Social Identity Theory 1 Running head : THE SELF - ESTEEM HYPOTHESIS

Distinctions are made between global and specific, personal and social, and trait and state self-esteem, and these are used to structure a review of over 40 studies concerning social identity theory’s hypothesis that (1) intergroup discrimination elevates self-esteem and (2) low self-esteem motivates discrimination. It is observed that researchers have tended to employ measures of global personal trait self-esteem in their investigations of this selfesteem hypothesis, and it is argued that measures of specific social state self-esteem are more consistent with social identity theory’s assumptions. Although no convincing evidence is found for the self-esteem hypothesis in its full and unqualified form, it is argued that this is due to a lack of specificity in its formulation and it is suggested that a more qualified and specific version of the hypothesis may be more appropriate. Social Identity Theory 3 Social Identity Theory’s Self-Esteem Hypothesis: A Review and Some Suggestions for Clarification Why do social groups engage in discriminatory behavior against one another? Usually, intergroup discrimination can be explained in terms of a realistic conflict of interests (see Campbell, 1965). For example, inequity in limited resources or differences in religious or political values can often be cited as legitimate causes of intergroup discrimination. However, in some cases discrimination may occur in the absence of sociological variables such as these, suggesting that they may not be necessary explanatory constructs. Social identity theory (Tajfel & Turner, 1979; see also Hogg & Abrams, 1988) proposes that salient social categorization coupled with ingroup identification are both necessary and sufficient for intergroup discrimination to occur along a mutually valued and relevant comparison dimension. The theory explains such apparently irrational discrimination in terms of fundamental social psychological processes that may occur in addition to, as well as in the absence of, any historical, ideological, or material justification for conflict. Social identity theory’s predictions have received empirical support from studies employing the minimal group paradigm (Tajfel, Billig, Bundy, & Flament, 1971; for reviews, see Brewer, 1979; Mullen, Brown, & Smith, 1991). Here, participants are randomly and anonymously divided into two groups (e.g., “Group A” and “Group B”) ostensibly on the basis of trivial or arbitrary criteria (e.g., preference for art or the toss of a coin). They are then asked to distribute a valuable resource (e.g., money or evaluative points) between other participants who are usually only identified by code number and group membership. No social interaction is allowed to occur between participants, and they are not given the opportunity to award points to themselves. These conditions are assumed to eliminate all rational motives, such as interpersonal pressures or self-interest, from the resourcedistribution task. Furthermore, the minimal social content of the newly formed groups is thought to exclude the influence of any group-specific norms that might prescribe conflict. Typically, participants in minimal group experiments have been found to allocate more money to members of the ingroup than to members of the outgroup. Notably, this intergroup discrimination is displayed even when it fails to optimize absolute ingroup rewards (e.g., Tajfel et al., 1971), suggesting that ingroup members wish to maximize ingroup profit relative to the outgroup, rather than simply accumulate as much money for the ingroup as possible. Oakes, Haslam, and Turner (1994) have summarized five variables that social identity theory assumes to be related to the degree of intergroup discrimination displayed in the minimal group paradigm: (1) the degree to which subjects identified with the relevant ingroup and (2) the salience of the relevant social categorization in the setting, (3) the importance and relevance of the comparative dimension to ingroup identity, (4) the degree to which the groups were comparable on that dimension (similar, close, ambiguously different), including in particular, (5) the ingroup’s relative status and the character of the perceived status differences between the groups (Tajfel, 1978b; Turner, R. Brown and Tajfel, 1979; Sachdev and Bourhis, 1987). (p. 83) One crucial variable has been excluded from this list: self-esteem. Social identity theory proposes that minimal group members discriminate in favour of their ingroup in order to imbue it with a “positive distinctiveness” (Hogg & Abrams, 1988, p. 23). The motivation behind this strategy is thought to be the desire to achieve, maintain, or enhance a positive social identity. It is assumed that, by establishing positive distinctiveness for the ingroup as a whole, ingroup members are establishing a positive social identity for themselves and hence positive self-esteem. Hogg and Abrams (1990; see also Abrams & Hogg, 1988) have derived two informal corollaries from Turner's (1982) exposition of this assumption, and these are taken to embody the self-esteem hypothesis: Social Identity Theory 4 1. Successful intergroup discrimination enhances social identity and thus elevates self-esteem. Self-esteem is a dependent variable, a product of specific forms of intergroup behaviour. 2. Depressed or threatened self-esteem promotes intergroup discrimination because of a need for self-esteem. Self-esteem is an independent variable, a motivating force for specific forms of intergroup behaviour. (p. 33) Research investigating the self-esteem hypothesis has tended to focus on either corollary 1 (Chin & McClintock, 1993, Expt.’s 1 & 2; Gagnon & Bourhis, 1996; Hewstone, Islam, & Judd, 1993, Expt. 2; Hogg & Turner, 1985a, 1985b; Hogg, Turner, NascimentoSchulze, & Spriggs, 1986, Expt. 1; Hunter, Platow, Howard, & Stringer, 1996; Hunter, Stringer, & Coleman, 1993; Islam & Hewstone, 1993, Expt. 2; Kelly, 1988; Mullin & Hogg, 1995; Nascimento-Schulze, 1993; Oakes & Turner, 1980; Turner & Spriggs, 1982, as cited in Lemyre & Smith, 1985; Vanbeselaere, 1991, 1996; Vickers, Abrams, & Hogg, 1985, as cited in Abrams & Hogg, 1988) or corollary 2 (Abrams, 1982, 1983, as cited in Abrams &Hogg, 1988, and Hogg & Abrams, 1990; Bourhis, 1994; Brockner & Chen, 1996; J. Brown,Collins, & Schmidt, 1988, Expt.’s 1 & 2; Crocker & Luhtanen, 1990; Crocker & Schwartz,1985; Crocker, Thompson, McGraw, & Ingerman, 1987, Expt.’s 1 & 2; Finchilescu, 1986;Hogg & Turner, 1987; Long & Spears, 1995, as cited in Long & Spears, 1997; Long,Spears, & Manstead, 1994; Meindl & Lerner, 1984, Expt. 1; Platow, Harley, Hunter,Hanning, Shave, & O’Connell, 1997, Expt.’s 1 & 2; Ruttenberg, Zea, & Sigelman, 1996;Sachdev & Bourhis, 1984, 1985, 1987, 1991; Seta & Seta, 1992, 1996; Sidanius, Pratto, &Mitchell, 1994; Smith & Tyler, 1997, Expt. 2). Relatively few studies have attempted toaddress both corollaries simultaneously (Branscombe & Wann, 1994; Hogg & Morkans,1989, as cited in Hogg & Abrams, 1990; Hogg & Sunderland, 1991; Hogg et al., 1986, Expt.2; Lemyre & Smith, 1985; Maass, Ceccarelli, & Rudin, 1996, Expt.’s 1 & 2; Meindl & Lerner,1984, Expt. 2; Verkuyten, 1997; Wagner, Lampen, & Syllwasschy, 1986). There are some serious limitations with studies that examine one corollary withoutalso investigating the other. Correlations between discrimination and self-esteem do notestablish the direction of causality between the two. Hence, positive correlations betweendiscrimination and posttest self-esteem could imply either that discrimination elevates self-esteem (as per corollary 1) or that it has no effect on self-esteem and that people with highpretest self-esteem tend to discriminate more than people with low pretest self-esteem.Similarly, negative correlations between pretest self-esteem and discrimination could beinterpreted either as evidence that low pretest self-esteem motivates discrimination (as percorollary 2) or as evidence that people who engage in high levels of discrimination tend topossess low self-esteem for some other reason, perhaps even as a consequence of theirdiscrimination.The presence of appropriate control conditions can alleviate these interpretationalproblems to some extent, however, any conclusive test of the self-esteem hypothesis should ideally examine both of its corollaries simultaneously. It is not enough to show thatlow self-esteem promotes discrimination or that it is elevated by discrimination. The self-esteem hypothesis makes the specific predictions that low self-esteem promotesdiscrimination and that this discrimination then elevates self-esteem. In this sense, studiesthat address one aspect of the hypothesis without the other have less than half theinterpretational power of those that investigate both together.Different Types of Self-EsteemMany researchers (Abrams & Hogg, 1988; Branscombe & Wann, 1994; Brewer &Miller, 1996; Chin & McClintock, 1993; Crocker & Luhtanen, 1990; Hogg & Abrams, 1990;Hogg & Sunderland, 1991; Hunter et al., 1996; Kelly, 1988; Long & Spears, 1997; Long etal., 1994; Wagner et al., 1986) have drawn attention to a mismatch between the type ofself-esteem that is typically measured in investigations of the self-esteem hypothesis andthe type of self-esteem that should be being measured according to social identity theory’s assumptions. Three distinctions between different types of self-esteem help to clarify thisissue. These included (a) global vs. specific self-esteem (e.g., Rosenberg, Schooler, Social Identity Theory 5 Schoenbach, & Rosenberg, 1995), (b) trait vs. state self-esteem (e.g., Brewer & Miller,1996; Heatherton & Polivy, 1991), and (c) personal vs. social self-esteem (e.g., Breckler &Greenwald, 1986; Crocker & Luhtanen, 1990; Long et al., 1994; Luhtanen & Crocker, 1991,1992).Global vs. Specific Self-EsteemRosenberg et al. (1995) point out that, “in general, self-esteem has not proved to bean impressive predictor of behavioral outcomes” (p. 144). They suggest that this lack ofpredictive power may be due, in part, to researchers’ ignorance of the global-specificdistinction. Global self-esteem refers to the esteem in which one holds one’s overall self-image, whereas specific self-esteem refers to the esteem in which one holds a particularself-image. Following Fishbein and Azjen (1975), it is argued that specific measures of self-esteem will prove to be more powerful predictors of related specific behavior than globalmeasures (see also Crocker, Luhtanen, Blaine, & Broadnax, 1994; Dutton & Brown, 1997). In the context of the self-esteem hypothesis, measures of pretest self-esteem shouldtherefore be made specific to the particular ingroup under investigation.In addition, particular instances of intergroup discrimination are liable to have more ofan impact on specific, rather than global, self-esteem. Hence, posttest measures shouldalso address the ingroup involved in the discrimination. These issues of specificity areespecially relevant in the minimal group paradigm, where the social categories that areemployed are likely to possess only very limited significance for participants’ overall self-images.Trait vs. State Self-EsteemAt any given moment an individual’s level of self-esteem can be divided into twoseparate components: trait and state self-esteem (for a related discussion, see Kline,1993). Trait self-esteem is the product of self-evaluations that are made over a relativelylong period of time, whereas state self-esteem is the product of self-evaluations carried out in the immediate present.Traditional self-esteem scales have tended to focus on self-esteem as afundamentally stable trait and to regard fluctuations in test-retest performance asinconvenient error variance. However, self-esteem may also be conceived as the product ofa dynamic process of self-evaluation that is constantly updating previous self-attitudesaccording to new experiences. From this perspective, trait self-esteem is more properlyregarded as the average of a series of state self-esteem values that vary across time.When considering the self-esteem hypothesis, the trait-state distinction is importantfor three reasons. First, it would seem commonsense to employ state self-esteem scaleswhen attempting to detect transitory changes in self-regard brought about through the sortof short-term discrimination that is exhibited in laboratory research. Second, as Oakes et al.(1994) noted, social identity theory makes it plain that the immediate salience of one’s self-image as an ingroup member influences the degree of subsequent discrimination. This implies that self-esteem should be addressed in the here and now of the experimentalcontext, rather than across time (see Long & Spears, 1997). Third, in the case of theminimal group paradigm social categorizations are more likely to have an impact on state,rather than trait, self-esteem because participants perceive their group membership to benew and transient (see Hogg & Sunderland, 1991).Personal vs. Social Self-EsteemThe distinction between personal and social self-esteem is the most important from atheoretical perspective because it reflects the distinction between personal and socialidentity that is so crucial to social identity theory’s nonreductionist account of intergroupbehavior. Social identity theory proposes that group members engage in intergroupdiscrimination in order to achieve, maintain, or enhance the positive distinctiveness of theirsocial identity. In other words, group members are motivated to manage their social self-esteem: the esteem in which they hold the shared self-image that constitutes their social psychological ingroup. The notion of social self-esteem involves a redefinition of self-esteem at the group level as an attitude concerning a collective self-image. This Social Identity Theory 6 reconceptualization of self-esteem leaves it commensurate with social identity theory’sgroup-level concept of positive distinctiveness and reinforces the idea that group membersare striving for positive self-esteem as ingroup members, rather than as individuals.To misinterpret the self-esteem hypothesis in terms of personal self-esteem would beto recast social identity theory as an individualistic theory of intergroup behavior (for adiscussion, see Long & Spears, 1997). Reductionist theories of this type argue that thesocial group is no more than an aggregation of separate individuals who behave as a groupin order to satisfy their own individual needs and motives (for a discussion, see Turner,Hogg, Oakes, Reicher, & Wetherell, 1987). Although this notion of functionalinterdependence may provide a plausible account of intergroup discrimination at theinterpersonal level, it has difficulty explaining discrimination that occurs in the absence ofinterpersonal contact, such as in the minimal group paradigm. Consequently, it is importantto differentiate this personal self-esteem hypothesis from social identity theory’s social selfesteem hypothesis.The various types of self-esteem discussed above (i.e., global vs. specific, personalvs. social, and trait vs. state) can be combined to form eight conceptually separate forms ofself-esteem. Example items are presented for each type of self-esteem in Table 1.1 -Insert Table 1 about here --Different Types of Self-Esteem ScaleIt seems clear that a fair test of the self-esteem hypothesis can only be accomplishedusing measures of specific social state self-esteem. However, there has been somedisjuncture between theory and research in this respect, and this may be attributed to thelack of appropriate self-esteem scales that are currently available.Most self-esteem scales tend to focus on global personal trait self-esteem (e.g.,Julian, Bishop, & Fiedler’s, 1966, evaluative dimension of the semantic differential scale,JBF; Kuhn & McPartland’s, 1954, Twenty Statements Test,2TST; Rosenberg’s, 1965, SelfEsteem Scale, RSES; Sherwood's, 1962, Self-Concept Inventory, as cited in Blascovich &Tomaka, 1991). Some scales have been designed to measure the trait self-esteemconnected with specific areas of one’s personal identity (e.g., Helmreich & Stapp’s, 1974,Texas Social Behaviour Inventory,3as cited in Blascovich & Tomaka, 1991; Marsh &O’Neill’s, 1984, Self-Description Questionnaire III, as cited in Hunter et al., 1996). A fewscales purport to tap global personal state self-esteem (Gergen’s, 1962, Self-ValuationTriads test, as cited in Major, Sciacchitano, & Crocker, 1993; Heatherton & Polivy’s, 1991,State Self-Esteem Scale).Only one scale has attempted to address social self-esteem: Luhtanen and Crocker(1992) have developed and tested (Crocker & Luhtanen, 1990) a scale designed tomeasure global social trait self-esteem. The Collective Self-Esteem Scale (CSES) consistsof four 4-item subscales assessing membership esteem (self-evaluation as an ingroupmember), identity esteem (evaluation of the subjective importance of the ingroup), public self-esteem (perceptions of other’s evaluations of the ingroup), and private self-esteem(self-evaluation of the ingroup as a whole).The CSES has been shown to have significant but low correlations with globalpersonal trait self-esteem (measured using RSES), “suggesting that although collective andpersonal self-esteem are related, the two constructs are relatively distinct empirically aswell as conceptually” (Luhtanen & Crocker, 1991, p. 223). Although this may be true, thescale does not qualify as the ideal instrument for testing the self-esteem hypothesis (for adiscussion, see Long & Spears, 1997).The first problem with the CSES relates to its content. As Luhtanen and Crocker(1992) readily admit, only the private subscale approaches the conceptualization of socialself-esteem implied by social identity theory. The remaining three subscales focus oninterpersonal evaluations of belonging, importance, and respect by others. Consequently,for the purposes of testing the self-esteem hypothesis the 16-item CSES actually reduces to a 4-item subscale. The small size of this subscale is bound to render it relativelyinsensitive. Social Identity Theory 7 A second problem with the CSES is that it is designed to assess trait, rather thanstate, self-esteem. This makes it unsuitable for detecting the type of short-term changes inself-esteem that may occur as a result of discrimination (see Maass et al., 1996).Finally, the scale’s instructions require participants to consider their “gender, race,religion, nationality, ethnicity and socioeconomic class” (Luhtanen & Crocker, 1992, p. 305,italics in the original) simultaneously while responding to each item. Presuming thatparticipants possess the cognitive capacity to accomplish this rather complex task, thescale should theoretically provide a measure of global, rather than specific, social self-esteem. Crocker et al. (1994) argue that the global nature of the CSES makes it possible toapply it to new groups, such as those created in the minimal group paradigm, however, it isdifficult to understand how global social self-esteem might relate to any specific instance ofintergroup behavior. As Abrams (1992) remarks, “there is no logical reason why evaluationsof all of one’s social-group memberships should be equally positive, or that positive regard of several implies positive esteem of any particular one” (p. 65).To be fair, Luhtanen and Crocker (1992) point out that their scale is flexible enough tobe adapted to address specific ingroups without compromising its psychometric properties.Some research can attest to the scale’s psychometric robustness in this respect (Crockeret al., 1994; Ethier & Deaux, 1990; Long et al., 1994; Phinney, Chavira & Tate, 1993;Ruttenberg et al., 1996; Verkuyten, 1997).All in all then, the private subscale of the CSES aims to elicit a global evaluation ofmany different ingroups across time, rather than an immediate evaluation of a specificsalient ingroup. To our knowledge, no purpose-built measures of either specific social traitor specific social state self-esteem exist.Collating the EvidenceMore than 20 studies have addressed the self-esteem hypothesis since the first majorreview in this area (Abrams & Hogg, 1988; see also Hogg & Abrams, 1990) and although subsequent minor reviews have dealt with some of this new evidence (e.g., Abrams, 1992;Crocker et al., 1993; Long & Spears, 1997), they have not attempted to evaluate old andnew evidence together. The aim of this article is to provide a comprehensive review ofresearch relevant to the self-esteem hypothesis.Content and Structure of the ReviewWe had some reservations about evaluating evidence relating to corollary 2 thatrelies on differential group status. Social identity theory predicts that members of low statusgroups will have relatively low specific social self-esteem, and they should thereforepossess greater motivation to engage in intergroup discrimination than members of highstatus groups (see Hinkle & Brown, 1990). However, some researchers have expressedconcerns about equating status with self-esteem too readily (Abrams & Hogg, 1988;Crocker et al., 1993; Hogg & Sunderland, 1991).Status hierarchies between ingroups and outgroups are usually anchored by relatively objective criteria (e.g., money or power). Consequently, there is likely to be relatively littlevariability between different group members’ perceptions of ingroup and outgroup status. Incontrast, specific social self-esteem derives from subjective satisfaction with an ingroupand is likely to be far more variable between members of the same group. The attitude thatone holds about a given ingroup may be influenced by others’ attitudes about that group,however, it will also retain a fair amount of independence from these attitudes. Hence, evenmembers of low status groups can possess high specific social self-esteem (for a review,see Crocker & Major, 1989). Consequently, there is theoretical grounding for treatingspecific social self-esteem as separate from ingroup status. We therefore decided todiscount evidence relating to corollary 2 that relies on differential group status (exclusionsinclude Bourhis, 1994; J. Brown et al., 1988, Expt. 2; Crocker & Luhtanen, 1990;Finchilescu, 1986; Platow et al., 1997, Expt. 1; Sachdev & Bourhis, 1984, 1985, 1987,1991; for a review of other unequal status group studies, see Sachdev & Bourhis, 1987).4 We also viewed evidence depending on normative discrimination as being potentiallyproblematic. Normative discrimination refers to discrimination that is prescribed by explicit Social Identity Theory 8 or implicit cues embodied in the intergroup situation. So, for example, discrimination thatoccurs between two groups that have a long history of conflict can be treated as beingrelatively normative because it is prescribed by previous intergroup relations.Turner (1980) argues that normative discrimination merely reflects the status quo ofthe current intergroup situation and so does not represent genuine social competition (seeTurner, 1975). In contrast, what we will call competitive discrimination (i.e., sociallycompetitive discrimination) operates against intergroup norms to bring about social changevis-à-vis positive ingroup distinctiveness. The self-esteem hypothesis only really applies tocompetitive discrimination. Normative discrimination is not thought to be linked to identity-contingent self-esteem (see Abrams & Hogg, 1988; for a related discussion, seeNascimento-Schulze, 1993).It remains possible, however, that normative discrimination provides a way ofsatisfying the need for positive self-esteem independent of social identity theory’s positive distinctiveness explanation. The self-satisfaction derived from “doing the right thing” bycomplying with relevant intergroup norms may enhance self-esteem (cf. corollary 1; for adiscussion, see Hogg et al., 1986). Furthermore, group members who possess low self-esteem may be motivated to exhibit greater normative discrimination in anticipation of thisenhancement (cf. corollary 2).Considering that the need for positive self-esteem may motivate, and be satiated by,both competitive and normative discrimination, evidence involving normative discriminationshould theoretically be discounted from this review in order to reduce the risk of acceptingfalse evidence in favour of the self-esteem hypothesis. However, most cases of intergroupconflict are likely to contain both forms of discrimination in varying degrees, and it is difficultto say with any certainty that one form has taken precedence over the other. For example,competitive discrimination may predominate between groups that have a history of conflictif, for some reason, group members perceive the intergroup situation to be relatively insecure and amenable to social change. Consequently, we decided to include evidencederiving from intergroup situations in which normative discrimination was possible, on theassumption that it was not inevitable. However, we also decided to treat this evidenceseparately from that deriving from situations in which competitive discrimination was mostprobable in order to provide a clearer picture of both sets of evidence.Hence, the following review is divided into two sections. The first section deals withexperiments in which normative discrimination is unlikely to occur (e.g., discriminationbetween groups of equal status, such as in the standard minimal group paradigm). Thesecond section contains evidence that derives from experiments that may confoundcompetitive discrimination with normative discrimination (e.g., discrimination betweengroups of unequal status, such as most real groups). In each section, supportive andunsupportive evidence for each corollary of the self-esteem hypothesis is presented in turn,followed by studies that address both corollaries simultaneously.5 Evidence For and Against the Self-Esteem HypothesisEvidence based on Competitive DiscriminationCorollary 1Supportive Evidence. Oakes and Turner (1980) conducted the first test of corollary 1.Participants were randomly assigned to one of two minimal groups ostensibly on the basisof their preference for abstract paintings by Klee and Kandinsky. Participants in theexperimental condition then completed an intergroup points-distribution task while those inthe control condition read a newspaper article about arranged marriages. All participantsthen completed a composite posttest measure of global personal state self-esteemcontaining items from the TST, JBF, and RSES (see previous).Significant intergroup discrimination occurred in the experimental condition.Furthermore, participants in the experimental condition reported significantly higher posttestself-esteem than those in the control condition. This result was reliable on the TST and JBF, but not the RSES. The researchers concluded that “minimal intergroup discriminationdoes tend to increase self-esteem” (p. 299). However, they were also aware of a number of Social Identity Theory 9 other explanations that may have accounted for their findings, and it is worth consideringthese in some detail because they provide a useful framework within which to evaluate laterstudies.First, Oakes and Turner suggested that the minimal group paradigm might containimplicit demand characteristics (Orne, 1962) that goad participants into exhibitingintergroup discrimination. Hence, it could be argued that discrimination is an artefact of theexperimental situation rather than a genuine social psychological phenomenon.Furthermore, according to this demand characteristics hypothesis, the elevated self-esteemreported in the experimental condition may be attributed to the self-satisfaction thatparticipants derive from “doing well” in the experiment by complying with its implicit goal ofeliciting discriminatory behavior (for a discussion, see Hogg et al., 1986).In their defence against this explanation, Oakes and Turner cite unpublished work bySt. Claire (1979, later published in St. Claire & Turner, 1982) that, in their opinion, contradicts the demand characteristics hypothesis. However, Berkowitz (1994) has recentlysubjected St. Claire and Turner’s (1982) study to criticisms that seem to refute their originalconclusions and hence to revive the demand characteristics debate (see R. Brown, 1986;Farsides, 1993; Hartstone & Augoustinos, 1995; Schiffmann & Wicklund, 1992; Tajfel et al.,1971). Consequently, compliance with demand characteristics and the concomitantincrease in self-regard associated with this compliance remain tenable explanations ofdiscrimination and differential self-esteem in the minimal group paradigm.Second, Oakes and Turner (1980) proposed that the constant reference to groupmemberships made throughout the intergroup points-distribution task may have increasedthe salience of participants’ social identity. It is argued that this increase in salience couldhave elevated self-esteem in the experimental condition independent of any effects ofdiscrimination.Subsequent research has been unsupportive of this salience per se hypothesis. Hogg et al. (1986, Expt.’s 1 & 2) categorized participants as minimal group members eitherexplicitly or implicitly (using a similar procedure to Oakes & Turner, 1980) and then askedthem to complete an intergroup points-distribution task. Both explicitly and implicitlycategorized participants engaged in significant intergroup discrimination and showed higherglobal personal state self-esteem (measured using Oakes & Turner’s, 1980, method) thanthose who had not discriminated, suggesting that category salience was not a significantfactor in determining posttest self-esteem.Further evidence against the salience per se hypothesis comes from Lemyre andSmith (1985) who compared the posttest global personal state self-esteem of participantswho had completed a standard intergroup points-distribution task with that of participantswho had completed intra-ingroup and intra-outgroup versions. It was reasoned that if theelevation of self-esteem was brought about as a sole consequence of the salience-enhancing properties of the points-distribution task, then participants in these three conditions would not differ significantly in terms of their posttest self-esteem. In fact,participants in the intergroup condition reported significantly higher self-esteem than thosein the intragroup conditions, suggesting that intergroup discrimination elevated self-esteemover and above any effects caused by increased salience.A third criticism is that the tasks in Oakes and Turner’s (1980) experimental andcontrol conditions were of unequal psychological significance. Hence, it is possible thatparticipation in a psychology experiment may have led to an increase in self-esteem forpeople in the experimental condition (see Lemyre & Smith, 1985). Alternatively, the task ofreading a newspaper may have led to a decrease in self-esteem for people in the controlcondition (see Abrams & Hogg, 1988).A fourth criticism relates to the effects of minimal social categorization independent ofdiscrimination. Lemyre and Smith (1985) found that the self-esteem reported byparticipants immediately after they had been categorized as minimal group members was lower than that reported by uncategorized participants. They proposed that this differencemay either be because categorization represented a threat to personal identity or because Social Identity Theory 10 participants experienced more uncertainty about the purpose of their categorization relativeto those who went on to employ it during the points-distribution task. Both possibilities areapplicable to the categorized participants in Oakes and Turner’s control condition.Messick and Mackie (1989) have pointed out that “if minimal categorization creates achallenge to self-esteem, and if intergroup bias is a consequence of that threat, thegeneralisability of studies using such categorization manipulations will be severelyrestricted” (p. 60). However, additional evidence concerning Lemyre and Smith’s (1985)categorization-depression hypothesis is equivocal, with some studies finding no support(Chin & McClintock, 1993) and others finding support but with qualifications (Hogg &Turner, 1987; Vanbeselaere, 1991; Wagner et al., 1986).Finally, the psychometric properties of Oakes and Turner’s (1980) self-esteemmeasure can be brought into question. This composite scale included items from threeseparate scales presented in a random order. As Abrams and Hogg remark, “the consequences of this are difficult to predict, but it is unlikely that the validity of any of thescales remained intact” (p. 319).Furthermore, Oakes and Turner instructed participants to respond in terms of howthey felt in the immediate situation in an effort to tap state, rather than trait, self-esteem.Again, there is a distinct possibility that this procedure compromised the validity of thescales concerned. For example, in the RSES participants were asked to indicate thedegree to which they agreed or disagreed with the statement “at times I think I am no goodat all”, however, they were also instructed to respond in terms of how they felt “right now”.These conflicting demands may have confused participants and led to a mixture of state-and trait-oriented responses.Further evidence that competitive discrimination elevates self-esteem comes fromother minimal group experiments. Mullin and Hogg (1995) found that minimal groupmembers who had been categorized on an explicitly random basis reported higher levels of global personal state self-esteem (measured using a single item) after completing anintergroup points-distribution task than uncategorized participants who had completed aninterpersonal points-distribution task. Similarly, Hogg et al. (1986, Expt. 1) found thatminimal group members who had engaged in intergroup discrimination reported higherlevels of global personal state self-esteem (TST, JBF, and RSE combined) than those whohad not discriminated.Chin and McClintock (1993) have put forward social value theory as an alternative tosocial identity theory in order to account for increments in self-esteem followingdiscrimination. Social value theory proposes that individuals can be classified aspossessing either prosocial or competitive social value orientations. During interpersonalpoints-distribution, prosocial individuals tend to minimize the difference between allocationsmade to the self and others, whereas competitive individuals tend to maximize thisdifference. Chin and McClintock (1993) propose that behavior that is congruent with one’s socialvalue orientation provides an important source of self-esteem. Consequently, they arguethat increments in self-esteem following minimal intergroup discrimination may be attributedto a consistency between intergroup behavior and a competitive interpersonal social valueorientation. This is quite different from social identity theory’s explanation, which accountsfor elevated posttest self-esteem in terms of intergroup, rather than interpersonal,processes. This theoretical discrepancy is highlighted by Chin and McClintock’s additionalpredictions that individuals who possess a prosocial orientation will experience higher self-esteem following fair intergroup behavior but lower self-esteem following discrimination.In a test of these predictions (Chin & McClintock, 1993, Expt. 1), prosocial andcompetitive individuals (classified according to pretest data obtained weeks beforehand)took part in a minimal groups experiment. Participants were categorized using the Klee-Kandinsky procedure (see Oakes & Turner, 1980) before completing an intergroup pointsdistribution task that either required them to be fair, to be discriminatory, or allowed them tochoose either strategy freely. In a fourth condition, participants were categorized but did not Social Identity Theory 11 discriminate, and in a fifth, baseline, condition participants were neither categorized norgiven the opportunity to discriminate. Posttest global personal state self-esteem wasmeasured using the RSES, JBF, and the Self-Concept Inventory (see previous) combined.Competitive participants engaged in significantly more discrimination than prosocialparticipants in the free-choice condition, supporting the hypothesis that interpersonal socialvalue orientation influences intergroup behavior. Furthermore, there was a significantinteraction between forced points-distribution strategy (i.e., fairness vs. discrimination) andsocial value orientation (i.e., prosocial vs. competitive) in terms of self-esteem. However,although secondary analyses revealed the self-esteem of competitive individuals to besignificantly higher than that of prosocial individuals in the forced discrimination condition, itwas not significantly lower in the critical forced fairness condition. These results thereforeseem more in line with corollary 1 of the self-esteem hypothesis than with social valuetheory. It is also interesting to note that, contrary to Lemyre and Smith’s (1985) categorization-depression hypothesis, Chin and McClintock found no difference in self-esteem between categorized and uncategorized participants who did not discriminate.In a second experiment, Chin and McClintock (1993, Expt. 2) used a posttestmeasure of global social self-esteem (CSES, presumably in its trait form) in place ofpersonal state measures. Again, self-esteem was found to be higher following forceddiscrimination than forced fairness or no discrimination. Notably, this effect did not interactwith social value orientation.Vanbeselaere (1991) categorized participants randomly, but ostensibly on the basisof the degree to which each of two pictures (scenes of a seashore and a forest) made thestronger impression on them. Participants then completed a perceptual ability task,estimating the lengths of lines, before proceeding to evaluate the performance of eachgroup in this task. Both preand post-test global personal state self-esteem (JBF) weremeasured immediately before and after discrimination using a between-subjects design. Significant intergroup discrimination occurred and, consistent with corollary 1, posttestself-esteem was significantly higher than pretest self-esteem. It was also found that theposttest self-esteem reported by categorized participants who had discriminated wasequivalent to that reported by uncategorized participants who had not discriminated, addingsupport to Lemyre and Smith’s (1985) categorization-depression hypothesis.Interestingly, participants who were categorized simultaneously as members of twogroups reported higher pretest self-esteem than those who had only been categorized assingle-group members. Reductions in self-esteem resulting from the imposition of minimalsocial categorizations do not, therefore, appear to occur in an additive fashion.Finally, Gagnon and Bourhis (1996) categorized participants on an explicitly randombasis (the toss of a coin) before asking them to complete an intergroup points-distributiontask. Ingroup identification (single item) and specific social state self-esteem (five itemstapping feelings about group membership) were then measured. In line with corollary 1, a regression analysis found that discrimination significantly predicted positive self-esteem.Additional regression analyses suggested that this association was only relevant insofar asparticipants identified with the ingroup.Unsupportive Evidence. Three studies of competitive discrimination have foundunsupportive evidence for corollary 1 (Hogg & Morkans, 1989, see later; Hogg &Sunderland, 1991, see later; Hogg & Turner, 1985a). Hogg and Turner (1985a) categorizedparticipants as members of two minimal groups on an explicitly random basis and thenasked them to complete the usual intergroup points-distribution task. Although significantdiscrimination occurred during this task, it was unrelated to posttest global personal traitself-esteem (RSES).Corollary 2Supportive Evidence. Only two studies of competitive discrimination have foundsupportive evidence for corollary 2 (Hogg & Sunderland, 1991, see later; Platow et al., 1997, Expt. 2). Platow et al. (1997, Expt. 2) categorized participants using the Klee-Kandinsky procedure before measuring their global personal state self-esteem (JBF) and Social Identity Theory 12 specific social state self-esteem (using two items focusing on relative ingroup worth).Participants then completed a standard intergroup points-distribution task.A significant interaction between personal and social self-esteem was obtained interms of discrimination. Participants who possessed high personal and low social self-esteem and participants who possessed low personal and high social self-esteem bothengaged in significant discrimination, whereas those who possessed high or low levels ofboth forms of self-esteem did not. However, follow-up analyses revealed that participantswith high personal and low social self-esteem discriminated more than those with highpersonal and high social self-esteem. This evidence lends support to the conceptualdistinction between global personal and specific social self-esteem and provides limitedsupport for corollary 2 vis-à-vis individuals who possess high personal self-esteem.Unsupportive Evidence. Eight experiments have found that, in direct opposition tocorollary 2, high pretest self-esteem is associated with greater competitive discrimination (Abrams, 1982, 1983, as cited in Abrams & Hogg, 1988, and Hogg & Abrams, 1990;Brockner & Chen, 1996; Crocker & Luhtanen, 1990; Crocker et al., 1987; Seta & Seta,1992, 1996; Sidanius et al., 1994). Abrams (1982, as cited in Abrams & Hogg, 1988, andHogg & Abrams, 1990) found that pretest global personal trait self-esteem (RSES)correlated positively with intergroup evaluative bias, reward bias, and trait differentiation.This last association for trait differentiation was replicated in a subsequent study (Abrams,1983, as cited in Abrams & Hogg, 1988, and Hogg & Abrams, 1990).Crocker et al. (1987, Expt. 1) categorized participants as minimal group members onan explicitly random basis by asking them to draw lottery tickets that designated groupmembership. Participants then completed a measure of global personal trait self-esteem(RSES). Specific personal state self-esteem was then manipulated across participants viafalse positive or negative feedback concerning individual performance on a personality test.Participants showed significant intergroup discrimination on a series of evaluative traits but this was unrelated to global personal trait self-esteem. Furthermore, no significantdifferences in discrimination were found between participants who had received positivefeedback and those who had received negative feedback. The authors concluded that “theextent of ingroup favouritism shown by subjects in a minimal intergroup situation may beunrelated to needs for self-enhancement as operationlized by either trait self-esteem ortransient threats to the self-concept” (p. 911).Participants in Crocker et al.’s (1987, Expt. 1) study went on to rate aboveandbelow-average personality test scorers. Although, overall, above-average scorers wererated more positively than below-average scorers, participants who received failurefeedback tended to be significantly more generous in their ratings of below-average scorersthan participants in a no feedback control condition. It can be argued that, since thecategory of “below-average scorers” represented an ingroup for participants who receivedfailure feedback, these findings represent ingroup favouritism.6Furthermore, since this ingroup favouritism was found to be significantly more pronounced for participants whopossessed high global personal trait self-esteem, these results contradict corollary 2.In a partial replication of Crocker et al. (1987, Expt. 1), Brockner and Chen (1996)found that American participants with high global personal trait self-esteem (measured onthe Revised Janis-Field Self-Esteem Scale; Eagly, 1967, as cited in Brockner & Chen,1996) showed greater discrimination following negative feedback than those with low self-esteem. This interaction between high global personal trait and low specific personal stateself-esteem was not found to be significant for participants from the People’s Republic ofChina, suggesting it to be specific to certain cultures.Seta and Seta (1992) modified Crocker et al.’s (1987, Expt. 1) design such that in aparticipant condition minimal group members received false feedback concerning theaverage personality test performance of the minimal ingroup and outgroup taking part.They were informed either that both groups had failed, both groups had succeeded, the ingroup had failed and the outgroup had succeeded, or the outgroup had failed and theingroup had succeeded. Following this feedback, participants evaluated the test Social Identity Theory 13 performance of the ingroup and outgroup on a set of traits. In an observer condition peopleonly read through the personality test before evaluating the performance of two minimalgroups to which they did not belong. Like Crocker et al.’s (1987, Expt. 1) no feedbackcondition, this control condition provided a measure of baseline (i.e., normative)discrimination against which the degree of competitive discrimination could be gauged. Theglobal personal trait self-esteem (RSES) of all participants was measured prior to testfeedback.Ingroup and outgroup evaluations did not differ significantly following either ingroupfailure or outgroup success feedback for participants, but they did differ significantly and ina direction consistent with each type of feedback for observers. Hence, a rather passive,but viable, form of discrimination occurred among participants compared with observers.Participants who possessed high personal self-esteem exaggerated the differencebetween their evaluative ratings of the ingroup and outgroup significantly in the condition where the ingroup succeeded and the outgroup failed. In contrast, participants whopossessed low personal self-esteem seemed to exhibit an outgroup-favouring bias underthese conditions. These results, which were replicated in a subsequent study (Seta & Seta,1996), are contrary to corollary 2, which predicts that individuals with low self-esteem willshow the greatest polarization of ratings in favour of the ingroup.Crocker and Luhtanen (1990) also ran an experiment similar to Crocker et al. (1987,Expt. 1) and found significant discrimination between minimal group members, but nosignificant relationship between this discrimination and either pretest global personal traitself-esteem (RSES) or global social trait self-esteem (the private subscale of the CSES).Sidanius et al. (1994) conducted a minimal groups experiment in which participantswere randomly divided into “overestimators” and “underestimators”, ostensibly on the basisof their performance on a dot-estimation task. Participants then rated each group along thedimensions of “able”, “intelligent”, “stupid”, and “incompetent” as well as rating how willing they were to engage in cooperative or competitive tasks with the outgroup in the future.Pretest global personal trait self-esteem (RSES) was measured prior to categorization.It was found that “the higher the subjects’ level of trait self-esteem, the greater thedifference in perceived competence between ingroup and outgroup [in favour of theingroup] and the less willing subjects were to cooperate than to compete with the outgroup”(p. 161). Again, these results would appear to be the opposite of those predicted bycorollary 2.Two studies of competitive discrimination report evidence that people with low globalpersonal trait self-esteem do not discriminate more than those with high self-esteem (J.Brown et al., 1988, Expt. 1; Crocker & Schwartz, 1985). Crocker and Schwartz (1985) havefound that people with low trait self-esteem report more negative evaluations of others ingeneral but do not engage in intergroup discrimination more than those with high self-esteem. Participants were asked to complete a measure of global personal trait self-esteem (RSES) before being categorized as minimal group members via the lottery ticketprocedure (see Crocker et al., 1987, Expt. 1). Participants then indicated their expectationsabout the personalities of ingroup and outgroup members on five positive and five negativeevaluative traits.Intergroup discrimination only occurred on positive evaluative traits and was notassociated with self-esteem. Instead, participants with low self-esteem rated both ingroupand outgroup members more negatively than did those with high self-esteem. Similarresults have been obtained by Crocker et al. (1987, Expt. 1) and Crocker and Luhtanen(1990).J. Brown et al. (1988, Expt. 1) provide evidence to suggest that people with differentlevels of pretest global personal trait self-esteem differ in the type, rather than the extent, ofdiscrimination that they display. Participants completed a measure of global personal traitself-esteem (the Texas Social Behaviour Inventory, see previous) before being categorized as “overestimators” or “underestimators” using the dot-estimation procedure (see Sidaniuset al., 1994). Half of the underestimator group and half of the overestimator group were Social Identity Theory 14 then taken into a different room. Participants in each room completed a brainstorming taskduring which their group had to suggest different applications for a fly swatter. They thenrated solutions to this task provided by the ingroup and outgroup from only one of therooms. This meant that participants rated ingroup solutions to which they had personallycontributed in one room, while in the other room participants rated ingroup solutions towhich they had not contributed. Hence, the design consisted of a 2 (origin of target solution:same room vs. different room) x 2 (target group solution: ingroup vs. outgroup) mixeddesign with repeated measures on the last factor.People who had high self-esteem engaged in significant discrimination in the sameroom condition but not in the different room condition. In other words, people with high self-esteem discriminated between ingroup and outgroup solutions when they had personallycontributed to the ingroup’s solutions but not when they had not contributed to thesesolutions. The opposite pattern was found for people who had low self-esteem: Significant discrimination occurred in the different room condition, but not in the same room condition.Furthermore, these results were due to significant differences in ingroup favouritism forpeople with high self-esteem and significant differences in outgroup derogation for peoplewith low self-esteem.The researchers concluded that people with high self-esteem seem to preferrelatively direct methods of self-enhancement (i.e., ingroup favouritism and discriminationdirectly linked to the self) whereas those with low self-esteem tend to opt for more indirectmethods (i.e., outgroup derogation and discrimination that is only tacitly linked to the self).This study should be held as being unsupportive of corollary 2, which implies a cleardistinction between individuals with high and low pretest self-esteem in terms of the degreeof discrimination they subsequently display.Experiments Addressing Corollaries 1 and 2 SimultaneouslyFour studies of competitive discrimination have tested the self-esteem hypothesis in its entirety (Hogg & Morkans, 1989, as cited in Hogg & Abrams, 1990; Hogg et al., 1986,Expt. 2; Hogg & Sunderland, 1991; Lemyre & Smith, 1985). Lemyre and Smith (1985)designed the most carefully controlled test of the self-esteem hypothesis to date. Theirminimal groups experiment controlled for the effects of categorization per se, points-distribution per se, and any interaction between the two in terms of increased groupsalience.The core of the experiment consisted of a 2 (categorization: present vs. absent) x 2(points-distribution task: interpersonal vs. intergroup) between-subjects design.Categorization was operationalized using the lottery ticket procedure. Preand post-testglobal personal state self-esteem were measured immediately before and afterdiscrimination in a between-subjects design using a composite scale consisting of theRSES, JBF, the Self-Concept Inventory, a single direct measure of self-esteem, and a half-length version of the TST. A regression analysis on a composite index of all five self-esteem scales showed thatthe posttest self-esteem of categorized participants was significantly predicted byintergroup discrimination, adding support to corollary 1. Contrary to corollary 2, however, asimilar regression analysis revealed no significant relationship between categorizedparticipants’ pretest self-esteem and their subsequent discrimination.Importantly, pretest self-esteem was found to be significantly lower for categorizedparticipants than for uncategorized participants on all the self-esteem measures apart fromthe Self-Concept Inventory and the direct single item, suggesting that categorization per sedepressed self-esteem (see previous). Furthermore, although the posttest self-esteem ofcategorized participants who had discriminated was found to be significantly higher thantheir pretest self-esteem, it was at an equivalent level to the pretest self-esteem ofuncategorized participants. Taken together, this evidence can be taken to support corollary1, but with an important caveat: Intergroup discrimination does seem to elevate self-esteem relative to its level immediately after minimal social categorization, but only to the level itwas at immediately before this categorization. Social Identity Theory 15 Hogg and Sunderland (1991) and Hogg and Morkans (1989, as cited in Hogg &Abrams, 1990) have conducted two very similar studies. Hogg and Sunderland (1991)manipulated specific personal state self-esteem across participants using bogus feedbackon a word association task to produce success and failure conditions. Minimal groupmembers (categorized on an explicitly random basis) and uncategorized participants thencompleted an intergroup points-distribution task. Global personal state self-esteem (JBF)was measured immediately before and after this task using a between-subjects design.In support of corollary 2, categorized participants who received negative feedbackabout their performance on the word association task exhibited significantly greaterdiscrimination than those who received positive feedback. However, contrary to corollary 1,“greater intergroup discrimination was not associated with higher post-test levels and/orsignificant increase [sic] in transitory self-esteem” (p. 58).In contrast to this study, Hogg and Morkans (1989, as cited in Hogg & Abrams, 1990) found no significant difference in discrimination between success and failure conditions.There was a slight tendency for posttest global personal state self-esteem (JBF) to behigher under failure than success conditions, however, since this difference was notaccompanied by differential levels of discrimination, it cannot be taken as evidence infavour of corollary 1 (see Hogg & Abrams, 1990).Finally, Hogg et al. (1986, Expt. 2) found no evidence for corollary 2 and limitedevidence for corollary 1. In this minimal groups experiment, pretest global personal traitself-esteem and posttest global personal state self-esteem were measured within-subjectsusing TST, JBF, and RSES.Contrary to corollary 2, high and low pretest trait self-esteem individuals did not differsignificantly in the extent of their discrimination. Only participants with low pretest trait self-esteem reported significantly higher posttest state self-esteem than participants in a controlcondition who had not discriminated. Participants with high pretest trait self-esteem did not differ significantly from those in the control condition in terms of their posttest state self-esteem. This evidence may be treated as being supportive of corollary 1 on theunderstanding that it only relates to certain types of individuals, namely people with lowpretest trait self-esteem.SummaryTable 2 sets out the supportive and unsupportive competitive discrimination evidencefor each corollary of the self-esteem hypothesis, grouping studies according to the type ofself-esteem that was measured or manipulated.79 experiments are supportive and 3 areunsupportive of corollary 1, whereas 3 experiments are supportive and 16 are unsupportiveof corollary 2. This lead us to the cursory conclusion that competitive discriminationenhances self-esteem but is not motivated by depressed self-esteem.-Insert Table 2 about here --Evidence based on Competitive and/or Normative Discrimination Corollary 1Supportive Evidence. Turner and Spriggs (1982, as cited in Lemyre & Smith, 1985)conducted a minimal groups experiment in which cooperative and competitive instructionswere manipulated orthogonally with interpersonal and intergroup points-distribution in abetween-subjects design. Posttest global personal trait self-esteem was measured usingthe TST, JBF, and RSES.As expected, the greatest discrimination occurred in the condition containingcompetitive instructions and intergroup points-distribution. Furthermore, self-esteem wassignificantly higher on the RSE and JBF under competitive conditions than undercooperative conditions.As Lemyre and Smith (1985) note, it is difficult to draw any firm conclusions from thisstudy because the competitive-cooperative manipulation confounds the experimentaldesign: Self-esteem may have been elevated by either intergroup discrimination per se or by the effects of competitive instructions independent of discrimination. It is also possiblethat the competitive instructions interacted with social categorization to produce normative Social Identity Theory 16 discrimination, and this may have elevated self-esteem independent of competitivediscrimination. Aside from these criticisms, this evidence can be read as being consistentwith corollary 1.Further support for corollary 1 comes from experiments employing more meaningfulsocial categorizations. Hewstone et al. (1993, Expt. 2) found significant intergroupdiscrimination between Muslim (majority) and Hindu (minority) participants on a series ofnonstereotypical evaluative traits. Furthermore, a regression analysis of posttest globalpersonal state self-esteem (JBF) on target group evaluations found that, consistent withcorollary 1, self-esteem was relatively enhanced by higher evaluations of the religiousingroup and lower evaluations of the religious outgroup.A relatively subtle test of corollary 1 has been proposed by Hewstone (1990). Drawingfrom Pettigrew (1979), it is predicted that the attribution of positive ingroup behavior andnegative outgroup behavior to internal causes serves to enhance self-esteem. Conversely, the attribution of negative ingroup behavior and positive outgroup behavior to externalcauses is thought to protect self-esteem.In a test of these predictions, Islam and Hewstone (1993, Expt. 2) asked Muslimstudents to attribute the cause of an actor’s behavior in a series of vignettes using a 2(actor’s religion: ingroup vs. outgroup) x 2 (actor’s nationality: ingroup vs. outgroup) x 2(actor’s behavior: positive vs. negative) between-subjects design. This created fourdifferent group memberships for the actor (double ingroup, double outgroup, and twocrossed groups; for a discussion, see Hewstone et al., 1993). Posttest global personal stateself-esteem (JBF) was measured immediately after these attributions.A regression analysis found that internal attributions made for the positive behavior ofdouble ingroup members significantly predicted high self-esteem. In addition, externalattributions made for the positive behavior of one of the two crossed groups also predictedhigh self-esteem. Assuming, in line with social identity theory, that this crossed group was perceived to be an outgroup (for a discussion, see R. Brown & Turner, 1979), this evidenceseems congruent with corollary 1.Hogg and Turner (1985b) categorized participants according to perceivedinterpersonal similarity and group distinctiveness and then asked them to take part in anintergroup points-distribution task. A principal component factor analysis found that posttestself-esteem (measured using a single item, presumably addressing global personal traitself-esteem) emerged as a weak factor that loaded on intergroup discrimination, providingsupport for corollary 1.Finally, three real group studies that focus on corollary 1 have found limitedsupportive evidence (Hunter et al., 1996; Kelly, 1988; Vanbeseleare, 1996). Kelly (1988)measured intergroup differentiation between the supporters of various political parties interms of perceived intragroup heterogeneity in favour of the ingroup, sympathy with theoutgroup’s views, and desire for close contact with the outgroup. Posttest global personal self-esteem was measured using the TST, JBF, and RSES (presumably in their trait forms).No significant correlations occurred between intergroup differentiation and self-esteem as measured by either the TST or JBF. Data from the RSES indicated that therewere significant positive correlations between the degree of perceived intragroupheterogeneity favouring the ingroup and self-esteem for Conservative supporters. Similarcorrelations occurred between all three measures of intergroup differentiation and theRSES for Social Democratic Party supporters.This evidence can only be taken as limited support for corollary 1 because evidencewas found for the supporters of some political groups (i.e., Conservative and SocialDemocratic Party) but not for others (e.g., Labour Party, Liberal Party, Communist Party),suggesting that the results may only apply to certain types of supporter or the normativeprescriptions of the party that they support.Hunter et al. (1996) asked Protestant (majority) and Catholic (minority) Northern Irish schoolchildren to rate members of each religious group on a series of evaluative traits. Pre-and post-test specific and global personal state self-esteem (the Self-Description Social Identity Theory 17 Questionnaire III, see previous) were measured using a within-subjects design. Significantintergroup discrimination occurred, and participants tended to show an increase in specificaspects of their personal self-esteem (e.g., the domains of academic ability, verbal ability,honesty, physical appearance, parental relations, and religion). Notably, this increase wasnot apparent using a global self-esteem subscale (consisting of six items from the RSES),supporting the researchers’ contention that only specific aspects of self-esteem aresusceptible to change following discrimination.Hunter et al.’s (1996) results are questionable for two reasons. First, participants inthe control group, who were not given the opportunity to discriminate, actually showed asignificant decrease in some aspects of their specific self-esteem (physical ability and sexrelations), and this casts doubt on the reliability of the findings in the experimentalcondition. The reduction of self-esteem in the control group is not particularly surprisingwhen one considers that pretest state self-esteem was measured three weeks before the main experiment. By definition, state self-esteem scales are more sensitive to situationalvariables than trait scales and are not supposed to have good test-retest reliability. Hence,any number of extraneous variables could have affected state self-esteem either duringpretest measurement or immediately before discrimination.Second, the self-esteem changes reported above were found only among Catholicchildren. This is particularly worrying because Protestant children exhibited roughly twice asmuch discrimination as Catholic children in the experimental condition and so, according tocorollary 1, would be expected to report the greatest difference in self-esteem.Finally, Vanbeselaere (1996) measured the global personal state self-esteem (JBF)of high, low, and neutral status group members after they had evaluated groupperformances on a line-length estimation task. Intergroup discrimination correlatedsignificantly with the self-esteem of people in the high status group but not with that ofpeople in the neutral or low status groups. Consequently, this study can only be said to find support for corollary 1 under certain conditions.Unsupportive Evidence. Vickers et al. (1985, as cited in Abrams & Hogg, 1988) foundthat participants who acted against a prescriptive norm of cooperation by engaging inminimal intergroup discrimination reported lower posttest global personal self-esteem (JBF,presumably in its trait form) than those who cooperated. The researchers suggest that thisreduction may have resulted from norm violation per se. Alternatively, participants whocooperated may have experienced elevated self-esteem as a result of complying with therelevant norm. Notwithstanding these possibilities, this evidence can be read as being indirect opposition to corollary 1.Two real group studies addressing corollary 1 have found unsupportive evidence(Hunter et al., 1993; Nascimeto-Schulze, 1993). Hunter et al. (1993) asked Protestant andCatholic schoolchildren to make attributions about the cause of an actor’s behavior in aseries of vignettes. A 2 (participant’s religion: Protestant vs. Catholic) x 2 (actor’s religion: ingroup vs. outgroup) x 2 (actor’s behavior: positive vs. negative) mixed design wasemployed with repeated measures on the last two factors. Posttest global personal traitself-esteem was measured using the general-self subscale of the Self-DescriptionQuestionnaire III.Catholic children made significantly more internal attributions for positive ingroupbehavior than for positive outgroup behavior and significantly more internal attributions fornegative outgroup behavior than for negative ingroup behavior. Protestant children’sattributions did not vary significantly as a function of target group. Self-esteem was notsignificantly associated with attributional differentiation either across or within groups. Thisevidence is therefore unsupportive of corollary 1.Finally, Nascimento-Schulze (1993) asked Brazilian bank clerks to compare theirbanks on a series of evaluative dimensions before completing two measures of globalpersonal trait self-esteem (RSES and TST). Contrary to corollary 1, no significant differences were found between the self-esteem of participants who discriminated and theself-esteem of those who did not discriminate. Social Identity Theory 18 Corollary 2Supportive Evidence. Long and her colleagues (Long & Spears, 1995, as cited inLong & Spears, 1997; Long et al., 1994) have investigated the relationship betweenpersonal and social self-esteem in terms of subsequent discrimination. Long et al. (1994;see also Long & Spears, 1997) asked Dutch students to complete three measures of globalpersonal trait self-esteem (JBF, RSES, and the Texas Social Behaviour Inventory) and ameasure of specific social trait self-esteem (CSES, modified to be specific to Dutchpsychology students). The students then took part in a brainstorming task during whichthey were asked to think of an advertising slogan encouraging the use of phosphate-freedetergents. They then evaluated one of their own group’s solutions, a solution produced byanother group of Dutch students (i.e., a second ingroup) and a solution produced by anoutgroup (German or Swiss students, depending on condition).Contrary to corollary 2, participants with high personal self-esteem engaged in significant discrimination whereas participants with low personal self-esteem did not.However, personal and social self-esteem interacted such that participants with highpersonal and low social self-esteem displayed greater discrimination than those whopossessed the other three possible combinations of personal and social self-esteem. Thisevidence is therefore consistent with that provided by Platow et al. (1997, Expt. 2) insuggesting that personal and social self-esteem are separate constructs and that corollary2 is only valid for social self-esteem when personal self-esteem is also taken into account.Long and Spears (1995, as cited in Long & Spears, 1997) conducted a similar studyto Long et al. (1994). Again, they found that participants who possessed high globalpersonal trait self-esteem (JBF, RSES, and the Texas Social Behaviour Inventory)discriminated significantly more than those with low self-esteem. However, no interactionbetween personal and social self-esteem was obtained. The CSES as a whole did notpredict discrimination, however, participants with low scores on the public subscale of the CSES were found to discriminate significantly more than those with high scores, addingsupport to corollary 2 in terms of specific personal trait self-esteem.Finally, Crocker et al. (1987, Expt. 2) found that pretest global personal trait self-esteem (RSES) interacted marginally with membership in different status sorority groups topredict ingroup favouritism. Congruent with corollary 2, people with low self-esteemexhibited the greatest ingroup favouritism in high status sororities, however, in opposition tocorollary 2, people with high self-esteem showed the greatest ingroup favouritism in lowstatus sororities. Consequently, this evidence only provides limited support for corollary 2.Unsupportive Evidence. Three real group experiments have found no relationshipbetween pretest self-esteem and discrimination (Hogg & Turner, 1987; Ruttenberg et al.,1996; Smith & Tyler, 1997, Expt. 2). Hogg and Turner (1987) found no relationship betweenpretest global personal state self-esteem (measured on 20 behavioral characteristics) andintergroup points-distribution between male and female groups. Interestingly, they found that categorization in terms of gender increased self-esteem for males, but decreased self-esteem for females (cf. Lemyre & Smith, 1985).Ruttenberg et al. (1996) asked Jewish and Arab participants to complete a measureof global social trait self-esteem (CSES) followed by a revised version of Levinson andSanford’s (1944, as cited in Ruttenberg et al.) Anti-Semitism Scale. They then rated thehumorous content of a selection of stereotype-based jokes and cartoons.No significant correlations occurred between self-esteem and the measure ofprejudice or discrimination for either ethnic group. Following a regression analysis, theauthors concluded that global social trait self-esteem “was not a strong predictor ofattitudes toward the outgroup or of reactions to humour disparaging the outgroup” (p. 218).Smith and Tyler (1997, Expt. 2) found that pretest global personal trait self-esteem(RSES) correlated negatively with negative ratings of both ingroup and outgroup membersof campus sororities. No significant correlations were found with positive ratings of either ingroup or outgroup. This evidence converges with that provided by minimal group studiesshowing that people with low self-esteem have more of a negative attitude towards others Social Identity Theory 19 in general, but do not discriminate more than those with high self-esteem (Crocker &Schwartz, 1985; Crocker & Luhtanen, 1990; Crocker et al., 1987, Expt. 1).Smith and Tyler (1997, Expt. 2) also found that pretest global social trait self-esteem(CSES) showed significant positive correlations with positive ratings of the ingroup but notwith positive ratings of the outgroup or negative ratings of either inor out-group. Thisevidence only suggests that people with high social self-esteem regard ingroup membersas possessing more positive qualities than do people with low self-esteem and does not,therefore, provide supportive evidence for corollary 2.Experiments Addressing Corollaries 1 and 2 SimultaneouslyFour studies have attempted to test both corollaries of the self-esteem hypothesisusing real groups (Branscombe & Wann, 1994; Meindl & Lerner, 1984, Expt. 2; Verkuyten,1997; Wagner et al., 1986). Branscombe and Wann (1994) found limited support for bothcorollaries of the self-esteem hypothesis. American participants watched a film clip of a boxing match between an American and a Soviet fighter. In an identity-threateningcondition participants were shown a clip in which the American fighter lost the match and inan unthreatening condition they were shown a clip in which he won. Pretesting four weeksprior to this manipulation established the extent to which each participant identified withAmerica and allowed the construction of a 2 (identification: high vs. low) x 2 (outcome ofboxing match: identity-threatening vs. unthreatening) between-subjects design.Following the film clip, participants were provided with the opportunity to derogate avariety of outgroups, including Russians. Preand post-test global social trait self-esteemwere measured in a within-subjects design using half the items from the private subscale ofthe CSES immediately before outgroup derogation and the other half immediately afterderogation.High identifiers showed significantly lower self-esteem after watching the Americanboxer lose the match than they did after he won, whereas low identifiers did not differ significantly in their level of self-esteem between these conditions. In addition, highidentifiers in the identity-threatening condition derogated the Russian outgroup significantlymore than those in the unthreatening condition. Again, this difference was not significantamong low identifiers. Hence, consistent with social identity theory, outgroup derogationwas promoted by a threat to identity-contingent self-esteem when the relevant identity washighly valued. The crucial question is whether this derogation led to an increase in self-esteem.Analyses using a median split on Russian derogation scores found that highderogators reported a significantly greater positive shift from preto post-test self-esteemthan low derogators in the identity-threatening condition. A similar shift was not apparent inthe unthreatening condition. It therefore appears that outgroup derogation leads to anincrease in global social self-esteem for people who have a high level of identification withthe ingroup and who have undergone a threat to that identification. Again, this is consistent with social identity theory’s assumptions.Wagner et al. (1986) found support for corollary 2 but not for corollary 1. An audiotape recording of a debate about television was played to undergraduate law students, whowere then informed that the discussion ability was of a medium standard. In a low statuscondition it was then mentioned that previous research had shown the discussion ability oflaw students to be worse than that of economics students. In a neutral conditionparticipants were informed that previous research had compared the discussion abilities ofdifferent departments but were not told about the results of this research. Finally, in acontrol condition no mention of previous research was made.Participants in the low status and neutral conditions evaluated the discussion ability ofmedical students, while those in the control condition evaluated the discussion ability ofstudents without being informed of their departmental affiliation. Preand post-test self-esteem were measured immediately before and after derogation in a between-subjects design using a semantic differential scale and a shortened and modified version ofCoopersmith’s (1967) self-esteem scale. Social Identity Theory 20 Four of the items from Coopersmith’s scale, concerning intellectual achievement atuniversity, were significantly different from the rest of the scale and from the semanticdifferential scale. These were analysed separately as achievement self-esteem (i.e.,specific personal trait self-esteem). The remaining Coopersmith items were combined withthe semantic differential scale to provide a measure of global personal trait self-esteem.Pretest achievement self-esteem was significantly lower in both the low status andneutral conditions compared with control conditions. It is therefore debatable whether thestatus manipulation contributed to the difference between the low status and controlconditions or whether this difference was simply the result of the categorization-depressioneffect that seems to have occurred in the neutral conditions. Whatever the case, nosignificant differences in outgroup evaluation appeared between the three conditions.However, pretest achievement self-esteem was significantly negatively correlated withoutgroup evaluation across conditions, lending some support to corollary 2. Contrary to corollary 1, no significant differences in either type of posttest self-esteem occurred frompreto post-test measures.One problem with Wagner et al.’s (1986) study is that the low status of the ingroup(law students) relative to one outgroup (economics students) was expected to affectrelations with regard to a second outgroup (medicine students). Devaluation of a secondoutgroup may not necessarily restore self-esteem that is contingent upon the statusrelations between the ingroup and a first outgroup. Indeed, Branscombe and Wann (1994)found that discriminating against outgroups that were irrelevant to the original identity threatactually caused a decrease in self-esteem.In defence of their design, Wagner et al. argue that because the ingroup’s statusrelationship with the second outgroup is more ambiguous than that between the ingroupand the first outgroup, it will be perceived as being more unstable and amenable to socialchange. Hence, their design represents an attempt to protect against the confounding influence of normative discrimination. As they put it, “the presence of a second outgroupwhich has no relevant relationship to the status hierarchy of the ingroup and the firstoutgroup, may provide an opportunity to obtain a ‘cleaner’ measure of ingroup bias” (p. 23).Further research needs to be conducted to ascertain the viability of this experimentaldesign.In contrast to Wagner et al. (1986), Maass et al. (1996, Expt. 1) have found supportfor corollary 1, but not for corollary 2. Intergroup discrimination was measured in terms ofthe degree to which different linguistic categories were employed. Northern Italian huntersand environmentalists were asked to select phrases that best described the actions ofcartoon representations of members of each group. Threat was manipulated prior to thistask such that half the participants received an antagonistic message from the outgroupand half received a cooperative message. Preand post-test measures of global personaltrait self-esteem (4 items from the RSE) and specific social trait self-esteem (the private and identity subscales of the CSES) were taken before the threat manipulation and afterdiscrimination using a between-subjects design.In line with research on intergroup attribution (e.g., Islam & Hewstone, 1993, Expt. 2)and previous research on linguistic intergroup bias, participants selected abstract (asopposed to concrete) phrases to describe the positive actions of ingroup memberssignificantly more than the positive actions of outgroup members (for a review, see Maass& Acuri, 1996). They also selected abstract phrases to describe the negative actions ofoutgroup members significantly more than the negative actions of ingroup members. Thisbias was found to be significantly more pronounced in the identity-threatening conditions.In support of corollary 1, linguistic intergroup discrimination correlated positively withboth types of posttest self-esteem, however, significant increases in self-esteem from pre-to post-test were only found in the identity-threatening condition in terms of specific socialtrait self-esteem. No significant increases in global personal trait self-esteem were identified from preto post-test. Contrary to corollary 2, no significant correlations were identifiedbetween linguistic discrimination and either type of pretest self-esteem. Social Identity Theory 21 In a second experiment employing Swiss students of either northern or southernItalian origin, Maass et al. (1996, Expt. 2) found no significant relationship between eitherpreor post-test specific social trait self-esteem and linguistic intergroup discrimination.Finally, two real group experiments find no support for either corollary of the self-esteem hypothesis (Meindl & Lerner, 1984, Expt. 2; Verkuyten, 1997). In contrast to socialidentity theory, Meindl and Lerner (1984) have suggested that the experience of personalfailure, and the lowered personal state self-esteem that results, engenders both ingroupand outgroup favouritism in the service of self-enhancement. They propose that outgroupfavouritism enhances an ingroup’s positive distinctiveness because any group that is in aposition to favour an outgroup is normally regarded as being superior to that outgroup.In tests of this hypothesis, Meindl and Lerner (1984, Expt.’s 1 & 2) manipulated thespecific personal state self-esteem of English-speaking Canadian students to form twoconditions. In a low self-esteem condition participants were asked to fetch a chair to sit on during the experiment. The chair was rigged to cause an “accident” involving a graduatestudent’s computer-card collection, which participants believed contained carefully arrangeddissertation data. In a control condition participants retrieved the chair without causing theaccident. Following this manipulation, participants indicated the extent to which they agreedwith aggressive and benevolent attitude positions concerning Quebec’s relationship with therest of Canada.Participants in the low self-esteem condition agreed significantly more with bothaggressive and benevolent attitude positions than those in the control condition. Theseresults seem to suggest that both ingroup and outgroup favouritism are reactions tolowered self-esteem and so cannot be taken to support corollary 2, which only predictsintergroup discrimination in favour of the ingroup as a response to low self-esteem.Meindl and Lerner (1984, Expt. 2) included a measure of posttest global personal traitself-esteem (RSES) in their second experiment. Contrary to corollary 1, posttest selfesteem was not found to be associated with evaluations of either aggressive or benevolentattitudes.Verkuyten (1997) measured the pretest specific social trait self-esteem of Dutchstudents (using the private subscale of the CSES, modified to be specific to Dutchstudents) before asking them to complete three social distance scales concerning thedegree of contact desired between Dutch and Moroccan students. Posttest global personalstate self-esteem was then measured (JBF).Pretest self-esteem was not correlated with subsequent group evaluations and, in aregression analysis, neither ingroup nor outgroup evaluations significantly predictedposttest self-esteem. Consequently, this evidence cannot be taken to support eithercorollary of the self-esteem hypothesis.In separate regression analyses it was found that positive ingroup evaluationspredicted positive posttest self-esteem for people with high pretest self-esteem, whereas negative outgroup evaluations predicted positive posttest self-esteem for people with lowpretest self-esteem. This evidence therefore seems congruent with J. Brown et al.’s (1988)proposal that people with high self-esteem tend to enhance their positive self-regardthrough ingroup favouritism whereas those with low self-esteem enhance their self-regardvia outgroup derogation.SummaryTable 3 sets out the supportive and unsupportive competitive and/or normativediscrimination evidence for each corollary of the self-esteem hypothesis, grouping studiesaccording to the type of self-esteem that was measured. 9 experiments are supportive and10 are unsupportive of corollary 1, whereas 4 experiments are supportive and 12 areunsupportive of corollary 2.-Insert Table 3 about here --Conclusions In line with previous work by Hogg and Abrams (1990) and Crocker et al. (1993), thecurrent review finds more evidence in favour of corollary 1 than corollary 2. Overall, 18 Social Identity Theory 22 experiments are supportive and 13 are unsupportive of corollary 1, whereas 7 experimentsare supportive and 28 are unsupportive of corollary 2. Furthermore, only one study(Branscombe & Wann, 1994) out of the eight that address the self-esteem hypothesis in itsentirety finds support for both corollaries.From this exhaustive review of the literature, it would seem as if intergroupdiscrimination leads to an increase in self-esteem but is not motivated by a need for self-esteem. This is contrary to social identity theory, which relies on a motivational explanationof discrimination. This explanation fails unless either both corollaries of the self-esteemhypothesis receive convincing support or corollary 2 receives support and some otherexplanation is offered to account for the redundancy of corollary 1. Consequently, it can beconcluded that the majority of evidence does not support social identity theory’s self-esteem hypothesis in its full and unqualified form.It is of interest to note that, contrary to corollary 2, nine studies found individuals with high global personal trait self-esteem to be the most discriminatory (Abrams, 1982, 1983,as cited in Abrams & Hogg, 1988, and Hogg & Abrams, 1990; Brockner & Chen, 1996;Crocker et al., 1987; Long & Spears, 1995, as cited in Long & Spears, 1997; Long et al.,1994; Seta & Seta, 1992, 1996; Sidanius et al., 1994). This evidence is congruent with theidea that individuals who possess generally low self-esteem lack the self-confidence toengage in self-enhancement via intergroup discrimination (see Abrams, 1992; Abrams &Hogg, 1988; Hogg & Abrams, 1990). However, it remains uncertain as to whether thishypothesis applies equally well to more specific facets of self-esteem (cf. Platow et al.,1997, Expt. 2).Table 4 compares the number of experiments that find supportive and unsupportiveevidence for each corollary in terms of the different types of self-esteem that weremeasured or manipulated. It can be seen that, for both corollaries, specific self-esteemprovides a higher proportion of supportive evidence than global self-esteem, social selfesteem provides a higher proportion than personal self-esteem, and state self-esteemprovides a higher proportion than trait self-esteem. This evaluation adds some support toour contention that the self-esteem hypothesis may be more applicable to specific, social,and state forms of self-esteem than to global, personal, and trait forms.-Insert Table 4 about here --A tentative reformulation of the self-esteem hypothesis can be stated as follows:1. Successful competitive intergroup discrimination enhances social identityand thus increases the specific state self-esteem associated with that identity.2. Low specific social state self-esteem promotes competitive intergroupdiscrimination because of a need for this type of self-esteem.Only two studies make any attempt to measure specific social state self-esteem(Gagnon & Bourhis, 1996; Platow et al., 1997, Expt. 2), and it is encouraging to note thatboth find supportive evidence for the self-esteem hypothesis. Future research should use more specific measures of self-esteem by either adapting existing scales (e.g., the CSES)or constructing new ones. However, two potential problems should be taken into accountbefore employing measures of specific social state self-esteem.First, it may be difficult to disguise the hypothesized connection between self-esteemand discrimination from participants. For example, in a between-subjects designparticipants in both preand post-test conditions are likely to discern some relationshipbetween the item “at the moment, I feel good about my group” and their treatment of theingroup. Furthermore, within-subject operationalizations of preand post-test measures, asadvocated by some researchers (Branscombe & Wann, 1994; Hogg et al., 1986, Expt. 2;Hunter et al., 1996; Verkuyten, 1997; cf. Hogg & Sunderland, 1991), are liable toexacerbate this problem.Blatant use of specific social state measures is therefore inadvisable because theyare liable to increase the impact of demand characteristics and self-presentational concerns. Recent work in the area of implicit social cognition seems to offer a way aroundthis problem. For example, Greenwald and Banaji (1995) discuss the possibility of implicit Social Identity Theory 23 measures of self-esteem. Similarly, relatively unobtrusive measures of intergroupdiscrimination have been developed (e.g., Fazio, Jackson, Dunton, and Williams, 1995;Islam & Hewstone, 1993, Expt. 2; Maass et al., 1996, Expt. 1).Second, as Abrams and Hogg (1988) note, “experiments which have reported ingroupbias in terms of trait adjective ratings, affective ratings and performance evaluationratings...are, in effect, directly tapping the relative esteem in which subjects hold their owngroup” (p. 323). In other words, some measures of discrimination may also act asmeasures of specific social state self-esteem. Employing such measures as indices ofdiscrimination to be related to explicit measures of specific social state self-esteemtherefore runs the risk of obtaining artefactual results. Consequently, it would seemappropriate to employ more behavioral measures of discrimination (e.g., intergroup points-or money-distribution) in conjunction with specific social state self-esteem measures.Suggestions for Further Clarification As mentioned previously, one reason for the confusion surrounding empirical supportfor the self-esteem hypothesis is the vast array of methods and experimental designs thathave been employed in its investigation. Further clarification and elaboration of the self-esteem hypothesis may help to alleviate this problem.First, there remains some ambiguity concerning the exact nature of changes in self-esteem brought about via discrimination. The notion that increments in self-esteem mayserve different functions seems to have been implicitly acknowledged in the literature,where reference is commonly made to achieving, maintaining, and enhancing positive self-esteem.Logically, achieving positive self-esteem consists of raising initially low or neutral self-esteem to a higher level. Maintaining positive self-esteem involves restoring initially highself-esteem following its reduction or retaining its original level throughout an esteem-threatening experience. Finally, enhancing positive self-esteem involves further increasing initially high levels of self-esteem.These distinctions are important because it is possible that discrimination is effectiveat fulfilling some functions more than others. For example, as Lemyre and Smith (1985)propose, it is possible that discrimination only restores self-esteem that has previously beenlowered, rather than bringing about any absolute increment. In addition, some sort of ceilingeffect may prevent the continual enhancement of positive self-esteem beyond a certainlimit.A second ambiguity concerns the distinction between the threat of low self-esteemand the actual experience of low or depressed self-esteem (see Farsides, 1995). Corollary2 of the self-esteem hypothesis seems to cover both eventualities and yet it remains anempirical question as to whether they involve comparable processes or producecomparable results.Third, Farsides (1995) has noted that researchers have neglected to investigate whether successful discrimination enhances self-esteem, as per corollary 1 of the self-esteem hypothesis. Presumably, the distinction between successful and unsuccessfuldiscrimination relates to the results of the discrimination concerned. Unsuccessfuldiscrimination involves the failure of discriminatory acts to bring about the desired positivedistinctiveness. One way to operationalize a clearer test of corollary 1, therefore, would beto include feedback conditions in which participants are informed about the effects that theirintergroup behavior has had upon the positive distinctiveness of each group.Fourth, as Brewer (1979) notes, some researchers do not distinguish betweeningroup favouritism and outgroup derogation so that “it is often ambiguous whether thecomparison rests on enhancement of the in-group, devaluation of the out-group, or both”(p. 321). Brewer (1979) argues that self-enhancement occurs mainly via ingroupfavouritism whereas Wills (1981) suggests that it results more from outgroup derogation.Again, a more precise analysis would facilitate a better understanding of any relationship that might exist between self-esteem and these two forms of discrimination (see J. Brownet al., 1988; Verkuyten, 1997). Social Identity Theory 24 A finer distinction is also required between ingroup identification and social self-esteem. Some researchers include ingroup identification subscales or items as part ofsocial self-esteem scales (e.g., Luhtanen & Crocker’s, 1992, CSES). Others include socialself-esteem subscales or items as part of ingroup identification scales (e.g., Brown,Condor, Mathews, Wade, & Williams, 1986). Considering this blurred distinction, it isunsurprising that research attempting to relate pretest measures of ingroup identification todiscrimination has met with similarly inconclusive results as investigations of the self-esteem hypothesis (for a review, see Hinkle & R. Brown, 1990). Indeed, failure todistinguish between ingroup identification and social self-esteem may have hinderedprogress in both areas of research. After all, social identity theory predicts that pretestidentification will correlate positively with discrimination, but that pretest self-esteem willcorrelate negatively with discrimination. Consequently, single scales that attempt tomeasure both dependent variables simultaneously are in danger of entangling competing effects.It would therefore seem prudent to separate operationalizations of ingroupidentification and social self-esteem. One possibility would be to restrict measures ofidentification to issues of importance and measures of self-esteem to issues of groupevaluation. The need to distinguish between identification and self-esteem is reinforced byresearch suggesting that the self-esteem hypothesis may only be applicable to individualswho possess a high degree of identification with the ingroup (Branscombe & Wann, 1994;Gagnon & Bourhis, 1996).Questions also remain concerning the relationship between the different varieties ofself-esteem that have been discussed. Long and Spears (1997) argue that pretest personaland social self-esteem interact such that people with high personal and low social self-esteem discriminate more than those of the other three possible combinations of personaland social self-esteem. Here, it is argued that low social self-esteem provides both a threat to positive personal identity and the motivation to improve the positivity of social identity.This argument seems to make most sense in terms of global personal trait and specificsocial state self-esteem.In general, individuals who possess high global and low specific self-esteem shouldpossess an additional drive for self-consistency that adds to their need for self-esteem inmotivating self-enhancement. Similarly, individuals who have high trait and low state self-esteem should have a double incentive to engage in self-enhancement compared withthose who possess the other three possible combinations of trait and state self-esteem.Adding these predictions together, it might be expected that individuals with high globalpersonal trait and low specific social state self-esteem will possess the greatest motivationto engage in intergroup discrimination compared with those possessing the other sixcombinations of self-esteem.Finally, the distinction between competitive and normative discrimination deserves greater attention. Ideally, the self-esteem hypothesis should be tested in the minimal groupparadigm in order to avoid the confounding effects of normative discrimination. Someresearchers argue against minimal groups in this area of research, suggesting that self-esteem is likely to have stronger links with real group memberships than transient andrelatively inconsequential group memberships (e.g., Crocker et al., 1987; Crocker &Luhtanen, 1990; cf. Nascimento-Schulze, 1993). Although this may be the case in terms ofglobal personal trait self-esteem, it does not seem unreasonable to expect measures ofspecific social state self-esteem to be sensitive enough to relate to the social identitiesprovided by minimal groups.This said, research involving real groups plays an important role in providingecological validity for social identity theory and so should never be discounted. Future real-group experiments should be designed to maximize the perception of relevant statushierarchies as being insecure and amenable to social change (e.g., Wagner et al., 1986) and new techniques should be used to account for any normative discrimination that might Social Identity Theory 25 occur alongside competitive discrimination (e.g., Crocker et al., 1987, Expt. 1; Seta & Seta,1992, 1996).The controvertible findings regarding the self-esteem hypothesis have led someresearchers to conclude that it has been overimplicated in social identity theory’s analysis(e.g., Abrams & Hogg, 1988; Hogg & Abrams, 1990). Extensions of social identity theoryhave tended to either ignore the self-esteem hypothesis (e.g., Turner et al.’s, 1987, self-categorization theory) or to curtail its generality and include other motivational processes inits place (e.g., Brewer, 1991, 1993; Hogg & Abrams, 1993). We have argued that the lackof firm empirical support for the self-esteem hypothesis may be due to its impreciseformulation. We have sought to provide some clarification using different types of self-esteem and suggest that future research employs measures of specific social state self-esteem. However, further clarification is required in order to allow a more thoroughinvestigation of the relationship between self-esteem and intergroup discrimination. Social Identity Theory 26 ReferencesAbrams, D. (1992). Processes of social identification. In G. Breakwell (Ed.), Socialpsychology of the self-concept (pp. 57-99). London: Academic Press.Abrams, D., & Hogg, M. A. (1988). Comments on the motivational status of self-esteem in social identity and intergroup discrimination. European Journal of SocialPsychology, 18, 317-334.Berkowitz, N. H. (1994). Evidence that subjects expectancies confound intergroupbias in the Tajfel minimal group paradigm. Personality and Social Psychology Bulletin, 20(2), 184-195.Blascovich, J., & Tomaka, J. (1991). Measures of self-esteem. In J. P. Robinson, P.R. Shaver, & L. S. Wrightsman (Eds.), Measures of personality and social psychologicalattitudes (pp. 115-160). New York: Academic Press.Bourhis, R. Y. (1994). Power, gender and intergroup discrimination: Some minimal group studies. In M. Zanna & J. Olson (Eds.), The psychology of prejudice: The 7th Ontariosymposium on personality and social psychology (pp. 171-208). Hillsdale, NJ: Erlbaum.Branscombe, N. R., & Wann, D. L. (1994). Collective self-esteem consequences ofoutgroup derogation when a valued social identity is on trial. European Journal of SocialPsychology, 24 (6), 641-657.Breckler, S., & Greenwald, A. G. (1986). Motivational facets of the self. In R. L.Sorrentino & E. T. Higgins (Eds.), Handbook of motivation and cognition (pp. 145-164).New York: Guilford.Brewer, M. B. (1979). Ingroup bias in the minimal intergroup situation: A cognitivemotivational analysis. Psychological Bulletin, 86, 307-324.Brewer, M. B. (1991). The social self: Being the same and different at the same time.Personality and Social Psychology Bulletin, 17, 475-482.Brewer, M. B. (1993). Social identity, distinctiveness, and in-group homogeneity. Social Cognition, 11 (1), 150-164.Brewer, M. B., & Miller, N. (1996). Intergroup relations. Buckingham: Open UniversityPress.Brockner, J., & Chen, Y. (1996). The moderating roles of self-esteem and self-construal in reaction to a threat to the self: Evidence from the People’s Republic of Chinaand the United States. Journal of Personality and Social Psychology, 71 (3), 603-615.Brown, J., Collins, R. L., & Schmidt, G. W. (1988). Self-esteem and direct vs. indirectforms of self-enhancement. Journal of Personality and Social Psychology, 55, 445-453.Brown, R. (1986). Social psychology: The second edition. New York: The Free Press.Brown, R., Condor, S., Mathews, A., Wade, G., & Williams, J. (1986). Explainingintergroup differentiation. Journal of Occupational Psychology, 59, 273-286.Brown, R. J., & Turner, J. C. (1979). The criss-cross categorization effect inintergroup discrimination. British Journal of Social and Clinical Psychology, 18, 371-383. Campbell, D. T. (1965). Ethnocentric and other altruistic motives. In D. Levine (Ed.),Nebraska symposium on motivation. Lincoln, NE: University of Nebraska Press.Chin, M. G., & McClintock, C. G. (1993). The effects of intergroup discrimination andsocial values on levels of self-esteem in the minimal group paradigm. European Journal ofSocial Psychology, 23, 63-75.Coopersmith, S. (1967). The antecedents of self-esteem. San Francisco: Freeman.Crocker, J., Blaine, B., & Luhtanen, R. (1993). Prejudice, intergroup behaviour andself-esteem: Enhancement and protection motives. In M. A. Hogg & D. Abrams (Eds.),Group motivation: Social psychological perspectives (pp. 52-67). London: HarvesterWheatsheaf.Crocker, J., & Luhtanen, R. (1990). Collective self-esteem and ingroup bias. Journalof Personality and Social Psychology, 58, 60-67.Crocker, J., Luhtanen, R., Blaine, B., & Broadnax, S. (1994). Collective self-esteem and psychological well-being among White, Black and Asian college students. Personalityand Social Psychology Bulletin, 20, 503-513. Social Identity Theory 27 Crocker, J., & Major, B. (1989). Social stigma and self-esteem: The self-protectiveproperties of stigma. Psychological Review, 96, 608-630.Crocker, J., & Schwartz, I. (1985). Prejudice and ingroup favoritism in a minimalintergroup situation: Effects of self-esteem. Personality and Social Psychology Bulletin, 11,379-386.Crocker, J., Thompson, L. J., McGraw, K. M., & Ingerman, C. (1987). Downwardcomparison, prejudice, and evaluations of others: Effects of self-esteem and threat. Journalof Personality and Social Psychology, 52, 907-916.Dutton, K. A., & Brown, J. D. (1997). Global self-esteem and specific self-views asdeterminants of people’s reactions to success and failure. Journal of Personality and SocialPsychology, 73 (1), 139-148.Ethier, K., & Deaux, K. (1990). Hispanics in Ivy: Assessing identity and perceivedthreat. Sex Roles, 22, 427-440. Farsides, T. (1993). Social identity theory: A foundation to build upon, not undermine:A comment on Schiffmann & Wicklund (1992). Theory and Psychology 3 (2), 207-215.Farsides, T. (1995). Why social identity theory’s self-esteem hypothesis has neverbeen tested and how to test it. Paper presented at the Social Section Conference of theBritish Psychology Society, York, September.Fazio, R. H., Jackson, J. R., Dunton, B. C. & Williams, C. J. (1995). Variability inautomatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline?Journal of Personality and Social Psychology, 69 (6), 1013-1027.Finchilescu, G. (1986). Effect of incompatibility between internal and external groupmembership criteria on intergroup behaviour. European Journal of Social Psychology, 16,83-87.Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior. Reading, MA:Addison-Wesley. Gagnon, A., & Bourhis, R. Y. (1996). Discrimination in the minimal group paradigm:Social identity or self-interest? Personality and Social Psychology Bulletin, 22 (12), 1289-1301.Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102 (1), 4-27.Hartstone, M., & Augoustinos, M. (1995). The minimal group paradigm:Categorization into two versus three groups. European Journal of Social Psychology, 15,179-193.Heatherton, T. F., & Polivy, J. (1991). Development and validation of a scale formeasuring state self-esteem. Journal of Personality and Social Psychology, 60, 895-910.Hewstone, M. (1990). The ‘ultimate attribution error’? A review of the literature onintergroup causal attribution. European Journal of Social Psychology, 20, 311-335.Hewstone, M., Islam, M. R., & Judd, C. M. (1993). Models of crossed categorization and intergroup relations. Journal of Personality and Social Psychology, 64, 779-793.Hinkle, S., & Brown, R. J. (1990). Intergroup comparisons and social identity: Somelinks and lacunae. In D. Abrams & M. A. Hogg (Eds.), Social identity theory: Constructiveand critical advances (pp. 48-70). New York: Harvester Wheatsheaf.Hogg, M. A., & Abrams, D. (1988). Social identifications: A social psychology ofintergroup relations and group processes. London: Routledge.Hogg, M. A., & Abrams, D. (1990). Social motivation, self-esteem and social identity.In D. Abrams & M. A. Hogg (Eds.), Social identity theory: Constructive and critical advances(pp. 28-47). New York: Harvester Wheatsheaf.Hogg, M. A., & Abrams, D. (1993). Towards a single-process uncertainty-reductionmodel of social motivation in groups. In M. A. Hogg & Abrams, D. (Eds.), Group motivation:Social psychological perspectives (pp. 173-190). London: Harvester Wheatsheaf.Hogg, M. A., & Sunderland, J. (1991). Self-esteem and intergroup discrimination in the minimal group paradigm. British Journal of Social Psychology, 30, 51-62. Social Identity Theory 28 Hogg, M. A., & Turner, J. C. (1985a). Interpersonal attraction, social identification andpsychological group formation. European Journal of Social Psychology, 15, 51-66.Hogg, M. A., & Turner, J. C. (1985b). When liking begets solidarity: An experiment onthe role of interpersonal attraction in psychological group formation. British Journal ofSocial Psychology, 24, 267-281.Hogg, M. A., & Turner, J. C. (1987). Intergroup behaviour, self-stereotyping and thesalience of social categories. British Journal of Social Psychology, 26, 325-340.Hogg, M. A., Turner, J. C., Nascimento-Schulze, C., & Spriggs, D. (1986). Socialcategorization, intergroup behaviour and self-esteem: Two experiments. Revista dePsicologia Social, 1, 23-38.Hunter, J. A., Platow, M. J., Howard, M. L., & Stringer, M. (1996). Social identity andintergroup evaluative bias: Realistic categories and domain specific self-esteem in a conflictsetting. European Journal of Social Psychology, 26, 631-647. Hunter, J. A., Stringer, M., & Coleman, J. T. (1993). Social explanations and self-esteem in Northern Ireland. Journal of Social Psychology, 5, 643-650.Islam, M. R., & Hewstone, M. (1993). Intergroup attributions and affectiveconsequences in majority and minority groups. Journal of Personality and SocialPsychology, 64, 936-950.Julian, J. W., Bishop, D. W., & Fiedler, F. E. (1966). Quasi-therapeutic effects ofintergroup competition. Journal of Personality and Social Psychology, 3, 321-327.Kelly, C. A. (1988). Intergroup differentiation in a political context. British Journal ofSocial Psychology, 27, 319-332.Kline, P. (1993). The handbook of psychological testing. London: Routledge.Kuhn, M. H., & McPartland, T. S. (1954). An empirical investigation of self attitudes.American Sociological Review, 19, 68-70.Lemyre, L., & Smith, P. (1985). Intergroup discrimination and self-esteem in the minimal group paradigm. Journal of Personality and Social Psychology, 49, 660-670.Long, K. M., & Spears, R. (1997). The self-esteem hypothesis revisited: Opposingeffects of personal and collective levels of esteem on intergroup differentiation. In R.Spears, P. Oakes, N. Ellemers, & S. A. Haslam (Eds.), The social psychology ofstereotyping and group life (pp. 296-317). Oxford: Blackwell.Long, K. M., Spears, R., & Manstead, A. S. R. (1994). The influence of personal andcollective self-esteem on strategies of social differentiation. British Journal of SocialPsychology, 33 (3), 313-219.Luhtanen, R., & Crocker, J. (1991). Self-esteem and intergroup comparisons: Towarda theory of collective self-esteem. In J. Suls & T. A. Wills (Eds.), Social comparison:Contemporary theory and research (pp. 211-234). Hillsdale, NJ: Erlbaum.Luhtanen, R., & Crocker, J. (1992). A collective self-esteem scale: Self-evaluation ofone's social identity. Personality and Social Psychology Bulletin, 18 (3), 302-318. Major, B., Sciacchitano, A. M., & Crocker, J. (1993). Ingroup versus outgroupcomparisons and self-esteem. Personality and Social Psychology Bulletin, 19 (6), 711-721.Maass, A., & Acuri, L. (1996). Language and stereotyping. In N. Macrae, M.Hewstone, & C. Stangor (Eds.), The foundations of stereotypes and stereotyping (pp. 193-226). New York: Guilford Press.Maass, A., Ceccarelli, R., & Rudin, S. (1996). Linguistic intergroup bias: Evidence forin-group-protective motivation. Journal of Personality and Social Psychology, 71 (3), 512-526.Meindl, J. R., & Lerner, M. J. (1984). Exacerbation of extreme responses to anoutgroup. Journal of Personality and Social Psychology, 47, 71-84.Messick, D. M., & Mackie, D. M. (1989). Intergroup relations. Annual Review ofPsychology, 40, 45-81.Mullen, B., Brown, R., & Smith, C. (1991). Ingroup bias as a function of salience, relevance and status: An integration. European Journal of Social Psychology, 22, 103-122. Social Identity Theory 29 Mullin, B. A., & Hogg, M. A. (1995). Dimensions of subjective uncertainty in socialidentification and minimal intergroup discrimination. Unpublished manuscript, University ofQueensland.Nascimento-Schulze, C. M. (1993). Social comparison, group identity andprofessional identity. A study with bank clerks. Revista de Psicologia Social, 8 (1), 69-82.Oakes, P. J., Haslam, S. A., & Turner, J. C. (1994). Stereotyping and social reality.Oxford: Blackwell.Oakes, P. J., & Turner, J. C. (1980). Social categorization and intergroup behaviour:Does minimal intergroup discrimination make social identity more positive? EuropeanJournal of Social Psychology, 10, 295-301.Orne, M. T. (1962). On the social psychology of the psychological experiment withparticular reference to the demand characteristics and their implications. AmericanPsychologist, 17, 776-783. Pettigrew, T. F. (1979). The ultimate attribution error: Extending Allport’s cognitiveanalysis of prejudice. Personality and Social Psychology Bulletin, 5, 461-476.Phinney, J. S., Chavira, V., & Tate, J. D. (1993). The effects of ethnic threat on ethnicself-concept and own-group ratings. Journal of Social Psychology, 133 (4), 469-478.Platow, M. J., Harley, K., Hunter, J. A., Hanning, P., Shave, S., & O’Connell, A.(1997). Interpreting in-group-favouring allocations in the minimal group paradigm. BritishJournal of Social Psychology, 36, 107-117.Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ:Princeton University Press.Rosenberg, M., Schoenbach, C., Schooler, C., & Rosenberg, F. (1995). Global self-esteem and specific self-esteem: Different concepts, different outcomes. AmericanSociological Review, 60, 141-156.Ruttenberg, J., Zea, M. C., & Sigelman, C. K. (1996). Collective identity and intergroup prejudice among Jewish and Arab students in the United States. Journal ofSocial Psychology, 136 (2), 209-220.Sachdev, I., & Bourhis, R. Y. (1984). Minimal majorities and minorities. EuropeanJournal of Social Psychology, 12, 307-314.Sachdev, I., & Bourhis, R. Y. (1985). Social categorization and power differentials ingroup relations. European Journal of Social Psychology, 15, 415-434.Sachdev, I., & Bourhis, R. Y. (1987). Status differentials and intergroup behaviour.European Journal of Social Psychology, 17, 277-293.Sachdev, I., & Bourhis, R. Y. (1991). Power and status differentials in minority andmajority group relations. European Journal of Social Psychology, 21, 1-24.Schiffman, R., & Wicklund, R. A. (1992). The minimal group paradigm and its minimalpsychology: On equating social identity with arbitrary group membership. Theory andPsychology, 2 (1), 29-50. Seta, C. E., & Seta, J. J. (1992). Observers and participants in an intergroup setting.Journal of Personality and Social Psychology, 63, 629-643.Seta, J. J., & Seta, C. E. (1996). Big fish in small ponds: A social hierarchy analysis ofintergroup bias. Journal of Personality and Social Psychology, 71 (6), 1210-1221.Sidanius, J., Pratto, F., & Mitchell, M. (1994). Ingroup identification, social dominanceorientation and differential intergroup social allocation. Journal of Social Psychology, 134,(2), 151-167.Smith, H. J., & Tyler, T. R. (1997). Choosing the right pond: The impact of groupmembership on self-esteem and group-oriented behavior. Journal of Experimental SocialPsychology, 33 (2), 146-170.St. Clare, L., & Turner, J. C. (1982). The role of demand characteristics in the socialcategorization paradigm. European Journal of Social Psychology, 12, 307-314.Tajfel, H., Billig, M. G., Bundy, R. P., & Flament, C. (1971). Social categorization and intergroup behaviour. European Journal of Social Psychology, 1, 149-178. Social Identity Theory 30 Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In S.Worchel & W. G. Austin (Eds.), The social psychology of intergroup relations (pp. 33-47).Monterey, CA: Brooks-Cole.Turner, J. C. (1975). Social comparison and social identity: Some prospects forintergroup behaviour. European Journal of Social Psychology, 5, 5-34.Turner, J. C. (1980). Fairness or discrimination in intergroup behaviour? A reply toBranthwaite, Doyle and Lightbown. European Journal of Social Psychology, 10, 131-147.Turner, J. C. (1982). Towards a cognitive redefinition of the social group. In H. Tajfel(Ed.), Social identity and intergroup relations. Cambridge: Cambridge University Press.Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. (1987).Rediscovering the social group: A self-categorization theory. Oxford: Blackwell.Vanbeselaere, N. (1991). The different effects of simple and crossed categorizations:A result of the category differentiation process or of differential category salience? In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 2, pp. 247-278).Chichester, England: Wiley.Vanbeselaere, N. (1996). The impact of differentially valued overlappingcategorizations upon the differentiation between positively, negatively, and neutrallyevaluated social groups. European Journal of Social Psychology, 26, 75-96.Verkuyten, M. (1997). Intergroup evaluation and self-esteem motivations: Self-enhancement and self-protection. European Journal of Social Psychology, 27, 115-119.Wagner, U., Lampen, L., & Syllwasschy, J. (1986). Ingroup inferiority, social identityand outgroup devaluation in a modified minimal group study. British Journal of SocialPsychology, 25, 15-23.Wills, T. A. (1981). Downward comparison principles in social psychology.Psychological Bulletin, 90, 245-271. Social Identity Theory 31 Footnotes1. One anonymous reviewer noted a possible conceptual confusion arising betweenthe global-specific distinction and univs. multi-dimensional measures of self-esteem.Certainly the two distinctions refer to similar ideas, however, the global-specific distinctionrefers to a type of self-esteem whereas the univs. multi-dimensional distinction refers to atype of scale. Hence, it is theoretically possible to measure both global and specific self-esteem as either unior multi-dimensional constructs. However, the potential for multi-dimensional scales is liable to decrease as the specificity of the self-aspect being evaluatedincreases. Hence, it will be easier to construct a multi-dimensional global personal trait self-esteem scale than a multi-dimensional specific social state self-esteem scale.2. The Twenty Statements Test requires respondents to generate twenty differentresponses to the question “who am I?” and then to rate their self-esteem in terms of eachresponse. This technique will inevitably elicit a list of social identities (e.g., “I am a man”). Although each response represents a measure of specific social trait self-esteem, when allresponses are combined to create an overall score it would seem more appropriate to treatthe scale as a measure of global personal trait self-esteem.3. The Texas Social Behaviour Inventory focuses on interpersonal, rather thanintergroup social self-evaluation.4. Bourhis (1994) and Sachdev and Bourhis (1984, 1985, 1987, 1991) also includeposttest measures of self-esteem in their studies, however, these cannot be legitimatelyrelated to corollary 1 because they are confounded by manipulations of power or status, orboth.5. The self-esteem hypothesis does not explicitly acknowledge distinctions betweendifferent types of self-esteem, and so these distinctions are only included for purposes ofcomparison and are not used to evaluate whether a particular piece of evidence issupportive or unsupportive. 6. Since ingroup favouritism and outgroup derogation were measured relative to abaseline condition in which the amount of normative discrimination prescribed by the statusdifferences of aboveand below-average scorers was ascertained, they representcompetitive discrimination unconfounded by normative discrimination.7. Given the diversity of methods, designs, and measures that have been employedto test the self-esteem hypothesis we opted for a qualitative rather than quantitative reviewin order to provide a more detailed analysis of the limited number of studies that areavailable. Social Identity Theory 32 Table 1Eight Types of Self-esteem and Example Items Type of self-esteemExample item Global personal traitUsually, I feel good about myself. Specific personal traitUsually, I feel good about my physicalappearance. Global personal stateAt the moment, I feel good about myself. Specific personal stateAt the moment, I feel good about my physicalappearance. Global social traitUsually, I feel good about my social groups. Specific social traitUsually, I feel good about my gender. Global social stateAt the moment, I feel good about my socialgroups. Specific social stateAt the moment, I feel good about my gender. Social Identity Theory 33 Table 2Evidence based on Competitive Discrimination Corollary 1Corollary 2 Supportive evidence Unsupportive evidence Supportive evidenceUnsupportiveevidence