereotypes, their underlyingattitudes and stereotypes are at worst unchangedand at best marked by unacknowledged ambivalence. Whites attitudestoward Blacks have been portrayed as markedby conflict betweenegalitarianvalues and unacknowledged egative feelings (Gaertner & Dovidio 1986b),betweenbeliefs that racism is bad and beliefs that Blacks are makingunfairdemands now that most inequalities have been addressed (Kinder 1986;McConahay986), and between the positive and negative affect about themthat has been internalized (Katz et al 1986).Social cognition techniques hold some promise of eliciting informationabout category content in a manner hat is relatively nonreactive. For ex-ample, subjects maymake valuative (How ood is this?) or nonevaluative (Isthis a real word?) judgments or simple responses (Name his word aloud)about trait words (AMBITIOUS,AZY) isplayed after presentation ofcategory label (WHITES,BLACKS). aster response times of such judg-ments ndicate greater associative strength betweencategory labels and traitsvarying in stereotypicality and affect (Gaertner & McLaughlin 983; Dovidioet al 1986). These studies have shownhat traits traditionally associated withBlack and White stereotypes (Katz & Braly 1933) are still associated withcategory labels, suggesting that attitudes and beliefs have not changedmarkedly. n addition, positive adjectives are more losely associated with thecategory Whites, although evidence that negative traits are still morecloselyassociated with the category Blacks is mixed(Gaertner & McLaughlin 983;Dovidio et al 1986).Despite hese indications that for white students traditionally related adjec-tives are still closely associated with Blackand Whitecategories and that thecategory White is the more positive, these techniques have methodologicaland conceptualproblems. Tasksdiffer in the extent to which hey are reactive,and delays between presentation of prime words and presentation of target


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


6/39

50 MESSICK & MACKIEwords (stimulus onset asynchrony, SOA)plus the repetitive nature of thetasks might provide ample opportunity for respondents to pre-manage heirresponses. Manipulation of SOAmayhelp distinguish automatic and atten-tional consequencesof category activation on responding (Neely 1977; Fazioet al 1986), as a more sensitive indicant of response management. nparadigm that reduces response management oncerns, Devine (1986) foundthat subjects primed(outside of consciousawareness)with the content of theirBlack stereotypes (musical, ostentatious, but not including hostility-relatedwords) were more likely to interpret a later ambiguousact as aggressive,suggesting that the idea of aggression is elicited by activation of the Blackcategory. On he other hand, finding that activation of traditionally stereotyp-ic attributes facilitates some responses mayshow only that the culturalstereotype is known, but not necessarily endorsed (Devine 1986). Multiplerepresentations of groups mayexist, any or all of which maybe activatedunder different circumstances. If somecategory labels automatically activateparticular content, but other material is available to override it attentionally,the cognitive, motivational, and contextual factors that inhibit or overrideactivation of one category in favor of another become mportant. In thisregard, Gaertner & Dovidio (1986b) note that clear normative pressures canproduceeither discrimination (Larwood t al 1984) or the lack of it; but whensituational normsare vague, discrimination increases (Frey & Gaertner 1986).Similarly, stereotypic judgments ncrease under conditions of task difficultyor complexity (Bodenhausen & Wyer 1985; Bodenhausen & Lichtenstein1987; cf Futoran & Wyer 1986) and time pressure (Freund et al 1985;Krnglanski & Freund 1983).Definitive interpretation of response atency studies also awaits more udi-cious choice Of both category (prime) and attribute (target) words. The latterneed to be chosen so that affective valence and degree of association with aparticular category can be independently manipulated if possible. Use ofmultiple target and multiple subject groups is important both to distinguishsuch effects as in-group/out-group bias (although see Sagar & Schofield 1980)and majority/minority status from the effects of specific representations, andbecause somecolors (black) elicit associations that are similar to thoseproduced by the corresponding group name Blacks; Frank & Gilovich 1988).Other work continues to explore the possibility that the content ofstereotypes consists not only of abstract trait concepts but includes visualimages (Brewer 1988; Brewer et al 1981; Lynn et al 1985), behavioralassociations, physical features, typical beliefs, attitudes, and feelings (An-dersen & Klatzky 1987; Hymes1986). For example, Deaux & Lewis (1983,1984) identified four componentsof gender stereotypes--traits, role be-haviors, physical characteristics, occupational preferences--and traced theinferential links among hem. Traits continue to be of interest because, like


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


7/39

INTERGROUP ELATIONS 51typical beliefs, preferences, etc, they are unobserved nferences (RothbartPark 1986) that perceivers seem prone to make about groups even whenexternal constraints are present (Allison & Messick1985; Mackie& Allison1987; Worth et al 1987).Recent esearch has also investigated howaffect is associated with categorystructures. Fiske & Pavelchak (1986) have argued that summary ffect (de-rived previously from ntegrating affectively laden category features) is storedwith category labels and is activated when he category is activated: Activa-tion of affect thus depends on the degree to whicha stimulus fits a storedcategory. Fazio (Fazio et al 1986) has shown hat somecategories (especiallythose with whichwe have extensive direct experience) automatically activatethe linked positive or negative affect. These developmentsmayhelp integratecontradictory findings about whether and whenaffective reactions to groupsoverwhelm or are overwhelmed by cognitive content and whether affectinfluences judgments independently of or in concert with cognition(Bodenhausen 1988; Brown& Geis 1984; Jackson & Sullivan 1987; Jacksonet al 1987; Taylor & Falcone 1982; Triplet & Sugarman1987).The automatic activation of affective material associated with a groupcategory mayhave both cognitive and motivational influence.s on furtherprocessing, resulting in the increased use of heuristics and biases (IsenDaubman 984; Stephan & Stephan 1985; Wilder & Shapiro 1988). Stephan& Stephan (1985; Dijker 1987) have suggested that intergroup interactionsaccompanied y negative affect [such as anxiety resulting from ignorance andscant previous contact (Stephan & Stephan 1985)] will involve greater re-liance on heuristics that promotecategory-level rather than individualizedjudgments [as might arousal (Kim & Baron 1988.)]. Stephan & Stephanpresent data indicating that the anxiety expressedby Hispaniccollege studentsabout interaction with Anglos s negatively related to the frequency of in-tergroup contacts and positively related to assumeddissimilarity, stereotyp-ing, and xenophobia.Affect can also influence whichof several features willbe utilized by the perceiver in making nitial judgments (Forgas & Bower1987)as well as further recategorizations (Erber 1985), and it can act as ain making other evaluative judgments (Schwarz & Clore 1988).Superordinate and Subordinate CategoriesDespite suggestions that the simultaneous activation of horizontally linkedconcepts (such as gender, race, and age) might produce nteractive effectsjudgments (Deaux & Lewis 1984; Futoran & Wyer 1986), little is knownabout relationships among orizontal structures. There has been increasedinterest in whethercategory representations include diagnostic attributes thatdistinguish the group from other categories at the same evel of generality(particularly groups with which he target category is frequently contrasted),


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


8/39

52 MESSICK & MACKIErather than only typical traits [(Klayman& Ha 1987; Trope & Bassok 1982;Trope & Mackie1987); see also the role of the social frame of reference inascertaining the prototypical memberof a category (Turner 1987)]. Forexample, he attributes thought to be prototypical of university professors maybe different dependingon whether they are compared o research scientists orto high school teachers. This possibility has important implications forcategorization. First, match of a target to a category prototype or exemplarmightnot be sufficient for categorization if alternative categories have similarfeatures or members.In this case, the importance of nonprototypical butdiagnostic features increases (Trope & Mackie1987). Second, it maybe thatprototypes are more flexible than earlier use of the concept connoted; ratherthan a prototype being a constant array of features equally activated by thecategory label, certain features or particular exemplarsmaybe differentiallyactivated by alternative categories considered for classification. Such deasare consistent with notions of gender and ethnic identity as socially negotiated(Deaux & Major 1987).More ttention has been paid to the nature and organization of verticallylinked category structures. Most current conceptualizations assume somehierarchical, progression from broad social categories (such as Blacks,Whites, females, males) through increasingly specific subcategories (suchstreetwise Black, Connecticut Yankee, career woman, macho man) withindividualized structures of particular individuals as the lowest level of thehierarchy [although exemplar information can of course be associated withcategory labels at any level in the structure (Billig 1985; Fiske 1988; MillerBrewer 1986; Rothbart 1988; cf Brewer 1988)]. Considerable attention hasbeen paid to the nature of subtypes, for at least two reasons.First, the possibility that middle-level structures mightconstitute a basiclevel of categorization [as in the nonsocial domain Rosch 1978)] and thuscarry the burden of social prediction has generated a focus on subtypes(Andersen & Klatzky 1987; Ashmore t al 1984; Brewer et al 1981; DeauxKite 1988; Deaux & Lewis 1984; Deaux et al 1985a,b; Hamilton & Trolier1986; Trzebinski 1985; Trzebinski et al 1985; Weber& Crocker 1983). Socialsubtypesare easily activated (even by indirect instructions), easily generated,and organize category-relevant information in memoryas shownby cluster-ing in recall (Martin 1987; Noseworthy & Lott 1984; Walker & Antaki1986)]. It has been suggested that subtypes are richer [although Deaux t al(1985b) found no evidence for this; see also Rothbart & John (1985)],visualizable (Brewer 1988), more internally coherent, and characterizedcloser associations betweencategory labels and associated attributes (Crocker& Park 1988) and affect (Fiske 1988) than broader social categories.On he other hand, vertically related social structures are not strictlyhierarchical (Deauxet al 1985b; Lingle et al 1984), unless only prototypical


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


9/39

INTERGROUPRELATIONS 53exemplars are considered (Hampson t al 1986). "Subordinate" categories("business woman") re often distinct from generic superordinate concepts("woman") nd perhaps from logical conjunctions of superordinate categories[such as "woman" nd "business executive" (Brewer 1988)]. They exhibitnumerous diosyncratic and distinctive associations (Andersen & Klatzky1987; see also Pryor et al 1984), domain pecific effects (Linville et al 1986,1988; Weber & Crocker 1983), and between-category overlap (DeauxLewis 1983, 1984; Deauxet al 1985b). For example, because manygenericlabels, like "American," re male oriented (Eagly & Kite 1987), the degreeoverlap between these generic and subordinate categories ("Americanathlete," for example) maybe higher for male categories than for femalecategories (Deaux et al 1985b); and categorization of womenmight be moreeasily influenced by context than is categorization of men(Deaux & Major1987; Eagly & Steffen 1984). Similar effects might be expected in thecategory structures of dominantversus minority groups. All of these featuressuggest hat no basic level of categorizationexists for social stimuli: The evelof social categorization that maximizeswithin-category similarity and be-tween-categorydifferences and thus has descriptive and predictive superiorityfor a particular task will vary across social situations and contexts (Lingle et al1984; Turner 1987).A second reason for interest in subtypes is the assumption hat the mostdetailed subtyping esults in individualization of the target, "accurate" socialperception, and by implication, improved intergroup relations. Rothbart &John (1985; Hampsont al 1986) have argued that there is a trade-off betweenthe numberof attributes that a category predicts and howwell it predictsthem. Broadcategories (such as gender stereotypes) link many raits to thecategory label but only loosely, so that the certainty that any member f acategory has the trait is reduced. In contrast, membersof more specificsubcategories are morecertain to have the (fewer) attributes associated withcategory membership. Subordinate categories are therefore seen as moreaccurate predictors of fewer attributes (see also Andersen& Klatzky 1987),and beliefs about these attributes association with the category maybe hardto change (Miller & Turnbull 1986). Both Rothbart & John (1985) and Sears(1983) note that individual members f groups are reacted to more positivelythan the group as a whole, suggesting the benefits of individualized process-ing. Locksley t al (1982) argued that stereotypic beliefs might be overriddenby individuating information about a single target (Heilman 1984; Miller1986) because perceivers fall prey to the base-rate fallacy and underutilizeprior probabilities. HoweverRasinski et al (1985) demonstrated hat subjectsin fact underrevised their ownprior probabilities in the face of the newinformation, particularly diagnostic information (Bodenhausen& Wyer1985;Deaux & Lewis 1984; Jackson & Cash 1985; Nisbett et al 1983; Wyer


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


10/39

54 MESSICK & MACKIE& Martin 1986). Krueger & Rothbart (1988) demonstrated that the diagnosticstrength of both category and target information combineadditively in judg-ments: Target information overrode initial categorization only whencategoryinformation was nondiagnostic for the required judgmentand individuatinginformation was both stable and diagnostic (see also Hinsz et al 1988). Giventhe difficulty of eliminating category effects from information processingcompletely, categorizations of the target at increasingly specific sublevelsmayhave the best chance of reducing any prejudicial effects of category-based processing.Despite any benefits that subcategorization might have for intergroupperception, there is increasing agreement hat social targets initially activateprimary or primitive generic categories such as race, gender, and age[although there is little direct evidence of this (Brewer1988; Bruner 1957a,b;Fiske & Neuberg 1988; Taylor 1981)]. For example, gender identificationapparently requires less extensive processing than trait judgments BowerKarlin 1974), and it interferes with subject-target similarity judgmentsonother dimensions Brewer 1988). Such primitive categorization has consider-able influence in that it constrains the range of subcategories that maysubsequently be activated (Brewer 1988). It may nhibit subtype activationentirely (Rothbart & John 1985) unless motivation and capacity for increasedprocessing are present (Fiske & Neuberg 1988; Langer et al 1985; Rodin1987). Frequentactivation of these primitive categories increases their likeli-hood of subsequent activation (Higgins et al 1985), further enhancing thepriority of superordinate categorization over subtype classification.When ubcategorization does occur, the level at which it ceases has beenstudied in terms of fit or matchbetween arget features and category features(Bruner 1957a,b; see Fiske & Neuberg1988, Oakes1987 for reviews) and theperceivers motivation (Fiske & Neuberg 1988; Neuberg & Fiske 1987;Omoto& Borgida 1988). Categories are activated and becomesalient whenthey are matchedby behaviors and attributes present in the target, relative toother targets and other categories available. Oakes 1987) has highlighted theimportance of the perceivers social goals in determining fit, arguing thatimmediate goals sometimes undermine the influence of features such asinfrequency, rarity, and novelty that reportedly make omecategories auto-matically distinctive (Oakes &Tumer 986; Nesdale & Dharmalingam1986;Nesdale et al 1987; cf McGuireet al 1978; Taylor 1981). In general, thepresence of category-consistent features in the target confirms categorizationat the initial level, whereas the presence of inconsistent features makesrecategorization or subtyping more probable (Fiske & Neuberg1988; Fiske etal 1987; Heilman 1984; Rothbart 1988; Rothbart & John 1985; see alsoHoffman 1986).


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


11/39

INTERGROUPRELATIONS 55THE OUT-GROUP HOMOGENEITY EFFECTWhen sked to judge the variability of social groups, people judge groups towhich hey do not belong to be morehomogeneoushan in-groups (Linville etal 1986, 1988; Mullen & Hu 1989; Quattrone 1986; Wilder 1984b). Evenwhen he perceived group is held constant (Judd & Park 1988; Worth et al1987), it is seen as more variable by its own members han by membersofother groups.Evidencefor this effect has been found with different conceptualizationsand measurementsof homogeneity(see Linville et al 1986, 1988; MullenHu 1989; Quattrone 1986 for reviews), with real and minimal groups (Simon& Brown1987; Judd & Park 1988; Worth et al 1987). Perceived homogeneityhas been associated with polarized evaluative judgments about group mem-bers [(Linville & Jones 11980); although this mayhold only when the di-mensions of evaluation are correlated (Judd & Lusk 1984)], with moreconfident member-to-group and group-to-member inferences (QuattroneJones 1980; Nisbett et al 1983), and with increased intergroup bias (Wilder1978). Considerable effort has therefore been directed at understanding howthe cognitive representations of in-groups and out-groups differ so as toinfluence judged variability.In-group vs Out-group RepresentationsLinville (Linville et al 1986, 1988) used an exemplarmodelof group repre-sentations to explain why n-groups are seen as more variable than out-groups. In this view, variability judgments depend on (a) the retrievalindividual exemplars rommemory, nd (b) the use of an availability heuristicto estimate the shape of the groupdistribution. Linville distinguishes betweenvariability (the degree to which group members re seen as being dispersed)and differentiation (the likelihood of distinguishing among roup members na particular attribute). Whereasariability reflects the spread of a distribution,differentiation reflects the number f attribute levels and their likelihood.Because ncreased contact with a group increases the numberof exemplarsaswell as the number f ways n which they differ, such contact should increaseboth perceived variability and differentiation. Because people have morecontact with in-group than out-group members,according to Linville, theformer are seen as more differentiated and variable.There is evidence from computersimulations of the model hat exposure tomultiple exemplars ncreases perceived differentiation, and to a lesser extent,variability (Linville et al 1988, Exp. 1). These esults are consistent with theknown tatistical relationships among ariability, dimensional complexity,and sample size. In addition, Linville et al (1988 Exp. 2) showedout-group


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


12/39

56 MESSICK & MACKIEhomogeneityeffects in groups where more in-group than out-group memberswere known such as with age cohorts and nationality), but not when cquain-tance included approximately equal numbers of in-group and out-groupmembers genders). More mportantly, increased familiarity over time withthe same n-group (college class) resulted in greater differentiation (Exp.There are reasons to suspect that the numberof in-group and out-groupexemplars known s not the sole mediator of the homogeneity ffect. First,with no knowledge, perceivers seem to assume homogeneity in minimalout-groups (Judd & Park 1988; Quattrone 1986; Worth et al 1987) comparedto minimal in-groups, although the difference appears primarily due to ex-aggerations of out-group homogeneity(Mullen & Hu 1989). One could arguethat there is always one known n-group exemplar, the self. Second, groupsize contributes to subjects judgments of variability independently of ex-emplar retrieval. For example,subjects estimates of the variability of mini-mal in-groups and out-groups of various sizes (where the self was the onlyexempl~ hat could be retrieved) showa positive relationship to size of thegroup (Simon & Brown 1987). A group of 50 seems more variable thangroup of 5. Third, there is no simple relationship between the number ofgroup members eportedly known nd variability estimates (Jones et al 1981;Linville 1982). Fourth, out-group homogeneityeffects have been reportedwith gender groups, where equal familiarity is assumed Carpenter & Ostrom1985; Park & Rothbart 1982).For these reasons it appears that how information about in-group andout-group members s processed maybe as important as how manymembersfrom each group are encountered. In particular, variability estimates aresensitive to whether information about individual exemplars s distinguishedand differentiated in the group representation (see section above; Linville et al1986). Representations of in-groups, for instance, involve moredifferentia-tion of individual members han representations of the out-group. Park &Rothbart (1982) have demonstrated that more information about distinctive(sub)category membershipsof individual in-group, compared o out-group,members can be recalled. Whereas women ecall the gender of male andfemale targets equally,well, for example, they are more likely to rememberalso the occupationof the female(in-group) than the male (out-group) targets.It is not clear whether his reflects encodingor retrieval differences (JuddPark 1988; Park & Rothbart 1982; Rothbart 1988; Rothbart & John 1985),although both imply differences in in-group and ou~-grouprepresentations.Information about in-group members s also more likely than informationabout out-group members o be organized by individuals (as measured byclustering of individual items in recall) (Carpenter & Ostrom1985; Sedikides& Ostrom1987). Thus factors that influence individuation of group members


Annu.

Rev.

Psychol.1989.4

0:45-81.D

ownloadedfroma


byDeutscheForschungsgemeins

chafton09/07/06.

Forpersonalu

seonly.


13/39

INTERGROUP ELATIONS 57should influence perceived homogeneity, independently of the number ofexemplars known.

Motivational FactorsSeveral plausible motivations to differentiate among ndividuals are con-ceptually independent of group membership.First, the possibility of futureinteraction has been shown o lead to more complex epresentations, and iffuture interaction is more ikely with the in-group than the out-group, differ-ences in complexity should result (Fiske & Neuberg1988). However, hisseems more relevant to laboratory groups, for example, than to gendergroups, since men have ample incentive to distinguish amongwomen,andvice versa.Second, argets uponwhomne is dependentare likely to be moredifferen-tiated. Althoughdependencys also often associated with in-group members,less powerful or minority groups must often depend on specific knowledge fthe majority out-groups preferences, behaviors, values, etc. Thus morecomplex representations about a majority out-group might be formed bymembers f a minority in-group, not only because the majority is greater innumber,but also because the in-group minority is dependenton them(Linvil-le et al 1986). At the same ime, perceived in-group homogeneitymight be anadaptive response to boost solidarity for a threatened minority ingroup (Simon& Brown1987). Simon& Brownassigned subjects to groups (based ostensi-bly on performance of perceptual tasks) and manipulated the majority andminority numerical status of the in-group and the out-group. Subjects madeestimates of the range of in-group and out-group scores (a measure ofvariability that is positively correlated with group size) on dimensions hatdiffered in their relevance to the classification task. Out-grouphomogeneitywas perceived only when he in-group was a majority: Whenhe in-group wasa minority, in-group homogeneitywas found. Although the minimalnature ofthe situation reduced true dependency,a dependencynterpretation of theseresults is consistent with other findings. Sedikides & Ostrom 1986) foundmore individualized representation of a high-status group (sophomores)both in-group and lower status out-group (freshmen) members.[A statusinterpretation mayalso help explain the increased differentiation of male,compared o female, targets noted by Linville et al (1988).]Interactions involving competition and cooperation may also influencedifferentiation of the in-group and out-group (Quattrone 1986; Wilder 1981).Although competition is usually associated with increased perception ofout-group homogeneity (Judd & Park 1988); and perhaps of in-groups as well(Simon& Brown 987)], it need not always result in undifferentiated views


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


14/39

58 MESSICK & MACKIEthe out-group. Consider a situation of high intergroup conflict in which thebest strategy for preservation of the in-group involves differentiating theout-group. The enemy roops makeone last desperate charge towards onesown ine. Under such conditions, motivation would be high to identify acommanding fficer, whose death will maximally undermine success of theattack. Intergroup competition could thus increase incentives for out-groupdifferentiation. Consistent with this possibility, Judd & Park found thatintergroup competition actually increased the amountof information recalledabout individual out-group members.The self as a member f the in-group ensures that there is always at leastone in-group exemplar for whom detailed and differentiated cognitivestructure exists. As noted above, this fact provides an explanation of out-group homogeneityeffects in minimal groups because one in-group exemplaris always available, even if no out-group exemplars are. The self is also areadily accessible exemplar, and thus mayhave disproportionate influence ongroup judgments Rothbart et al 1984; Judd & Park 1988). In addition, ParkRothbart (1982) suggest that differentiation of in-group members t morespecific levels occurs in order that other in-group membersmaybe dis-tinguished from the self, although this idea awaits direct testing. The con-tribution of the self to increased heterogeneity of the in-group may hus bemediated by both cognitive and motivational processes.Other Processing EffectsIn criticizing exemplar models, Park & Hastie (1987) have argued thatvariability judgments about a group can be made on-line and thus maynotdepend otally on recall of exemplars. [Linville et al (1988) acknowledgehatgroup judgments made frequently might be stored and retrieved withoutrecalculation.] This raises the issue of whether variability judgmentsaboutin-groups or out-groups are more ikely to be madeon-line, and if so, whetheron-line judgments are any more likely to reflect greater variability thanmemory-basedudgments. If on-line judgments are more likely for in-groupsthan out-groups, and if on-line judgments reflect greater variability thanmemory-based udgments, this would provide a processing explanation forthe out-group homogeneity ffect. Currently, there is no definitive evidencefor either of these suppositions. There maybe more at stake in processingin-group than out-group information, and this might lead to more on-lineprocessing. On the other hand, on-line variability judgments aboutnonmembership roups has also been reported under laboratory conditions(Park & Hastie 1987). Similarly, it is not clear whether on-line or memory-based judgmentsare more accurate indicators of actual group variability. Asnoted above, on-line judgments re not influenced by retrieval biases and, forthis reason, might be moreaccurate. On he other hand, the exemplars hat are


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


15/39

INTERGROUP ELATIONS 59likely to have a recall advantage consistent, salient, or extreme ones, forexample)are also the ones likely to receive on-line attention and to over-contribute to on-line judgments.In addition, it is possible that expectationsthat in-groups are variable and out-groups homogeneous irects attentiontowards differentiating information about the in-group and awayfrom suchinformation about the out-group (Worth 1988). Under many conditions,therefore, on-line judgments may not differ markedly from memory-basedjudgments.INTERGROUPBIASTajfel et al (1971) discovered that the merecategorization of a group of boysinto two subgroups, subgroups that were randomly determined by trivialpreferences, was sufficient to elicit behavior from the boys that favoredmembers f their newly defined in-group over out-group membersintergroupbias). This discovery challenged the idea that intergroup discrimination re-suited froma real conflict of interest between he two groups(Sherif 1967).also initiated a voluminousesearch effort to replicate and clarify the bias.Brewer (1979), Tajfel (1982), and Brewer & Kramer (1985) have reviewedmuch f this work, eaving little doubt hat the trivial or random lassificationof a group of people into two subgroups s sufficient to induce people in oneof the subgroups o favor others in that group relative to those in the othergroup. In this summaryf recent work on the intergroup bias, we focus firston theoretical and then on methodological issues.Theoretical IssuesThe most prominent theory guiding research on intergroup bias is socialidentity theory (SIT) (e.g. Tajfel & Turner 1986). In brief, social identitytheory proposes that peoples self-evaluations are shaped in part by theirgroup memberships. Furthermore, as part of a pervasive need to maintainpositive self-regard, people want to view the groups to which hey belong in apositive light. Because it is the relative position of ones owngroup incontrast to another that is important, self-esteem can be enhanced f peoplecan makea favorable comparison of their owngroup to another. Intergroupbias or discrimination, according to SIT, is such a favorable comparison.TheSIT interpretation of intergroup bias is difficult to test. It may e truethat people want to maintain positive self-esteem and that they view member-ship groups positively, but it remains to be shown hat these are the causes ofintergroup bias. A direct test of such a causal relation is, in fact, hard toimagine. However, SIT does indirectly imply that an act of intergroupdiscrimination should increase the actors self-esteem. Oakes & Turner


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


16/39

60 MESSICK & MACKIE(1980) report an experiment hat claims to demonstrate ust this. Subjects whowere categorized into two subgroups and whohad a chance to makealloca-tions that favored in-group members,manifested higher subsequent self-esteem scores than subjects whowere similarly categorized, but whohad noopportunity to makea group-favoring choice (cf Wagner t al 1986). Notingthat this experiment lacked someessential controls, Lemyre& Smith (1985)attempted to test the self-esteem hypothesis more adequately. Subjects wereeither categorized into two subgroupsor not, and roughly half tile members feach group were given a chance to make ntergroup point allocations beforemeasuresof self-esteem were collected. For the other half of the subjects,measuresof self-esteem were taken before the subjects were given the oppor-tunity to discriminate. The results indicated that subjects whohad beencategorized and who had displayed intergroup discrimination had higherself-esteem scores than subjects whohad been categorized but whohad nothad the chance to discriminate. While this finding provides support for SIT,an additional finding was unexpected. Subjects whowere simply categorizedshowed ower self-esteem scores than noncategorized subjects. As Lemyre&Smith note, mere categorization in these minimal-group tudies may hreatenself-esteem and intergroup discrimination may restore it. If minimalcategorization creates a challenge to self-esteem, and if intergroup bias is aconsequence of that threat, the generalizability of studies using suchcategorization manipulations will be severely restricted [although similareffects of trivial and important categorization have been found (Moghaddam& Stringer 1986)].Numerous tudies using SIT as an interpretive framework have providedonly mixedevidence for other key derivations from the theory. Brown t al(1986) found little evidence of a positive association between he degreein-group identification and the extent of positive intergroup differentiation.Suchresults can be explained by hypothesizing that the relationship will bemediated by the saliency of group membership nd the security of in-groupidentity (Smith 1985). However, tudies examining he idea that discrimina-tion wouldbe greater to the extent that group memberships ere made alientdid not discover much support for SIT. Both Ng (1986) and SachdevBourhis (1985) found bias for nonsalient groups only, and Sachdev & Bourhis(1987) foundwfoalience effects. Similarly, studies investigating the effectsstatus on discrimination have yielded conflicting results. Sachdev& Bourhis(1987) found that high- and equal-status groups discriminated more thanlow-status groups, whereas Finchilescu (1986) found that groups assignedlow status were more discriminatory. Ng(1985) found no differences in biasas a function of group status~ Espinoza & Garza (1985) report that minoritygroup members Hispanics) who are in a numerical minority in a group,discriminate more (are more competitive) than majority group members


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


17/39

INTERGROUPRELATIONS 61Hispanics in a majority. Simon& Brown(1987) claim that a group innumerical minority discriminates by enhancing ts perception of its relativehomogeneity[although Park & Rothbart (1982) and Linville & Jones (1980)found no relationship betweenperceptions of homogeneity nd evaluation].Simon& Brownfailed to find evidence of direct discrimination by theminority.

An mportant study that failed to find support for SIT was reported byVanbeselaere (1987). At issue in this study was the effect of cross-categorizing subjects into two different sets of subcategories. SIT offers noobvious reason why uch cross-categorization should reduce intergroup bias,but Deschamps& Doise (1978) reported data suggesting that it did. Van-beselaeres experiment, designed to correct possible flaws that Brown&Turner (1979) noted in the Deschamps& Doise procedures, closely replicatedthe original results: Simultaneously categorizing subjects on two crosseddimensions eliminated intergroup bias, measuredboth by performanceevalu-ations and by general attitude questions. Although Vanbeselaere offers notheoretical rationale for this result, it is clearly incompatiblewith SIT, sincein-group members pparently feel no need to make heir owngroup positivelydistinct from any of the others. AlthoughSIT has been responsible for almostsingle-handedly reviving intergroup research, invoking the concept of self-esteem has not provided a definitive understanding of in-group discrimination(see also Taylor & Moghaddam 987).More ecently Turner (1987) has described a self-categorization theory(SCT) that is broader than SIT and from which SIT may be derived. SCTappears to place greater emphasis than SIT on the nature of categorizationprocesses per se. In this sense the theory reflects an earlier concernwith theeffects of perceptual accentuation (Tajfel & Wilkes 1963; Doise 1978; Eiser& Stroebe 1972). According to Turner, people perceive themselves to bemembers f somegroups within a hierarchical structure of categories. Hu-mansare distinguished from nonhumans.Within humans, different groupingsare distinguished on the basis of relative intraclass similarities and interclassdifferences. Finally, within groups, the unique properties of individuals aredifferentiated.Categorization leads to perceptual distortions in that objects in the samecategory appear more similar to one another and more different from objectsin another category than they would f not categorized (see Wilder 1986a orreview; Herringer & Garza 1987). In addition, groupings hat contain the selfare special. Not only are they easily activated, they are also positivelyregarded. Theypossess "positive distinctiveness." Ethnocentrism t the grouplevel is analogous o self-esteem at the individual level.Perceptual accentuation effects coupled with positive regard for the in-group might well contribute to ethnocentrism. Perceptual accentuation effects


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


18/39

62 MESSICK & MACKIEon distinguishing dimensions maygeneralize to evaluative dimensions, orevaluative judgments might be accentuated or polarized whendistinguishingand evaluative dimensionsare correlated (Doise 1978; Eiser & Stroebe 1972;Judd & Lusk 1984). Such an interpretation might explain Vanbeselaeres1987 results by arguing that cross-classifying people simultaneously on twodimensions reduces or eliminates the perceptual distortion that simplecategorization induces, and that without the perceptual distortion, the evalua-tive gradient disappears.In summary,nearly 20 years after the discovery that mere categorizationproduced ntergroup bias, an adequate theory of the phenomenonas yet to bedeveloped. Both perceptual accentuation effects and self-esteem maintenanceseem ikely to be part of the story, but empirical findings have not definitivelyclarified their necessary or sufficient roles.Another factor, discovered in a different paradigm, s the extent to whichpeople behave, or expect to behave, in terms of their group memberships.McCallum t al (1985) discovered that when dyads played a PrisonersDilemmaGame PDG), they made more competitive choices than individualsplaying the same game. The competitive choice in the PDG ot only maxi-mizes the choosers payoff, it also maximizes the choosers competitiveadvantage, the difference between he choosers payoff and that of the other.McCallum t al (1985) reported a second experiment that ruled out thepossibility that groups were simply trying to maximizeheir owngains. Thusgroups seemed o have a more competitive orientation than individuals.Insko et al (1987) asked whether the increased group competitiveness wasdue to the interdependenceof the payoffs of the group members r to the factthat the groups in the McCallum t al study had met only through grouprepresentatives. They found that neither of these features could account forthe enhanced group competitiveness.Insko et al (1988) next reported that neither mere intragroup contact norintragroup discussion was sufficient to produce the group competitiveness.However,having to reach an intragroup consensus about the choice that thegroup memberswould make, even though the intergroup contact was in-dividual, did produce he competitivegroup effect. At this point, the evidencesuggests that the major factor involved with this so-called "discontinuity" incompetitiveness between individuals and groups (Brown1954) is whether thechoices are made ndividually (with or without discussion, visual contact, orpayoff sharing), or whether hey are made or the group (i.e. whetheron eachinteraction trial, only one choice is madeby the group or group members).Enhancedcompetition appears to characterize situations in which the groupmembers ct in lockstep. These results are consistent with Turners argumentthat group formation is an antecedent rather than a consequence of sucheffects, although Insko & Schopler (1987) offer other interpretations.


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


19/39

INTERGROUPRELATIONS 63Methodological IssuesThecentral issue to be discussed in this section has to do with the variety ofdependentvariables that have been used to measure intergroup bias. Beforereviewing these measures, however, we note a pervasive but problematiccharacteristic of nearly all of the experiments that have investigated thisphenomenon. his commonlement is a symmetry n the manipulation of theindependentvariable. In research with categorization the symmetrys that thesubjects are typically categorized into 2 (or more)groups, making t impossi-ble to determine if the bias emanatesfrom individuals whohave been catego-rized in a commonroup, or if it is directed toward others whohave been socategorized, or, of course, both. In the PDG xperiments summarized bove,the symmetrys that both groups of subjects in an experimental condition aregovernedby the same ules (e.g. there are no conditions in which individualsplay against group representatives). Thus t is impossible to say whether heincreased level of competition associated with group responding esults fromthe fact that the others are respondingas a unit, or becauseones wngroup sdoing so. Of somerelevance to this issue is the study of Rehm t al (1987),whofound that handball teams composed f 11-year-old boys whowere givenbright orange jerseys to wear during the gamewere moreaggressive than theiropponents whowore only their personal street clothes. Enhancing he groupidentity of one of the groups appeared to increase the competitivenessof theboys in that group. This kind of study, in whichmembershipn a single groupis spotlighted, is muchneeded.

Intergroup bias has been observed with an impressive array of dependentvariables. Rating measures include performance evaluations (see HinkleSchopler 1986 for a review; Sachdev & Bourhis 1987; Vanbeselaere 1987),attributions (Bond et al 1985; Brown& Wade1987; Stewart et al 1985),general evaluative ratings (Brewer & Silver 1978), and trait ratings (Rosen-baum& Holtz 1985). Behavioral measures have included direct moneyorpoint allocations (Ng 1985, 1986), allocations madeusing the Tajfel matrixes(TMs), which Bourhis & Sachdev (1986) have carefully explained (BrewerSilver 1978; Finchilescu 1986; Sachdev & Bourhis 1985), allocations madeusing the Brewer& Silver matrixes (BSMs) Brewer & Silver 1978; Herringer& Garza 1987), allocations made using the multiple alternative matrixes(MAMs) Bornstein et al 1983a,b), and choices in the PDGand(McCallumet al 1985; Insko et al 1987, 1988). This broad spectrumdependent measures would be reassuring about the pervasive nature of theintergroup bias if the evidence suggested hat these measuresall assessed thesame hing. Unfortunately, this is not the case.Brewer & Silver (1978), for instance, used the BSMs,TMs, and generalevaluative ratings to assess bias as a result of categorization and intergrouporientation. Whileboth of the behavioral measures ndicated less bias with a


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


20/39

64 MESSICK & MACKIEcooperative than with a competitive or independentorientation, the evaluativeratings showed small bias that was constant across orientations. Ng(1985)found no evidence of bias with direct monetaryallocations to in-group andout-group members,but whensubjects were asked how o weight two tasks todetermine payments, they tended to place a greater weight on the task onwhich the in-group was superior. The most controversial divergence ofmeasures is with the use of the TMsand the MAMssee Bomstein et al1983a,b and Turner 1983a,b). Typical results with the TMssuggest thatintergroup bias results from trying to achieve an in-group payment hat is aslarge as possible (maximumn-group profit or MIP)and trying to achievepaymenthat is larger than the payment o the out-group regardless of absolutesize (maximizing the difference or MD). The typical MAMindings showprevalence of maximizinghe joint payoff for the two groups, so long as thein-group gets more than the out-group (maximizing joint gain with owngroupahead, or MJO), and very little MIP and MD.The experimental study of intergroup discrimination faces the basic choreof developing a comprehensive theory of its dependent variables, and ofdetermining what the various indexes measure and how they relate to oneanother and to other theoretical concepts. Here we mention a few ideas thatmaybe useful in this pursuit.Oneof the first studies indicating that there mightbe important differencesbetween methods of measuring intergroup bias was reported by LaPiere(1934). While this study is frequently cited in research on the connectionbetweenattitudes and behavior (see, for instance, Ajzen 1987), it is rarelyseen in contemporarystudies of intergroup bias. This classic study shouldserve as a reminder hat measuresof attitudes toward a group do not necessar-ily predict behavior toward that group.

As we noted in a previous section, recent research suggests that attitudestoward racial or ethnic out-groups may nowbe more complicated than theyonce were. Gaertner & Dovidios (1986b) concept of aversive racism, impliesthat although people mayhold negative views of racial minorities, they alsocondemnacial prejudice and shun overtly racist behavior. Similarly, subjectsin mere categorization experiments rarely discriminate maximally gainst theout-group~their choices are "tempered with fairness" (Wilder 1986a:312)--suggesting either an ambivalence toward discrimination or the presence ofnormative controls on the magnitudeof discrimination. In both the attitudeand behavioral domains, therefore, there appear to be impediments to thedirect assessment of intergroup bias that wouldargue in favor of the use ofmore ndirect unobtrusive measures. Ng(1985) attributed to such factors thefailure to find discrimination with direct allocations when e did find it withan indirect measure.In allocation studies, people maynot only be reluctant to discriminate


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


21/39

INTERGROUP ELATIONS 65overtly, their allocations may e influenced by a variety of factors other thangroup membership Leventhal 1976). Ng (1986) has shown equity effects,larger allocations to groups that performed better, that tend to overrideintergroup bias. When roup performance is equal, direct allocations alsotend to be equal (Ng 1985). Moreover,when variable like status is manipu-lated by varying perceived performance (e.g. Sachdev & Bourhis 1987),self-assigned performance (e.g. Finchilescu 1986), or actual performance(e.g. Ng1986), allocations may eflect performancedistinctions rather thangroup discrimination. Thus, if a high-status group (which is always pairedwith a low-status group) shows ntergroup discrimination and a low-statusgroup showsa "negative" bias (see Sachdev& Bourhis 1987), the implicationis simply that both groups give more o the group that did better than to thegroup that did worse. In cases like this, techniques are needed hat allow forthe simultaneous assessment of tendencies to reward good performance andtendencies to overreward ones owngroup.Bourhis & Sachdev 1986) persuasively argue that the TMs re sensitivethe subtle effects of merecategorization. However,n the studies reported byMcCallumt al (1985), Insko et al (1987), and Insko et al (1988), manipula-tions that wouldhave had the effect of categorizing the subjects into those onone side of the corridor and those on the other did not result in increasedcompetition. When ompetition did occur in these studies it was more blatantand overt than that manifested n simple categorization experiments. So, justas it is necessary to use different instruments to measure emperatures fromwidely different sections of the temperature scale, it maybe that the TMs ndthe PDG re most appropriate for studying intergroup orientations that differin their intensity. Thecultivation of this idea would equire that the situationsand manipulations, as well as dependent variables, be ordered on a scale ofseverity or intensity. It would hen be possible to test the proposal hat certainkinds of measuresare moreappropriate for somekinds of situations than forothers.

Finally, we applaud the kind Of exchange that occurred between Bornsteinet al (1983a,b) and Turner (1983a,b). It can only be beneficial to examinesuch detail the possible goals or strategies of subjects in categorizationexperiments nd the relationship of these goals to the set of choices available.The issues that were raised in these papers run deeper than those raised inmanyrecent articles because they deal with the bedrock questions of howtheoretical concepts (like discrimination) are expressed in behavior. Onereason why the relative superiority of the MAMsnd TMswas not de-finitively resolved stems from different assumptions hat Bornstein et al andTurner make bout the set of motives, goals, or strategies that subjects canpursue in minimal-groupexperiments. Both Bornstein et al and Turner seemto accept the premise that subjects will choose he option that best satisfies


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


22/39

66 MESSICK & MACKIEtheir goals. They seem to differ with regard to assumptions that they makeabout what those goals can be. Bomsteinet al appear to hold that the goalsmust be one of seven orientations that include MIP, MD,and MJO.WhileTurner is somewhatvague, he views these goals as "continuous variables"(Turner 1983a:358), which we take to mean evaluative strategies in whichtrade-offs can be madebetweenone pure strategy and others (e.g. a willing-ness to exchangesomeamountof in-group payoff for an increase in relativeadvantage, or to give up some relative advantage to gain something infairness). This dispute is not about the connection of underlying theoreticalstates (strategies or orientations) to choices, but rather about the possibletheoretical states themselves. In view of the complexity that has beenobserved in studies of peoples choices of payoffs for themselves and others(MacCrimmon Messick 1976; Messick & Sentis 1985; Lurie 1987), we seeno reason to restrict the set of theoretically possible goals to the finite setproposed by Bomstein et al. Trade-offs among"pure" states are surelypossible.The exchangebetweenTurner and Bomsteinet al illustrates the major pointof this section: Careful thought about dependent variables often leads tocentral conceptual issues.IMPROVING INTERGROUP RELATIONSFrom ts very inception, the study of intergroup relations has aimednot onlyto understand but also to improve ntergroup relationships. In this section, wereviewrecent contributions toward his goal. Our irst focus is on the role thatintergroup contact mayplay, after which we mention contributions madebyother approaches.Intergroup ContactThe contact hypothesis is the proposal that under the right circumstances,direct interpersonal contact betweenmembers f two antagonistic groups willlead to a reduction in the negativity of intergroup attitudes (Allport 1954).While his principle "aspired to the role of dragon slayer" (Stephan 1987:15)early on, it has acquired the qualities of a "bag lady . . . encumberedwithexcess baggage" (Stephan 1987:17) or of a "laundry list" (Pettigrew1986:171). This metamorphosis ppears to have resulted from the accumula-tion of facts about the conditions under which ntergroup contact does or doesnot have beneficial consequences,with little parallel development f theory(Hewstone & Brown 1986b). Several authors have recently attemptedredress this imbalance. Indeed, the last few years have witnessed an explosionof theoretical articles about intergroup contact (Brewer& Miller 1984; Hew-


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


23/39

INTERGROUP ELATIONS 67stone & Brown1986b; Pettigrew 1986; Rothbart & John 1985; Stephan 1985,1987; Stephan & Stephan 1985; Wilder 1986a,b; Worchel 1986).Pettigrew (1986) has noted that the theoretical frailty of the contact hypoth-esis is commono other social psychological theories. It is logically loose,narrowly cognitive, statically focused on isolated rather than cumulativeimpacts, and mute about generality. Pettigrew documents hese charges byreference to the chapters in Hewstone& Browns (1986) important volumeand, in so doing, he mapsfruitful directions for conceptual development.At a minimum, heoretical development equires the organization of themultitude of variables that are known o influence the effectiveness of in-tergroup contact. Two ndependentbut similar efforts have recently appeared(Hewstone & Brown1986b; Stephan 1987). Both of these highlight broadercontextual factors, situational details, and the kinds of psychological pro-cesses that are evoked. Hewstone& Brownattribute great significance towhether the contact is interpersonal or intergroup, and they further examinethe perceptual and attributional consequencesof the contact as well as out-come udgments. Both the intergroup-interpersonal distinction and the rela-tively detailed attention afforded to outcomesunderscore the importanceofthe issue of generalizing from interpersonal contact to changes in intergroupattitudes.The process of generalizing from interpersonal contacts to intergroup atti-tudes is a central issue for theory and research. Brewer & Miller (1984)propose that intergroup contact will be maximallysuccessful when he groupor category memberships f the participants are as inconspicuousas possibleand when he interaction is intimate. Decategorization, according to theseauthors, is promotedby differentiation among ut-group members s well asthe personalization of intergroup contact. Miller et al (1985) and MillerBrewer 1986) summarize esearch supporting this position. The objectivedecategorizationwould ppear to be a society that is devoidof cultural, racial,or other intergroup differences--a colorblind society. A numberof theoristshave questioned the desirability of this as a societal goal. Schofield (1986)notes that suppressing ace as a meaningful opic in a biracial public school inthe United States not only obscured real differences betweenblack and whitechildren--so that when classes were organized according to performancelevel, for instance, the groupings tended to be racially homogeneous--butalso created an atmosphere in which it was taboo for children as well asteachers to discuss racial similarities and differences. Berry (1984) andHewstone& Brown 1986b) note that intergroup homogenization maybe notonly impossible but also undesirable: impossible because the activation ofprimitive categories like race or gender maybe automatic, and undesirablebecauseof the attendant loss of subcultural differences that enrich the texture


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


24/39

68 MESSICK & MACKIEof society. These heorists take the position that positive beliefs derived frominterpersonal encounters will not generalize to the group level unless theintergroup character of the interaction is madesalient (Hewstone& Brown1986). Suppressing intergroup categorization thus impedes generalization.Rothbart & John (1985) and Wilder (1986b) attack the generalization problemfrom somewhatdifferent cognitive perspectives. The former view the prob-lem as one of associating new features with a category stereotype. Perhapsbecause "poor" exemplars of a category are less accessible through thecategory label for a prejudiced person, a black scientist is more ikely to bestored with the category "scientist" than "black"--they seem to have lessimpact on judgments about the group. Wilder (1986b) employs information-processing concepts to isolate various waysgeneralization could fail. Both,however,discuss the paradox hat for an impression of a person to generalizeto that persons group, the person must be perceived as a "typical" groupmember, which implies that the person may be encumbered with negativestereotypical connotations that will retard the formationof a positive impres-sion in the first place (Rothbart & Lewis 1988; Wilder 1984a).

As a final illustration of the recent permeation of theory in studies ofintergroup contact, we note the chapter by Miller & Davidson-Podgorny(1987). These authors focus on classroom learning situations like Jigsaw(Aronson et al 1978), Learning Together (Johnson & Johnson 1975),TeamsGamesTournament Edwards et al 1972) that are designed to promoteintergroup cooperative learning [see Slavin (1985) for an assessment of theseefforts]. The authors tease out implications of three different theoreticalpositions-~expectation states theory (Cohen 1982), the ignorance model(Stephan & Stephan 1984), and the social categorization model developedBrewer & Miller (1984)--and then compare these implications to themetaanalytic findings of a numberof pertinent studies. A major contributionof the chapter is the application of social psychological heory to this area.Despite the conceptual progress made in the last few years, Pettigrew(1986) warns hat we mayhave created unreasonably high expectations for thegood that intergroup contact can achieve. Whatcan be expected of programsto bring Catholic and Protestant children together briefly in NorthernIrelandwhenboth denominations insist on separate schools (Trew 1986)? FosterFinchilescu (1986) argue that the Black-White contact that does occurSouthAfrica will do little to alter interracial attitudes so long as an explicitlyracial status hierarchy, apartheid, is the law of the land. Reicher (1986:164)espouses the extreme position that research on interracial contact not onlyoffers no hopebut is itself "part of the problem" o the extent that it acceptsracial categories, themselvessymptomsf racism, as valid. [In a similar vein,Stein (1988) claims that the odious racial views of the Third Reich weresimply lifted from contemporaryscientific thought and not invented by the


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


25/39

INTERGROUP ELATIONS 69National Socialists.] If intergroup contact per se offers only modesthope ofimproving ntergroup relations, what other alternatives are available?Other ApproachesINSTITUTIONALND EGISLATIVEHANGEn societies that allocate privi-leges differentially to different groups, be they Whites versus Blacks, An-glophone versus Francophone, Jewish versus Arab, or Protestant versusCatholic, efforts must be made o change he social structure in ways hat willpromote intergroup harmony.These changes are probably the most importantof all since other efforts to promote ntergroup peace are unlikely to succeedin societies that condone nstitutional discrimination. The elimination ofsocietal barriers betweengroups usually entails the replacementof one set ofrules, procedures,and institutions with another set; and it is here that socialpsychological research can be useful in illuminating the strengths and weak-nesses of various alternatives (see, for instance, Gerard & Miller 1975;Brewer & Miller 1984).CONFLICT RESOLUTION~ BARGAINING~ AND NEGOTIATION To the extentthat there is a real conflict of interest between wo groups, attempts to settlethe dispute fairly and efficiently mayprevent the disagreement rom escalat-ing into intergroup hostility (Pruitt &Rubin1986). It strikes us as curious thatthe immenseiterature on conflict management, literature large enough osupport two scholarly journals--The Journal of Conflict Resolution and TheNegotiation Journal--remains by and large apart from the literature onintergroup relations and vice versa. An nspection of the reference lists of tworecently published texts---Social Conflict (Pruitt & Rubin 1986) and Theoriesoflntergroup Relations (Taylor & Moghaddam987)~reveals little commoncontent, despite the fact that dispute managementechniques like thosedescribed by Fisher & Ury (1981) or Raiffa (1982) may be thought of"preventive" intergroup relations when pplied to intergroup conflicts (Fisher1983). Recent books on negotiation and conflict managementhave beenpublished by Lewicki & Litterer (1985), Lewicki et al (1986), Rangarajan(1985), and Roth (1985). Notable exceptions to this insularity are the volumeon intergroup conflict edited by Stroebe et al (1988), particularly the chapterby Morley et al (1988), and Worchel & Austins edited book (1986).Conflict management t the international level may nvolve efforts tounderstand he causes, pitfalls, and consequences f foreign policies (Tetlock1986). Psychologists can contribute to policy formation by studying thepsychological consequences of various policy options. Gergen (1974), forexample, showedwhy t was naive to expect countries receiving US foreignaid to feel unadulterated gratitude for the aid. Allison &Messick 1985, 1987)


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


26/39

70 MESSICK & MACKIEhave shown that group decisions, such as are made by governments, areassumedo reflect the views of the citizens, even n situations wherecitizensare perceived to have little influence on governmentdecisions. Governmentpolicies and decisions may herefore become mportant determinants of peo-ples beliefs about the compatibility of their interests with those of othernations. Bar-Tal & Geva(1986) pinpoint such belief incompatibility asnecessary condition for international conflict.Our position is that the principles of dispute management nd conflictresolution need to be explicitly woven nto the fabric of intergroup relations.CATEGORIZATIONrevious research on intergroup contact has failed togenerate great optimism.If the consequencesof c6ntact are unreliable, and ifit is not feasible or desirable to eliminate intergroup categories as Brewer&Miller (1984) recommend, what options are left? We mention severalapproaches that derive from the emerging focus on categorization. Wilders(1986b) excellent review is recommendedor more detail.Changing he out-group stereotype As mentioned earlier, direct efforts tochangebeliefs and attitudes toward out-groups, either through direct contactor information campaigns, have been carefully analyzed by Rothbart & John(1985) and Wilder (1986a). Both of these analyses focus on the difficultygeneralizing from a positive interaction with an out-group membero theout-group tself. Rothbart &Park (1986) further argue that the nature of traitadjectives associated with stereotypes maymakestereotypes differentiallyresistant to change. Negative traits in particular are easy to confirm butdifficult to disconfirm. Disconfirming evidence may not be available ifcontact with outgroup members s avoided.Weakening ntergroup boundaries To the extent that intergroup boundariesare blurred or weakened,ntergroup interaction will be more ikely to occur interms of personal characteristics than category labels, and intergroup bias willbe reduced. Intergroup boundaries can be weakened n manyways, including,for instance, by cross-cutting category memberships o that an out-groupmember n one categorization is an in-group member on another (Van-beselaere 1987), by reducing cues to category membership Worchel 1979),by disrupting the assumed belief dissimilarity to the out-group (Wilder1986b), and by highlighting superordinate categories (Kramer 1988). Gaert-ner (1985; cited in Gaertner & Dovidio 1986a) reports a study in which theseating pattern of A and B group members was varied from segregated(AAAABBBB),hrough partially integrated (BAABABBA),o fully inte-grated (ABABABAB).embersof the more integrated groups experiencedtheir mergedgroups as a unit, showed ess intergroup bias in leader choice,


Annu.Rev.

Psychol.1989.4

0:45-81.

Do

wnloadedfroma



afton09/07/06.

Forpersonaluseonly.


27/39

INTERGROUPRELATIONS 71expressed more satisfaction with group membership, and cooperated morethan membersof segregated groups.However,even the assimilation of new members nto a single group mayevoke categorization. Moreland 1985) found that in the initial stagesintegration, new members erceived themselves as an in-group and saw theold members s an out-group, with manyof the attendant consequences ofintergroup categorization.Diminishing the intensity of ingroup identification Intergroup boundariesare likely to be more salient whenones membershipn the in-group servesimportant personal goals, including the maintenanceof positive self-regard(Tajfel & Turner 1986). Thus reducing the instrumental importance of groupmembership, perhaps by providing alternative routes to goal achievement,maydecrease the tendency o perceive and to interact with others in categori-cal terms. Weknowof little experimental research on this point.CONCLUDING COMMENTSSpace constraints prevent us from covering all of the important research thathas been conducted n recent years. In these concluding remarkswe first notethree such domains hat deserve attention.Intergroup theory, and particularly social identity theory, has helped forgea newsubdiscipline at the interface of language, communication, nd socialpsychology (for reviews see Clark 1985; Giles & Wiemann 988; Giles et al1987). The predominant guiding framework in the area has been speechaccommodationheory (SAT), which deals with the cognitive, motivational,and affective processes that underlie speech convergence adaptation to oth-ers speech) and divergence (accentuation of linguistic differences). Speechdivergence, which can be viewed as a symbolic tactic for maintaining in-tergroup distinctiveness, might be usefully considered as a moresubtle mea-sure in studies of in-group bias.An ntergroup perspective has also provided an impetus to theory develop-ment n social influence. Moscovicis 1980) treatment of the issue of minor-ity influence as an intergroup problemand Mugnys 1982) application of SITto minority persuasion attempts injected social influence research with a vigorit has lacked since the 1950s. This work s thoroughly reviewed in Chaiken&Stangor (1987), Levine & Russo (1987), Maass et al (1987), MoscoviciMugny1987), and Wolf (1987). Referent identity theory (Turner 1982,1987), which posits that recognition of group membership s a necessarycondition for influence, has had less impact but has been successfully appliedto an integrative understanding of group polarization (Mackie1986; Wetherell1987) and conformity (Hogg & Turner 1987).


Annu.R

ev.

Psychol.1989.4

0:45-81.

Downloadedfroma


by


afton09/07/06.

Forpersonaluse

only.


28/39

72 MESSICK & MACKIEFinally, ideas from intergroup relations have begun o seep into the study ofgroup decision-making, in particular in social dilemma ituations (MessickBre

Date post:	03-Apr-2018
Category:	Documents
Upload:	nini345
View:	235 times
Download:	0 times

Messick Und Mackie

Documents