Cooperative Interaction and Intergroup Bias: Effects of Numerical Representation and Cross-Cut Role Assignment.

Allport (1954) argued that to increase intergroup acceptance, members of distinct groups (e.g., ethnic groups) should have opportunities to interact with each other. To this end, he specified key structural features that are essential for positive intergroup contact. Among the most important of these features are cooperative interdependence, equal status in the contact setting, interpersonal intimacy, and sanction among those in authority (Allport, 1954; Cook, 1984). In support of Allport's theorizing, numerous studies have shown that cooperation between members of different social groups reduces both biased favoritism toward in-group members and animosity toward out-group members (e.g., Desforges et al., 1991; Gaertner, Mann, Dovidio, Murrell, & Pomare, 1990; Rabbie, Benoist, Oosterbann, & Visser, 1974; Sherif, Harvey, White, Hood, & Sherif, 1961). In addition, several qualitative and quantitative reviews of school-based research document the potential for cooperative learning strategies to improve interethnic acceptance within desegregated classrooms (Johnson & Johnson, 1975; Johnson, Johnson, & Maruyama, 1984; Miller & Davidson-Podgorny, 1987).

Nevertheless, several studies reveal that when numerical representations of different groups are markedly different, cooperation alone is not sufficient for favorable intergroup outcomes (e.g., Rogers, Hennigan, Bowman, & Miller, 1984; Worchel, Andreoli, & Folger, 1977). More specifically, being in a numerical minority appears to prevent members' attitudes from benefiting from cooperative interaction (Bettencourt, Charlton, & Kernahan, in press; Bettencourt, Miller, & Hume, 1997; Miller, Brewer, & Edwards, 1985; Rogers et al., 1984). Similarly, laboratory studies that include intergroup contact, but not cooperation, also show that numerical minority groups express more in-group bias than do numerical majority groups (e.g., Brewer, Manzi, & Shaw, 1993; Gerard & Hoyt, 1974; Mullen, Brown, & Smith, 1992; Sachdev & Bourhis, 1984, 1991).

The finding that the minority may not benefit from cooperation alone is important because some developers and proponents of cooperative-team learning interventions, designed to improve interethnic acceptance in classrooms, assume that the benefits of intergroup cooperation override the effects of other variables, such as minority status. For example, it is often suggested that numerical representation within cooperative teams should reflect the relative frequency of the respective social categories in the classroom (Aronson, Stephan, Sikes, Blaney, & Snapp, 1978; DeVries & Edwards, 1974; Johnson & Johnson, 1987; Slavin, 1986). However, a meta-analysis of the effects of cooperative-team learning on intergroup acceptance in school settings contradicts this suggestion (Miller & Davidson-Podgorny, 1987). Consistent with the findings of laboratory studies, Miller and Davidson-Podgorny (1987) found that unequal numerical representation of groups within cooperative teams was correlated with a reduction in intergroup acceptance.

Several researchers (Bettencourt et al., in press; Miller et al., 1985; Rogers et al., 1984) have attempted to resolve this problem by structuring the intergroup interaction in ways that should maximize the potential for cooperation and thus reduce the negative impact of minority status. However, the results of their studies have shown little improvement in the attitudes of minorities. For example, Bettencourt et al. (in press) instructed cooperative teams, composed of a distinct numerical majority and minority, to adopt a focus during the cooperative task. Although such a focus had been found to greatly reduce bias among equal size groups (Bettencourt, Brewer, Croak, & Miller, 1992), Bettencourt et al.'s (1992) results showed only minimal reduction in the in-group bias of minorities in the interpersonal focus condition, compared to those in a no-focus control condition. Perhaps an interpersonal focus failed to have the positive effect that it had on equal size groups because the salience of the minority category remained high during the cooperative interaction. Bettencourt and her colleagues further concluded that intergroup cooperation can only have beneficial effects on the intergroup attitudes of numerical minorities if the interaction is structured in ways that greatly reduce the salience of category boundaries. In general, members of numerical minorities are more aware of their social category than are members of numerical majorities (McGuire, McGuire, Child, & Fujioka, 1978; McGuire, McGuire, & Winton, 1979; Mullen, 1983), and groups for whom category salience is high are more biased in their intergroup attitudes (Haunschild, Moreland, & Murrell, 1994; Hong & Harrod, 1988).

The purpose of the present work was to investigate whether the intergroup attitudes of the numerical minorities are improved by a structural intervention that is designed to reduce category salience in a cooperative setting. Specifically, we examined whether the role assignments for completing a cooperative task influence the in-group bias of minority and majority groups. We focus here, as have other researchers (e.g., Brewer et al., 1993; Gerard & Hoyt, 1974; Sachdev & Bourhis, 1984), on the specific effects of numerical minority representation as distinct from that of ethnic minority status, and we hold constant other variables related to status and power. Although in real-world settings, belonging to a numerical minority is often confounded with other status and power variables, this is not always the case, and the effects of these variables are not always the same (e.g., Clark & Clark, 1947; Hewstone & Jaspers, 1982; Patchen, Hofmann, & Davidson, 1976; Sachdev & Bourhis, 1984, 1991).

In the following sections, we review the literature on the effects of cross-cutting role assignment, specifically with respect to cooperative settings. Moreover, we discuss why such assignments may have the potential to reduce biased in-group favoritism among numerical minorities. Finally, we report the attitudinal outcomes of two experiments in which the role assignments and numerical representations of social category members were varied within cooperative teams.

CROSS-CUT ROLE ASSIGNMENT TO REDUCE CATEGORY SALIENCE

One structural intervention purported to reduce the salience of intergroup boundaries involves cross-cutting social category membership with role assignments to different tasks that ultimately contribute to a cooperative effort (Brewer & Miller, 1984; Marcus-Newhall, Miller, Holtz, & Brewer, 1993). Cooperative team members are assigned to tasks to ensure that members with different in-group identities share task roles. By crossing roles with groups, team members are provided a new basis for differentiating among one another, and salience of prior social category boundaries is reduced (Marcus-Newhall et al., 1993; also see Messick & Mackie, 1989, for a review). Brewer and Miller (1984) argue that reducing category salience allows for the intergroup contact to be person-based and encourages information processing that is interpersonal and self-relevant--a situation likely to induce greater out-group acceptance. By contrast, when category salience is high, contact is category-based and characterized by in-group members viewing outgroup members as depersonalized, interchangeable representatives of their social category--a situation likely to encourage in-group favoritism. When situational features make social category distinctions highly salient in a cooperative context, intergroup interaction will be category-based (Brewer & Miller, 1984; Miller & Harrington, 1990).

A reduction in social category salience should mitigate category-based interaction and decrease in-group bias. Supporting this reasoning, studies show that in-group bias is higher when people differ on two real social categories, such as ethnicity and gender, than when they differ on one category but share another (i.e., crossed categorization; see Brewer, Ho, Lee, & Miller, 1987; Commins & Lockwood, 1978; Islam & Hewstone, 1993, Urban & Miller, 1997). These findings also have been supported by laboratory studies in which two nominal group categories, both equally represented, were crossed (Deschamps, 1977; Deschamps & Doise, 1978; Vanbeselaere, 1987, 1991).

Most relevant to our research, however, is one study (Marcus-Newhall et al., 1993) in which social category memberships were crossed with role assignments to a cooperative team effort. In that study, Marcus-Newhall et al. (1993) assigned members of two equal size groups to either cross-cut or convergent role conditions. In the cross-cut condition, one member from each of the two social categories completed a list of cognitive traits that were associated with a good astronaut and another member from each social category completed a separate list of emotional traits. In the convergent condition, the members of one social category completed the list of cognitive traits and the members of the other social category completed the list of emotional traits. After this role assignment phase, all four members cooperated in producing a team product that was based on the two trait lists. In general, this study showed that compared to convergent role assignments, cross-cut role assignments were associated with reduced in-group bias. A second experiment, however, showed that if the team interaction was constrained, so there was little person-based interaction, no additional benefit from cooperation was produced by cross-cut assignment. Marcus-Newhall and her colleagues concluded that cross-cutting role assignments reduced intergroup bias through a process of reduced category salience, along with an opportunity for personalized interaction among team members. The fact that cross-cutting categories reduces in-group bias through personalization has been supported in a meta-analysis conducted by Urban and Miller (1997).

Because cross-cutting role assignments with social category membership reduces the salience of category boundaries, we reasoned that this strategy may have the potential to reduce intergroup bias among numerical minorities. Researchers have yet to examine whether cooperative role assignments affect the in-group bias of numerical minority and majority groups.

THE IMPORTANCE OF CATEGORY SALIENCE FOR NUMERICAL MINORITIES

Reducing the salience of category boundaries is particularly important for numerical minority groups because both theory (Brewer, 1979; Brewer & Miller, 1984; Duval & Duval, 1983; Duval & Wicklund, 1972; McGuire et al., 1978) and research (McGuire et al., 1978, 1979; Mullen, 1983) suggest that the social category of the numerical minority is more salient than that of the majority. Similarly, research assessing the effects of token status in intergroup settings reveals that tokens have more salient category memberships than nontokens (Cohen & Swim, 1995; Crocker & McGraw, 1984; see Moreland & Levine, 1992, for a review). It is theorized that the perceptual qualities of unequal social category representation direct the members of both the minority and majority to focus their attention on the smaller subgroup (Brewer, 1979; Brewer & Miller, 1984; Duval & Duval, 1983; Duval & Wicklund, 1972), which makes the social category of the minority more salient than that of the majority. When category distinctions are highly salient in an intergroup contact situation, group members are more apt to respond in ways that are category-based (Brewer & Miller, 1984; Hong & Harrod, 1988; Oakes, 1987; Tajfel, 1978; Wilder & Shapiro, 1989a, 1989b). However, there is little direct evidence that category salience is a mediator of the relationship between numerical representation and in-group bias.

Recently, Abrams, Thomas, and Hogg (1990) have argued "that it is the meaning attached to being in a minority" (p. 91), not the proportion of the representation per se, that affects the salience of the minority category. They suggest that group members may perceive that marked differences in numerical representations are associated with varying levels of power or status. In a similar vein, Festinger (1954) argued that "members of minority groups ... should be somewhat less secure in their self-evaluations" (pp. 136-137). We agree with Abrams et al.'s (1990) perspective (see also Mummendey & Simon, 1989; Sachdev & Bourhis, 1984; Simon & Brown, 1987) and assume that when members of a social category are represented by a numerical minority and they compare their group to the numerical majority in the setting, all else being equal (e.g., status), the minority group's numerical disadvantage will often result in comparisons that are more negative for the minority group. We interpret the results of Islam and Hewstone (1993), who found that minority Hindus feel more anxious in intergroup contact situations than do majority Muslims, as supportive of this assumption. Generally, groups that experience greater anxiety in an intergroup situation have more biased attitudes (Stephan & Stephan, 1985; Wilder & Shapiro, 1989a, 1989b).

Several theorists (Abrams et al., 1990; Mummendey & Simon, 1989) argue that the numerical disadvantage associated with minority representation induces greater cohesion among the numerical minorities, which, in turn, leads to stronger feelings of in-group identification toward the group. A few researchers have shown that minorities identify more strongly with their in-group than do majorities (Mummendey & Simon, 1989; Sachdev & Bourhis, 1984; Simon & Brown, 1987), and they reasoned that in-group identification mediates the relation between numerical representation and in-group bias. Although several studies suggest that social category salience and in-group identification are linked (e.g., Hinkle, Fox-Cardamone, Haseleu, Brown, & Irwin, 1996; Hogg, 1992; Struch & Schwartz, 1989; Turner, 1985; van Knippenberg & Ellemers, 1993), to our knowledge, no published work has shown that in-group identification actually mediates the relationship between minority representation and in-group bias.

PRESENT STUDIES

As we noted previously, no study has explicitly crossed role assignment to a cooperative team task with minority and majority membership. Because cross-cutting role assignment in a cooperative contact setting appears to reduce the salience of social category boundaries, we theorized that cross-cutting roles might mitigate in-group bias among numerical minority members, compared to convergent roles.

Therefore, the purpose of the present studies was to examine the moderating effect of cross-cut role assignments on in-group biases among numerical minorities and majorities. As Islam and Hewstone (1993) have advised, we also examined the process through which cross-categorization effects occur. Specifically, in Study 2, we investigated whether social category salience and in-group identification mediate the effect of cross-cut role assignment on in-group bias. Although Marcus-Newhall and her colleagues (1993) found evidence of more personalization in the cross-cut condition, there was no direct evidence that category salience was reduced in the cross-cut condition when compared to the convergent condition.

Using Brewer and Miller's (1984) experimental paradigm, which is an analog of the cooperative interventions used in desegregated schools, our two studies examined the effects of numerical representation and role assignment on intergroup attitudes. The …