Effects of intergroup similarity on intergroup relations

European Journal of Social Psychology, Vol. 23,581-595 (1993) zyx z zyxw Effects of intergroup similarity on intergroup relations SONIA ROCCAS and SHALOM H. SCHWARTZ zyxwv The Hebrew University of Jerusalem, Israel Abstract We examined the impact of intergroup similarity on two aspects of intergroup relations. Drawing on social identity and belief congruence theory, we hypothesized that - at high levels of intergroup similarity - increasing similarity has dual, seemingly opposed effects: It increases ingroup favouritism in evaluations but also increases readiness for social contact with the outgroup. We further hypothesized that both effects are moderated by the strength of individuals’ identiJication with their ingroup. Finally, we hypothesized that there is ingroup favouritism on dimensions relevant for defining the group, but outgroup favouritism on dimensions irrelevant for this purpose. One hundred and forty-nine students from two prestigious high schools, who were assigned to one of three levels of manipulated similarity between their schools, evaluated both schools on dimensions relevant and irrelevant to the school context and expressed their readiness for social contact with the other school. Ingroup favouritism appeared on relevant dimensions and outgroup favouritism on irrelevant dimensions. As predicted, for those highly identified with their ingroup, intergroup similarity led to greater ingroup favouritism in evaluations on relevant dimensions but to increased readinessfor outgroup social contact. Implications for interpreting inconsistent results of past research and for specgying conditions for intergroup bias are discussed. zyxw INTRODUCTION The assumption that similarity between groups improves intergroup relations has received broad research support. For example, research dealing with racial discrimination revealed that perceived similarity of beliefs consistently reduced discrimination towards outgroup members (Insko, Nacoste and Moe, 1983). However, this assumption is sometimes contradicted. For instance, Genthner, Shuntich and Bunting (1975) zyxwvu zyxwvutsr zyxwv This research was supported by grant No. 88-00085 from the United States-Israel Binational Science Foundation (BSF), Jerusalem, Israel, and by a grant from the Office of Science and Development of the National Council for Research and Technology (Israel), both to the second author. Reprint requests should be addressed to Shalom H. Schwartz, Department of Psychology, The Hebrew University, Jerusalem, Israel, 91905. 0046-2772/93/060581-15$12.50 0 1993 by John Wiley & Sons, Ltd. Received 21 April 1991 Accepted 20 January 1992 582 zyxwvutsr zyxwvu zyxw S. Roccas and S. H. Schwartz found that highly prejudiced white subjects were more aggressive towards a black with similar beliefs than towards a black with dissimilar beliefs. The current study seeks to clarify some of the conditions that determine whether similarity improves or damages intergroup relations. Research that examines the effects of similarity on intergroup relations has drawn mainly upon two theories, belief congruence and social identity. We briefly sketch these theories and relevant research, noting aspects helpful for identifying the conditions that interest us. Belief congruence Belief congruence theory (Rokeach, 1960)extrapolates from the similarity-attraction paradigm of interpersonal relations (Byrne, 1971) to the effects of similarity on relations between members of different groups. Research in this framework shows that perceived similarity between own beliefs and those of an outgroup member is negatively related to numerous types of discrimination (Insko, et al., 1983). However, three characteristics of this research limit generalizing from its conclusions to the effects of intergroup similarity on intergroup relations. Focus on individuals Experiments based on belief congruence theory typically inform subjects about the amount of belief similarity between self and individual target persons either from their ingroup or from a known, other group. Three aspects of this procedure limit inferences to the intergroup situation. First, the information about similarityconcerns a target individual and not a group. But, people tend to perceive information about individuals that contradicts group stereotypes in a biased way that preserves the stereotypes (Wilder, 1986). Consequently, information that single members of an outgroup are similar may have different effects than information that the outgroup itself is similar. Second, in these experiments, varied information is sometimes provided about multiple stimulus individuals. This focuses attention on individual characteristics and reduces the salience of group membership (Brown, 1988). Third, behaviour toward individuals rather than toward groups is examined. But, behaviour toward specific members of an outgroup may differ from behaviour toward the outgroup as a whole. Nature of the dependent variable Belief congruence research usually studies affective and behavioural dependent variables (e.g. liking, readiness for social contact) and not perceptual bias toward the outgroup. But, behaviour toward an outgroup is not necessarily connected with the way the outgroup is perceived in comparison to the ingroup. For example, Struch and Schwartz (1989) reported that the behavioural intention to be aggressive toward an outgroup was virtually uncorrelated with the tendency to perceive it as inferior to the ingroup. It is therefore likely that the conditions that promote the expression of outgroup discrimination in behavioural and affective modes differ from those that promote its expression in perceptual biases. zy zyxwvut zyxw zyxwvu Intergroup similarity 583 Amount of similarity between groups Only increases in similarity at low levels have been shown to promote positive responses to target persons in belief congruence studies. However, it is quite possible that increases at higher levels of similarity may yield a negative relation between similarity and responses. This is suggested by extension from uniqueness theory (Snyder and Fromkin, 1980), optimal distinctiveness theory (Brewer, 199l), and other models of individuation (e.g. Codol, 1984; Lemaine, 1974). These theories postulate that people are motivated to feel moderately different from others, to find an optimal balance between assimilation with and differentiation from others. Very low as well as very high similarity with others promote negative responses. To determine whether very high levels of intergroup similarity are also aversive, higher levels of similarity than those previously studied are required. Social identity According to social identity theory (Hogg and Abrams, 1988; Tajfel, 1974), individuals are motivated to derive a positive social identity from their group membership by perceiving their ingroup as positively distinctive compared to other groups. Perception of great intergroup similarity can be threatening because it blurs the distinction between groups, making it difficult for individuals to derive a distinctive positive social identity. Mere categorization into different groups elicits social comparison that leads directly to intergroup competition for prestige. This competition is the reason for ingroup bias -the widespread tendency to view the ingroup more favourably than outgroups (Mullen, Brown and Smith, 1992). Ingroup favouritism can be expressed materially (in allocation of tangible rewards: e.g. Tajfel, 1974; Brewer and Silver, 1978; Allen and Wilder, 1975) or cognitively (perceiving the ingroup as better on important dimensions: Brewer and Silver, 1978; Mummendey and Schreiber, 1983, 1984a). Group membership may produce intergroup bias even when groups have no objective reason for antagonism toward one another. Studies in the social identity framework have also examined the effects of similarity on intergroup relations. These studies have yielded inconsistent outcomes. Allen and Wilder (1975) found that degree of belief similarity with members of the ingroup and of the outgroup had no effect on the allocation of a desired resource. Moghaddam and Stringer (1988) found the usual ingroup bias when subjects perceived themselves to be similar both to the ingroup and to the outgroup, but outgroup favouritism when ingroup selection was based on chance and the outgroup was perceived to be similar. Turner (1978) studied the effect of similarity with the outgroup under conditions of stable and unstable status relations between groups. When status relations were stable, more ingroup bias was found for similar groups than for dissimilar groups. The opposite results were found when status relations were unstable. Mummendey and Schreiber (1984b) replicated the Turner study, but obtained very different results. Diehl (1988) manipulated both interpersonal and intergroup attitudinal similarity. When interpersonal similarity was manipulated, subjects favoured a similar outgroup member. In contrast, when intergroup similarity was manipulated, they favoured members of a dissimilar outgroup. Finally, Brown and Abrams (1986) examined the effects of similarity on liking 584 zyxwvutsr zyxwvut zyxwvuts zy S, Roccas and S. H. Schwartz for the outgroup and on evaluation of its performance. Students informed that outgroup members were similar to them either in their attitudes or in their scholastic abilities evaluated the performance of the outgroup more highly than did those told that outgroup members differed from them on both dimensions. Thus, similarity increased evaluation. However, evaluation of the outgroup was least positive among subjects told that the outgroup was similar to them in both its attitudes and its scholastic abilities. Brown (1 988) interpreted this as indicating that intergroup similarity beyond a certain threshold is aversive and threatening. Interestingly, in all experimental conditions, similarity was positively related to liking for the outgroup. In sum, studies in the social identity framework have shown some - albeit far from consistent - evidence of a negative impact of similarity on intergroup bias. This negative impact contradicts hypotheses derived from belief congruence theory. Two other variables studied in research stemming from social identity theory are relevant to our concerns, importance of the dimensions of group comparison and identification with the ingroup. We discuss these variables next. Importance of the dimensions of comparison Intergroup bias is related to the importance of the dimensions on which the groups are evaluated (Turner, 1978; Van Knippenberg, 1978). Mummendey (Mummendey and Schreiber, 1984a; Mummendey and Simon, 1989) found greater ingroup favouritism on dimensions of high importance to the ingroup than on those of low importance to it. Moreover, on dimensions of low importance to the ingroup but high importance to the outgroup, she found outgroup favouritism. The latter shows that intergroup bias may entail enhancement rather than devaluation of the outgroup. It contradicts the need for a positive social identity and requires another motivation to explain it'. Wilder (1986) proposed such a motivation - a norm of fairness. The fairness norm motivates people to compensate ingroup favouritism on important dimensions by outgroup favouritism on unimportant dimensions, if there is no cause to discriminate against the outgroup. ZdentiJicationwith the ingroup According to social identity theory, the motivation to positively differentiate the ingroup arises only if individuals have internalized their group membership as an aspect of their self-concept, that is, if they subjectively identify with the ingroup (Tajfel and Turner, 1979). Hinkle and Brown (1990) review results of 14 studies of the effects of degree of ingroup identification on ingroup bias. In most studies, ingroup identification correlated positively with ingroup favouritism, but the pattern was inconsistent and significant negative correlations were also found. Thus, the relation of ingroup identification to intergroup bias clearly merits further research. We will consider identification as a moderator of the effects of other variables. This is suggested by the finding that various predictors of intergroup aggression were ' zyxwvutsr Hinkle and Brown (1990) cite other examples of outgroup favouritism as well as proposals to explain it that they find problematic. In their meta-analysis, Mullen et al. (1992) report that the occurrence of ingroup versus outgroup favouritism varies as a joint function of ingroup status and importance of the dimension. zy zyxwv Intergroup similarity 585 significantly stronger among respondents who identified highly with their ingroup (Struch and Schwartz, 1989). THE PRESENT RESEARCH The present research sought to overcome the limitations of belief congruence studies noted above. To that end, we focused on the intergroup rather than the interpersonal level: We manipulated similarity information about the outgroup as a whole, rather than about a specific outgroup member. We also assessed attitudes toward the group as a whole. Furthermore, we studied high levels of similarity. This is because we assumed that subjects would feel threat to their unique social identity only at very high levels of similarity. In order to enable the subjects to express outgroup as well as ingroup favouritism, we measured intergroup bias using dimensions with varying importance to the definition of the group. We tested three hypotheses. The first concerned effects of the nature of the dimensions of similarity on the direction of intergroup bias; the second concerned effects of degree of similarity on group evaluation; the third concerned effects of similarity on readiness to interact with the outgroup. We discuss each in turn. In past research, the direction of intergroup bias depended upon whether the dimensions of evaluation were important to the ingroup or to the outgroup. In the present research, very similar groups were examined, so the same dimensions were important to both groups. We postulate that the relevance of the dimensions for dejining the group has effects similar to those of importance*. Perceiving the ingroup as superior to an outgroup on dimensions relevant for group definition is likely to enhance social identity. However, perceiving the ingroup as superior on dimensions irrelevant to group definition may confer no benefit, and perceiving the outgroup as superior may pose no threat. We also assume, following Wilder (1986), that intergroup bias represents a compromise between motivations to achieve a positive social identity and to act fairly. We therefore hypothesize: There is (1A) ingroup favouritism on dimensions relevant for defining the group, but (1B) outgroup favouritism on dimensions irrelevant for this purpose. Following social identity theory, we assume that very high similarity threatens the sense of distinctive social identity, motivating subjects to discriminate between groups and to perceive the ingroup as superior. However, we postulate that identification with the ingroup moderates the relation between intergroup similarity and ingroup bias. If a person does not identify with the ingroup, the distinctiveness of his or her ingroup will not affect that person’s social identity. Consequently, the person will not feel threatened by learning that an outgroup is very similar to the ingroup. Moreover, as theorized above, only similarity on dimensions relevant for defining the group poses a threat because it is superiority on these dimensions that contributes to a positive social identity. We therefore hypothesize: (2) For group members who identify with their ingroup, increased similarity on dimensions relevant for defining the ingroup leads to increased ingroup favouritism in evaluation. Note that this is the usual social identity theory prediction, but with the specification of two limiting conditions. zyxw zyxwvuts Our concept of relevance for defining the group differs from the use of relevance in Mullen e f al. (1992), where it is equivalent to importance. zyxwvuts zyxwvuts 586 S. Roccas and S. H. Schwartz Ingroup favouritism and a desire to avoid social contact with the outgroup do not necessarily occur together (Struch and Schwartz, 1989). Ingroup bias apparently can preserve the sense of distinct positive social identity without promoting hostility to the outgroup. Although similarity with members of an outgroup may threaten one’s social identity, such similarity can also make the prospect of interaction with members of that same outgroup more desirable. Interaction with similar others is more likely to provide the interpersonal rewards of reassurance and self-confirmation (Byrne and Clore, 1970). Moreover, the expectation that those who are more similar to us will reward us with more liking also makes interaction with them more attractive (Condon and Crano 1988). Hence, increasing similarity with outgroups, even at high levels, is likely to increase readiness to engage in social contact with their members. This prediction conforms, of course, with belief congruence theory. We expect identification with the ingroup to moderate the effect of intergroup similarity on readiness for social contact. Identification with one’s ingroup reflects, in part, a sense of similarity to ingroup members (Skevington, 1981). For those not identified with their ingroup, intergroup similarity implies little similarity between self and outgroup members. Consequently, for them, interaction with a similar outgroup may promise little personal reinforcement. We therefore hypothesize: (3) Increased similarity leads to greater readiness for contact with the outgroup among those who identify with the ingroup. METHODS Sample and design zyxwv One hundred and forty-nine 10th and 1lth grade students from the three most prestigious high schools in Jerusalem, Israel responded to questionnaires during regular class sessions. Prestigious schools were selected in order to permit the assumption that attending these schools was a source of pride for students. A 2(comparison dimension: relevanvirrelevant) x 3(similarity: very high/high/moderately high) factorial design was employed. Comparison dimension was a within-subjects factor and similarity a between-subjects factor. Students were randomly assigned to similarity condition. Procedure The experiment was presented as a study of students’ views of different high schools. After completing the experimental questionnaire, students were debriefed. They then completed a questionnaire concerning ethical aspects of the research. The following variables were manipulated or measured in the experimental questionnaire. Manipulation of similarity Subjects read that researchers had considerable objective information about various schools and were now interested in the students’ evaluations of their school as compared to other schools. Students were then provided with information about the degree of similarity between their school and one other school with regard to success Intergroup similarity zy 587 in national examinations, choices of topics for national examinations, readiness to exert themselves during compulsory military service, percentage volunteering for combat units, percentage of graduates fit to serve as army officers, choices of careers, and academic success. In the ‘moderately high’ similarity condition, subjects were told that graduates of both schools were very similar on the aforementioned characteristics. In the ‘high’ similarity condition, they were told, in addition, that previous research had revealed that graduates of these two schools report similar school experiences that are different from the experiences reported by graduates of other high schools, and that students of other schools consider these two schools to be virtually identical. In the ‘very high’ similarity condition they were told, in addition, that-due to budget deficits-the Ministry of Education was considering unifying various high schools and that these two schools were candidates for future unification because they were so similar. zyxwv zyx zyxwvu zyxwvut zyx Evaluation of the ingroup and the outgroup Subjects estimated the proportion of students in their school (ingroup) and in the other school (outgroup) who do each of the following six things: (1) Study hard, (2) are involved in politics, (3) are interested in popular music, (4) are active in youth organizations, ( 5 ) want to go to college, (6) consider the teachers in their school better than average. Three items were chosen from a pool of activities that judges assessed a priori as relevant to the school context (1,5,6) and three from a pool assessed as important to high school students but irrelevant to the school context (2,3,4). Responses were given on a 7-point scale labelled at its extremes: 1-Almost no one, and 7-almost everyone. Subjects also indicated whether they considered the type of behaviour described by each of the items as positive, neutral, or negative3. Importance of the dimensions To provide an estimate of the importance of each dimension for evaluating a school, subjects rated each of the six items on a scale that ranged from 1 - not important at all to 7 -very important. Indexes of bias Two indexes of bias in intergroup evaluation were computed, one based on items relevant to the school context and the other based on the items irrelevant to this context. For each item, the evaluation of the outgroup was subtracted from the evaluation of the ingroup. This difference score was then multiplied by the importance rating the subject gave to the item as a dimension for evaluating a school. The mean of these weighted difference scores for the three relevant items (alpha = 0.65) zy Subjects evaluated the schools on an additional item - cheating on examinations. This item was not included in analyses because student’s appraisals of the desirability of this behaviour varied greatly: 1 1 per cent considered cheating positive, 42 per cent negative, and 47 per cent neutral. All other items were rated as non-negative by at least 97 per cent of subjects. 588 zyxwvuts zyxw zyxwvu zyx S. Roccas and S. H. Schwartz and for the three irrelevant items (alpha = 0.30) served as measures of intergroup bias4. Readiness to engage in social contact For each of the following six activitiesthat might be carried out together with another school, subjects were asked if they would rather participate in the activity with members of the designated other school which they had evaluated or with members of another, unspecified school: (1) A graduation party, (2) a youth delegation to travel abroad, (3) an organization for volunteer activities, (4)seminars on social problems, (5) a week of pre-military training, (6) a programme of extracurricular courses. An index of readiness for social contact with the designated other school (the outgroup) was computed by summing the number of activities students preferred to be engaged in with that school (alpha = 0.90). Identijication with the ingroup To measure identification, we summed responses on 7-point scales to the following three questions: (1) How proud are you of attending your school? (2) How satisfied are you with having chosen your school? (3) How willing are you to recommend your school to other potential students? The internal reliability of this index of identification was alpha = 0.89. Subjects above and below the median on the identification index were defined as constituting the high and low identification groups. PILOT STUDY A pilot study was conducted to check whether the manipulation of intergroup similarity was effective, and to validate the a priori classification of items as relevant or irrelevant to the school context. Thirty-two high school students were randomly assigned to one of three similarity conditions parallel to the conditions in the main study: Very high/high/moderately high. The manipulation of similarity differed only in that the two schools described were not named. Subjects were asked to evaluate the similarity between the two schools on two items, using a scale anchored by 0 (not at all similar) and 100 (completely similar). An ANOVA on the perceived similarity (average of two items) of the schools confirmed the effectiveness of the manipulation. Perceived similarity was highest in the very high condition ( M = 87.8), lower in the high condition ( M = 77.8), and lower still in the moderately high condition ( M = 69.8) (F(2,29df) = 8.75, p < 0.05; all means different by Newman-Keuls test, p <0 .05). As intended, moreover, the perceived level of similarity was quite high in all conditions. We also assessed the relevance to the school context of the six items used to measure evaluation of the ingroup and the outgroup in the main study. Subjects rated the relevance of each item as a determinant of the prestige of a school, using a 7-point scale. A principal components factor analysis of the responses, constrained to a two-factor solution, accounted for 57.3 per cent of the variance. The first factor zyxwvut Given the low alpha for the index of bias on the irrelevant dimensions, analyses of irrelevant items were run separately on each item. Intergroup similarity zy zy 589 was composed of the three items classified a priori as irrelevant (31 per cent of the variance, loadings 0.62 to 0.84) and the second factor of the three items classified a priori as relevant (26 per cent, loadings 0.48 to 0.82). RESULTS Direction of intergroup bias on relevant and irrelevant dimensions Before examining the effects of similarity on intergroup bias, we sought to establish whether the usual ingroup favouritism was present here. We hypothesized that (1A) ingroup favouritism is found on dimensions relevant to the school context, and (1 B) outgroup favouritism is found on irrelevant dimensions. In order to measure intergroup bias, the evaluation of the outgroup was subtracted from the evaluation of the ingroup, for each item. Positive scores on intergroup bias indicate ingroup favouritism and negative scores indicate outgroup favouritism. A MANOVA on the bias scores for the six items was significant (F(6,141 df) = 12.76, p < 0.05). Column one of Table 1 presents the ingroup bias scores for each item, and column two presents the univariate F-values. As hypothesized (1A), ingroup favouritism was expressed on all three items relevant to the school context. In support of hypothesis lB, outgroup favouritism was expressed on two of the three items irrelevant to the school context. The third item showed no bias. zy zy Table 1. Ingroup bias scores and importance ratings for dimensions relevant and irrelevant to the school context ~ Items Relevant to the school context Want to go to college Consider teachers above average Study hard Irrelevant to the school context Active in youth organizations Interested in popular music Involved in politics Bias score 0.18 0.46 0.60 0.01 -0.46 -0.37 Univariate F test (df1,146) ~~ Importance ratings zyxwv 4.30* 13.14* 36.14* 4.92 6.27 4.85 0.00 17.17* 12.80* 3.57 3.16 4.92 *p < 0.05 for the difference between evaluation of the ingroup and the outgroup. Recall that the rationale for these hypotheses developed in the Introduction was based on the idea that only dimensions relevant for defining the group affect social identity. Past theorizing and findings on ingroup and outgroup favouritism referred to the importance rather than the relevance of dimensions for social identity. In order to establish whether our findings might reflect the impact of importance rather than the hypothesized effects of relevance, we examined the importance ratings of the items. These ratings, shown in column three of Table 1, reveal that - on average -items relevant to the school context were evaluated as more important than irrelevant items (M = 5.36 and 3.88 respectively, t(152’d = 1 2 . 6 7 , ~< 0.05). However, one of the items irrelevant to the school context (involvement in politics) was also 590 zyxwv zyxwvutsr S. Roccas and S. H. Schwartz one of the more important items. This suggests at least partial overlap of relevance with importance in the current study. Moreover, on this item, outgroup favouritism was shown, as predicted from its irrelevance, rather than ingroup favouritism, as predicted from its high importance. Effects of manipulatedsimilarity on intergroup bias We hypothesized (2) that for group members who identify with their ingroup, increased similarity on dimensions relevant for defining the ingroup leads to ingroup favouritism. That is, in the presence of high initial levels of similarity, manipulated similarity increases ingroup favouritism. Note that this hypotheses is limited to people who identify with the ingroup and refers to dimensions relevant for defining the group. The relevant findings are presented in Table 2. The bias indexes are the means of the weighted difference scores for the sets of school relevant items. Positive scores indicate ingroup favouritism, negative scores outgroup favouritism. zyxw Table 2. Intergroup bias on dimensions relevant to group definition as a function of similarity level and identification with the ingroup Level of manipulated similarity High Very high High Moderately high 2.99 1.95 -0.20 Identification (N) Low (N) (30) (26) (33) (22) (21) (17) 5.29 3.04 4.98 zyxwvu zyxw zyxw The bias scores in column one indicate that, as predicted, the greater the manipulated similarity between ingroup and outgroup, the greater the ingroup favouritism. To test the reliability of the predicted difference on ingroup favouritism among those highly identified with the ingroup, we conducted a planned comparison, using the error term from the 3 x 2 (similarity X identification) between-subjects ANOVA. This comparison revealed that, as predicted, highly identified subjects expressed significantly more ingroup favouritism in the very high similarity condition than in the moderately high similarity condition (t(140df) = 2 . 1 4 , ~< 05)’. The overall ANOVA also revealed a difference that was not predicted: Those low in ingroup identification showed stronger ingroup bias than those high in identification (F(1,140 df) = 7.92, p < 0.05). Phrased differently, the less strongly students identified with their school, the more they biased their evaluations in favour of their school6. Effects of manipulatedsimilarity on readiness for social contact We hypothesized (3) that increased similarity leads to greater readiness for social contact with the outgroup, among those who identify with their ingroup. Table ’ Analyses of bias on the irrelevantitems revealed no significanteffect of similarityor of identification. The other effects in the overall ANOVA (similarity and the two-way interaction) were not significant. zy zyxwv Intergroup similarity 591 3 presents the mean index of readiness for social contact as a function of similarity level and identification with the ingroup. Table 3. Readiness for social contact with the outgroup as a function of similarity level and identificationwith the ingroup ~~ Level of manipulatedsimilarity Very high High Moderately high ~ High Identification (N) Low (N) 4.67 4.56 4.12 (30) (26) (33) (22) (21) (17) 3.39 3.84 4.28 zyxwv zyxw As shown in column one of the table, which refers to those highly identified with their ingroup, the experimental groups were ordered as expected on the index of readiness for outgroup contact. The appropriate planned comparison, using the error term from the 3 x 2 (similarity x identification)between-subjectsANOVA, indicated that those told the two schools were very highly similar, were significantly more ready to engage in outgroup contact than those told the schools were only moderately highly similar (t(143 df) = 1 . 6 6 , ~< 0.05)7. DISCUSSION The major goal of this research was to examine the impact of intergroup similarity on two aspects of intergroup relations: Biased perception (evaluation) and readiness for social contact. We first sought to clarify the types of dimensions on which perceptual bias appears. Relevance and importance of dimensions of evaluation Ingroup favouritism was found on dimensions of evaluation relevant to the school context and outgroup favouritism on two of three dimensions irrelevant to this context. These results parallel Mummendey’s findings for dimensions varying in importance: Ingroup favouritism on dimensions important to the ingroup and outgroup favouritism on dimensions important to the outgroup (Mummendey and Schreiber, 1983, 1984a; Mummendey and Simon, 1989). Mummendey examined only importance, whereas we measured both importance and relevance. Relevance for group definition predicted the direction of intergroup bias better than importance did: Political involvement, which was highly important but irrelevant, yielded outgroup rather than ingroup favouritism. Our findings raise the possibility that, in general, the relevance of dimensions rather than their importance accounts for the direction of intergroup bias. Since relevance and importance are often confounded, research explicitly designed to tease out possible differences between the effects of these two variables is necessary. Social identity theory does not explain why there was outgroup favouritism on Neither the main effect nor the interaction was significant in the overall ANOVA. 592 zyxwvuts S. Roccas and S. H. Schwartz dimensions irrelevant to group definition. We based outgroup favouritism hypothesis on Wilder’s (1986) idea that subjects may also be motivated by a fairness norm in intergroup situations. Our findings support the view that subjects compensate for ingroup favouritism on relevant dimensions with outgroup favouritism on irrelevant dimensions. What conditions affect the tendency to compensate in this way? To the best of our knowledge, this phenomenon appears robustly only in studies of existing (non-experimental) groups in which subjects evaluated an outgroup whose basic characteristics were similar to their own. This suggests that compensating ingroup with outgroup favouritism may occur only in circumstanceswhere important bases of identification are shared by the ingroup and the outgroup. A second condition that may be necessary is the absence of real intergroup conflict (cf. Hinkle and Brown, 1990). In the presence of such conflict, compensatory outgroup favouritism may not emerge because the perceived conflict may justify discrimination. zy Effects of similarity We hypothesized that increased intergroup similarity has dual, seemingly opposed, effects on the two aspects of intergroup relations: It increases readiness for outgroup contact but also increases bias in favour of the ingroup. This research identified several conditions that effect the associations of intergroup similarity with these aspects of intergroup relations. We discuss each condition in turn. As predicted, subjects who identified with the ingroup showed greater ingroup favouritism on dimensions relevant to the school context the greater the manipulated similarity. For subjects who did not identify with the ingroup, ingroup favouritism did not vary systematically as a function of similaritylevel. These results are congruent with the social identity theory analysis that high similarity between an ingroup and an outgroup threatens distinctive social identity (Brown and Abrams, 1986). In past research, manipulated intergroup similarity had contradictory impacts on ingroup favouritism. We will integrate past results with our findings to clarify some conditions that moderate the influence of similarity on intergroup bias. zyxwvutsr Type of group Similarity has promoted intergroup bias in studies of existing groups but not of groups created for research purposes. The social identity linked to membership in the latter is transitory and probably of little significance for the self-concept (Abrams and Hogg, 1988). Intergroup similarity apparently poses a more serious and effective threat to social and hence to personal identities in existing groups, where emotional and self-concept investments in membership are greater. Degree of similarity Brown and Abrams (1986) found that manipulated similarity led to a lower evaluation of the outgroup only when subjects were informed that the groups were similar both in attitudes and status. Brown (1988) interpreted this as indicating that only high levels of similarity pose a threat to distinctive social identity. Our research supports this assertion. It also demonstrates a monotonic (positive) relation between zy zyxwvut zyxw zy Intergroup similarity 593 similarity and intergroup bias for the first time by manipulating three relatively high levels of similarity. Importance of group distinctiveness One other condition that may influence whether intergroup similarity is experienced as threatening is the importance of distinctiveness to the group’s justification for its existence. Political parties, for example, justify their existence by pointing to their distinctive ideologies. Hence, perceived similarity with another party is likely to be highly threatening. In contrast, the establishment of a public elementary school is primarily justified by the number of children in the neighbourhood, not by the distinctiveness of the goals of the school. Hence, teachers told that their school is very similar to another may experience little identity threat. This condition has yet to be studied empirically, but a related idea - the degree to which a group has a ‘comparativeideology’ -was suggested by Hinkle and Brown (1990). Identification with the ingroup Identification with the ingroup was negatively related to ingroup favouritism: Students less identified with their own school showed more ingroup favouritism. This unpredicted finding replicates findings in previous studies conducted in occupational settings (Brown, Condor, Wade and Williams, 1986; Brown and Williams, 1984; Oaker and Brown, 1986). It contrasts, however, with findings in research on political parties (Kelly, 1988, 1990). We discuss why schools are more similar to work-groups and then seek to explain this pattern of findings. Following Kelly (1988), we suggest that social identification is less central to membership in occupational than in political groups. Work-groups are mainly sources of income and friendship. In contrast, political party membership is largely based on identification with group goals. For students, schools are clearly more analogous to work-groups than to political parties. Students rarely enter schools out of personal identification with the schools’ goals. Low identification with a group probably reflects dissatisfaction with the benefits of group membership, including identity benefits. Tajfel(l978) suggested three alternative member responses when a group does not contribute positively to social identity. (1) Leave the group, if possible; (2) engage in social action to promote desirable changes, if leaving is impossible, or (3) reinterpret group attributes so that unwelcome features are either justified or made acceptable. In both work-groups and schools, dissatisfied members usually face serious barriers to leaving, inclining them to engage in social action or, more likely given the constraints, in reinterpretation. In contrast, it is usually much easier to quit or decrease investment in a political party. These responses may explain the negative relation between identification and ingroup favouritism among students and work-group members. Dissatisfied group members, unable to leave or to change the system, report low identification. When evaluating the group, however, they reinterpret its attributes, viewing them as more positive than the attributes of similar groups of which they are not members. This ingroup bias enables them to achieve some degree of positive social identity despite their dissatisfaction with their own group. zyxwvutsrq zyxwv zyxwvuts 594 S. Roccas and S. H. Schwartz Readiness to engage in social contact with the outgroup As with intergroup bias, the effects of manipulated similarity on readiness to engage in social contact with the outgroup depended upon the degree of identification with the ingroup. For highly identified subjects, similarity increased readiness for contact. For not especially identified subjects, readiness for contact was unrelated to similarity level. The results for readiness for contact are compatible with belief congruence theory. They join the results of a long series of studies demonstrating that similarity leads to attraction. This study adds three new points. First, similarity increases readiness for contact even when the similarity concerns a whole group and not a specific outgroup member. Second, similarity relates monotonically to readiness for contact across three levels of similarity. Third, similarity increases readiness for contact with an outgroup even as it simultaneously increases ingroup favouritism at the expense of the outgroup. REFERENCES zyx Abrams, D. and Hogg, M. (1988). ‘Comments on the motivational bias of self-esteem in social identity and intergroup discrimination’, European Journal of Social Psychology, 18: 317-334. Allen, V. L. and Wilder, D. A. (1975). ‘Categorization, belief similarity and group discrimination’, Journal of Personality and Social Psychology, 32:971-977. Brewer, M. B. (1991). ‘The social self:Being the same and different at the same time’, Personality and Social Psychology Bulletin, 17:475482. Brewer, M. B. and Silver, M. (1978). ‘Ingroup bias as a function of task characteristics’, European Journal of Social Psychology, 8: 393-400. Brown, R. J. (1988). Group Processes: Dynamics Within and Between Groups, Basil Blackwell, London. Brown, R. J. and Abrams, D. (1986). ‘The effect of intergroup similarity and goal interdependence on intergroup relations’, Journal of Experimental Social Psychology, 22:78-92. Brown, R. J., Condor, S., Wade, G. and Williams, J. (1986). ‘Explainingintergroup differentiation in an industrial organization’, Journal of Occupational Psychology, 59: 273-286. Brown, R. J. and Williams, J. (1984). ‘Group identification: The same thing to all people? Human Relations, 37:547-564. Byrne, D. (1971). The Attraction Paradigm, Academic Press, New York. Byme, D. and Clore, G. L. (1970). ‘A reinforcement model of evaluation processes’, Personality: An International Journal, l: 103-128. Codol, J.-P.(1984). ‘Social differentiation and nondifferentiation’. In: Tajfel, H. (Ed.) The Social Dimension, Cambridge University Press, Cambridge. Condon, J. W. and Crano, W. D. (1988). ‘Inferred evaluation and the relation between attitude similarity and interpersonal attraction’, Journal of Personality and Social Psychology, 54: 789-797. Diehl, M. (1988). ‘Social identity and minimal groups: The effects of interpersonal and intergroup attitudinal similarity on intergroup discrimination’, British Journal of Social Psychology, 27:289-300. Genthner, R. W., Shuntich, R. and Bunting, K. (1975). ‘Racial prejudice, belief similarity and human aggression’, Journal of Psychology, 91: 229-234. Hinkle, S. and Brown, R. (1990). ‘Intergroup comparisons and social identity: Some links and lacunae’. In: Abrams, D. and H o g , M. (Eds) Social Identity Theory: Constructive and Critical Advances, Wheatsheaf, Hemel Hempstead. Hogg, M. and Abrams, D. (1988). Social Ident$cations, Routledge, London. Insko, C., Nacoste, R. and Moe, J. (1983). ‘Belief congruence and racial discrimination:Review of the evidence and critical evaluation’, European Journal of Social Psychology, 13:153-174. zyxwvut zy Intergroup similarity zy zy 595 Kelly, C. (1988). ‘Intergroup differentiation in a political context’, British Journal of Social Psychology, 27: 319-332. Kelly, C . (1990). ‘Social identity and intergroup perception in minority-majority contexts’, Human Relations, 43: 583-592. Lemaine, G. (1974), ‘Social differentiation and social originality’, European Journal of Social Psychology, 4: 17-52. Moghaddam, F. and Stringer, P. (1988). ‘Outgroup similarity and intergroup bias’, Journal of Social Psychology, 128: 105-1 15. Mullen, B., Brown, R. and Smith, C. (1992). ‘Ingroup bias as a function of salience, relevance, and status: An integration’, European Journal of Social Psychology, 22, 103-122. Mummendey, A. and Schreiber, H. J. (1983). ‘Better or just different? Positive social identity by discrimination against or by differentiation from outgroups’, European Journal of Social Psychology, 13: 389-397. Mummendey, A. and Schreiber, H. J. (1984a). “Different just means ‘better’: Some obvious and some hidden pathways to in-group favoritism”, British Journal of Social Psychology, 23: 36S367. Mummendey, A. and Schreiber, H. J. (1984b). ‘Social comparison, similarity and ingroup favoritism’, European Journal of Social Psychology, 14: 231-233. Mummendey, A. and Simon, B. (1989). ‘Better or different? 111: The impact of importance of comparison dimension and relative in-group size upon intergroup discrimination’, British Journal of Social Psychology, 28: 1-1 6. Oaker, G. and Brown, R. (1986). ‘Intergroup relations in a hospital setting’, Human Relations, 3 9 767-778. Rokeach, M. (1960). The Open and ClosedMind, Basic Books, New York. Skevington, S. (1981). ‘Intergroup relations and nursing’, European Journal of Social Psychology, 11: 43-59. Snyder, C. R. and Fromkin, H. L. (1980). Uniqueness: The Human Pursuit of Dz@erence, Plenum Press, New York. Struch, N. and Schwartz, S. H. (1989). ‘Intergroup aggression: Predictors and distinctiveness from ingroup bias’, Journal of Personality and Social Psychology, 56: 364-373. Tajfel, H. (1974). ‘Social identity and intergroup behaviour’, Social Science Information, 13: 65-93. Tajfel, H. (1978). ‘Social categorization, social identity and social comparison’. In: Tajfel, H. (Ed.) Differentiation Between Social Groups: Studies in the Social Psychology of Intergroup Relations, Academic Press, London. Tajfel, H. and Turner, J. C. (1979). ‘The social identity theory of intergroup behavior’. In: Worchel, S. and Austin, W. G. (Eds) Psychology of Intergroup Relations, Nelson-Hall, Chicago. Turner, J. C. (1978). ‘Social comparison, similarity and ingroup favouritism’. In: Tajfel, H. (Ed.) Differentiation Between Social Groups: Studies in the Social Psychology of Intergroup Relations, Academic Press, London. Van Knippenberg, A. (1978). ‘Status differences,comparative relevance and intergroup differentiation’. In: Tajfel, H. (Ed.) Differentiation Between Social Groups: Studies in the Social Psychology of Intergroup Relations, Academic Press, London. Wilder, D. A. (1986). ‘Social categorization: Implications for creation and reduction of intergroup bias’. In: Berkowitz, L. (Ed.) Advances in Experimental Social Psychology, vol. 19, Academic Press, Orlando, FL. zyxwvut zy zyxwvut All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them i