11

Gender Outgroup Homogeneity

The Roles of Differential Familiarity, Gender Differences, and Group Size MARK RUBIN University of Newcastle, Australia

MILES HEWSTONE University of Oxford

RICHARD J. CRIPS University of Birmingham

ALBERTO VOCI Università di Padova

ZOË RICHARDS Cardiff University

O

utgroup homogeneity (OH) is the tendency to perceive members of one’s ingroup as being more variable than members of an outgroup. The OH effect has been identified using a variety of different measures (for a review, see Park & Judd, 1990). Measures of perceived group stereotypicality assess the degree to which group members are perceived to possess stereotypic versus counterstereotypic traits. OH is evident if outgroup members are perceived to be more stereotypical of the out-group than in-group members are perceived to be of the ingroup. Measures of perceived intragroup dispersion assess the degree to which group members are perceived to be 203

Yzerbyt-HB-184690619-ch-11.p65

203

8/29/03, 8:45 AM

204

THE PSYCHOLOGY OF GROUP PERCEPTION

dispersed around the group’s mean position. OH is evident if outgroup members are perceived to be less dispersed than in-group members. Measures of perceived intragroup similarity assess the degree to which group members are perceived to be similar to one another within their group. OH is evident if outgroup members are rated as being more similar to one another than ingroup members are to one another. Finally, name-matching measures assess intraand inter-category confusions when attempting to recall matches between individual group members and information to which they have been previously linked (e.g., Lorenzi-Cioldi, Eagly, & Stewart, 1995). OH is evident if more intracategory confusions are made for outgroup members than for ingroup members. The OH effect has been found using a variety of natural social groups, including Australians and Americans (S. A. Haslam, Oakes, Turner, & McGarty, 1995), Hindus and Muslims (Hewstone, Islam, & Judd, 1993), Labour party and Conservative party supporters (Kelly, 1988), young and old people (Linville, Fischer, & Salovey, 1989, Experiment 1), and firemen and funeral employees (Codol, 1984; for reviews, see Devos, Comby, & Deschamps, 1996; Linville, 1998; Mullen & Hu, 1989; Ostrom & Sedikides, 1992; Voci, 2000). Given the pervasive nature of the OH effect among natural groups, one might expect to find that the effect also occurs for gender groups, especially considering that gender provides a primary basis for social categorization (Brewer & Lui, 1989; J. A. Williams, 1984). However, there has been some debate concerning the evidence for gender OH. Some researchers have found small but significant gender OH effects (e.g., Park & Judd, 1990), whereas others have found no significant effects (e.g., Linville et al., 1989, Expt. 3). Two explanations have been put forward for these equivocal findings. First, Linville and colleagues (e.g., Linville et al., 1989) proposed that OH is driven by differences in familiarity with ingroup and outgroup members. They argued that strong gender OH effects are unlikely to occur because there is usually very little difference in the familiarity that people have with men and women. Second, Lorenzi-Cioldi and colleagues (e.g., Lorenzi-Cioldi et al., 1995) proposed that gender OH is perceived by men but that the reverse effect of in-group homogeneity (IH) is perceived by women. Consequently, overall gender OH effects (i.e., the average effects shown by male and female participants) are unlikely to occur because the IH shown by female participants subtracts from the OH shown by male participants to leave a null effect overall. Lorenzi-Cioldi et al. attribute these gender differences in gender OH to status differences between genders. In this chapter, we review the evidence supporting each of these two explanations for null gender OH effects. Specifically, we look at the roles of (a) differential intergroup familiarity and (b) gender differences in explaining null gender OH effects. We then go on to consider how group size might influence

Yzerbyt-HB-184690619-ch-11.p65

204

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

OH. Finally, we report some of our own research investigating the roles of differential familiarity, gender differences, and group size in determining gender OH, and we conclude by discussing a tentative model of OH.

EXPLANATIONS FOR NULL GENDER OH EFFECTS Differential Intergroup Familiarity Linville and colleagues (Linville, 1998; Linville & Fischer, 1993; Linville et al., 1989; Linville & Jones, 1980; Linville, Salovey, & Fischer, 1986) proposed that in most cases people are more familiar with ingroup members than with outgroup members. This differential intergroup familiarity may cause OH for three reasons. First, people have a more diverse and articulated set of experiences with familiar others. Second, people tend to have more incentives for making distinctions among familiar others. Third, people possess a larger sample of exemplars for familiar others and, according to basic statistical principles, greater variability is found among larger samples than among smaller samples. Evidence for the differential intergroup familiarity hypothesis is mixed. Some studies have found supportive evidence (Islam & Hewstone, 1993b; Linville et al., 1989, Experiments 1, 2, & 4), but other studies have found null effects (Askevis-Leherpeux & Bastounis, 1998; E. E. Jones, Wood, & Quattrone, 1981; Park, Ryan, & Judd, 1992) or opposite effects (Oakes, Haslam, Morrison, & Grace, 1995; Ryan & Bogart, 1997). One simple method of testing whether differential familiarity is necessary for OH to occur is to establish whether OH occurs in intergroup situations in which there is no differential familiarity. Minimal groups and gender groups provide two examples of such intergroup situations. Minimal Group Evidence. In the minimal group paradigm (e.g., Tajfel, Billig, Bundy, & Flament, 1971), 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). Importantly, these minimal groups are new and unfamiliar to participants, and group members’ personal identities remain anonymous throughout the experiment. Consequently, participants in the minimal group paradigm are as (un)familiar with the in-group and its members as they are with the out-group and its members.1 Hence, evidence of OH in the minimal group paradigm goes against the idea that differential familiarity is necessary for the occurrence of OH (e.g., Judd & Park, 1988). In 1992, Ostrom and Sedikides reviewed the evidence for OH in the minimal group paradigm and concluded that it was “quite meager” (p. 546). However, since this time, evidence for OH in the minimal group paradigm has grown considerably (Badea & Askevis-Leherpeux, 2002; Brewer & Weber, 1994,

Yzerbyt-HB-184690619-ch-11.p65

205

8/29/03, 8:45 AM

205

206

THE PSYCHOLOGY OF GROUP PERCEPTION

Experiment 1; Brown & Ross, 1977, Experiment 3, as cited in Ostrom & Sedikides, 1992; Ellemers, Spears, & Doosje, 1997, Experiment 1; Judd & Park, 1988; Lorenzi-Cioldi, 1995b, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi, 1997, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi, Deaux, & Dafflon, 1998, Experiments 2 & 3; Mackie, Sherman, & Worth, 1993, Experiment 2; Rubin, Hewstone, & Voci, 2001, Experiment 1; J. W. Sherman, Klein, Laskey, & Wyer, 1998, Experiment 1; Simon & Brown, 1987; Simon & Hamilton, 1994, Experiments 1 & 2; Simon & Mummendey, 1990; Simon & Pettigrew, 1990; Wilder, 1980, as cited in Wilder, 1981; Wilder, 1984, Experiment 1; Wilder & Thompson, 1980; Willis, Rubin, & Hewstone, 2000, Experiments 1 & 2). Relatively few minimal group studies have reported null OH effects (Brewer, Weber, & Carini, 1995, Experiment; Brown & Ross, 1977, Experiment 1, as cited in Ostrom & Sedikides, 1992; Doosje, Ellemers, & Spears, 1995, Experiment 2; Doosje, Spears, & Koomen, 1995, Experiments 1 & 2; Ellemers et al., 1997, Experiment 2; Karasawa & Brewer, 1989, as cited in Ostrom & Sedikides, 1992; Simon, 1990, as cited in Ostrom & Sedikides, 1992). Gender Group Evidence. The differential intergroup familiarity hypothesis also predicts little or no OH for gender groups because in most cases men and women have an approximately equal amount of familiarity with one another. Contrary to this hypothesis, there is now a large body of evidence demonstrating gender OH (Cabecinhas & Amâncio, 1999, Experiment 2; Carpenter & Ostrom, 1985, as cited in Messick & Mackie, 1989; Fiedler, Semin, & Finkenauer, 1993; Frable & Bem, 1985; Lorenzi-Cioldi, 1988; Lorenzi-Cioldi, 1992, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi, 1993; Lorenzi-Cioldi et al., 1995; Mackie, Sherman, & Worth, 1993, Experiment 1; Ostrom, Carpenter, Sedikides, & Li, 1993, Experiment 1; Park & Judd, 1990, Experiment 1; Park & Rothbart, 1982, Experiments 1 & 2; Powlishta, 1995; Young, van Knippenberg, Ellemers, & de Vries, 1999).2 Relatively few experiments have reported null gender OH effects (Beauvais & Spence, 1987; Jackson & Hymes, 1985; Linville et al., 1989, Experiment 3; Linville, Fischer, & Yoon, 1996; Ostrom et al., 1993, Experiment 2; Taylor, Fiske, Etcoff, & Ruderman, 1978, Experiments 2 & 3). In light of the minimal group and gender group evidence cited above, we can conclude that although differential intergroup familiarity may be sufficient to cause OH it is not always necessary (Linville et al., 1989). Given this limitation, the differential intergroup familiarity hypothesis cannot be regarded as providing a convincing explanation of null gender OH effects (e.g., Linville et al., 1989, Experiment 3; Linville et al., 1996).

Gender Differences in Gender OH Lorenzi-Cioldi et al. (1995) proposed that (1) overall gender OH effects are unlikely because (2) there are gender differences in gender OH such that men

Yzerbyt-HB-184690619-ch-11.p65

206

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

show OH and women show no effect or IH. Contrary to Lorenzi-Cioldi et al.’s first prediction, several studies have found overall gender OH to be a relatively reliable phenomenon (Fiedler et al., 1993; Frable & Bem, 1985; Lorenzi-Cioldi, 1993; Mackie et al., 1993, Experiment 1; Ostrom et al., 1993, Experiment 1; Park & Judd, 1990, Experiment 1; Park & Rothbart, 1982, Experiments 1 & 2; Powlishta, 1995; Young, van Knippenberg, Ellemers, & de Vries, 1999). Nonetheless, there is also evidence consistent with Lorenzi-Cioldi et al.’s second prediction that men show OH and women show no effect or IH. In particular, several studies using the name-matching measure of OH have found that men show significantly greater gender OH than women (Cabecinhas & Amâncio, 1999, Experiment 2; Lorenzi-Cioldi, 1988; Lorenzi-Cioldi, 1992, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi, 1993; Lorenzi-Cioldi et al., 1995; Young et al., 1999; see also Frable & Bem, 1985, footnote 4). Furthermore, studies using dispersion measures of OH have found the same pattern of results to be either significant (R. Brown & Smith, 1989) or marginally significant (Linville et al., 1989, Experiment 3; Park & Judd, 1990, Experiment 1). Lorenzi-Cioldi (1998) explains these gender differences in gender OH in terms of differences in group status. Gender and Group Status. People often perceive men to have a higher status than women (Eagly, 1987; Eagly & Wood, 1982; Fiske, 1993; Geis, 1993; Ridgeway & Diekema, 1992; Stewart & Vassar, 2000; Stewart, Vassar, Sanchez, & David, 2000, Experiment 3; Wagner & Berger, 1997). Lorenzi-Cioldi (1998) put forward a variety of social psychological processes to explain how this perceived status difference might account for gender differences in gender OH. We elaborate on three of these processes here. First, Lorenzi-Cioldi (1998, p. 39) proposed that members of high status groups are motivated to enhance the salience of their personal identity in order to claim personal credit for the high status that their group membership entails. This “personalizing strategy” is thought to be a way for group members to legitimize their high status because it implies that their status derives from their personal characteristics rather than from the privileges that their group membership bestows upon them (see also Branscombe, 1998). Hence, members of high status groups should perceive themselves in terms of their personal identity more than members of low status groups do. Some theorists have suggested that the presence of the individuated (personal) self in the ingroup increases perceived ingroup variability relative to perceived outgroup variability (Park & Judd, 1990; Simon, 1993). Hence, greater personal identity salience for members of high status groups may lead them to show greater OH than members of low status groups (Voci, 2000). Second, Lorenzi-Cioldi (1998, pp. 35–36) proposed that people tend to treat members of low status groups in terms of their group membership more than they do members of high status groups. This differential treatment can

Yzerbyt-HB-184690619-ch-11.p65

207

8/29/03, 8:45 AM

207

208

THE PSYCHOLOGY OF GROUP PERCEPTION

lead to differences in social identity salience for high and low status group members. Specifically, social identity salience should be higher for members of low status groups than for members of high status groups. As Voci (2000) noted, increasing social identity salience can lead either to perceived equivalence between ingroup and outgroup variability (see also Mullen & Hu, 1989, p. 235) or, under conditions of ingroup threat, to IH (Simon, 1992b). In either case, enhanced social identity salience for members of low status groups should lead them to show less OH than members of high status groups. Finally, Lorenzi-Cioldi and colleagues (1993, 1998; Lorenzi-Cioldi et al., 1995) proposed that high status groups are generally perceived to be more variable than low status groups. This “low status group homogeneity effect” (see Lorenzi-Cioldi, 1998, p. 34; Stewart et al., 2000) is thought to occur independent of the perceiver’s own group membership. In other words, this effect is predicted to be equally applicable for group members and nonmembers alike. The “low status group homogeneity effect” may occur because people in Western societies equate individuality with high status (Snyder & Fromkin, 1980), and it may account for OH because people are motivated to perceive their ingroup as having a higher status than other groups (Tajfel & Turner, 1979). A Matter of Interpretation. Lorenzi-Cioldi and colleagues (Lorenzi-Cioldi, 1998; Lorenzi-Cioldi et al., 1995) predicted an interaction between participants’ gender (male/female) and target group (ingroup/outgroup) such that men show OH and women show no effect or IH. There are four theoretical interpretations for this interaction, and it is worthwhile considering each in turn in order to appreciate the various confounds that are involved in this type of research. First, it is possible that the interaction between participants’ gender (male/ female) and target group (ingroup/outgroup) reflects a general tendency for men to show greater OH than women, regardless of the target groups involved (Lorenzi-Cioldi et al., 1995; Young et al., 1999). According to this explanation, not only would men be expected to show greater OH than women when the target groups were gender groups, but also when the target groups were minimal groups, occupational groups, national groups, and so on. Relatively few nongender group OH studies have tested for this general gender difference. However, the evidence from these few studies does not support the existence of a general gender difference in OH (Hewstone et al., 1993, Experiments 1 & 2; Simon & Pettigrew, 1990; Wilder, 1990, Experiments 1 & 2; Young, van Knippenberg, Ellemers, & de Vries, 1997). Second, it is possible that the interaction reflects a specific tendency for men to show greater gender OH than women. In other words, men may perceive their gender ingroup to be more variable than their gender outgroup to a greater extent than women. Following Lorenzi-Cioldi (1998), this tendency may be due to men’s enhanced personal identity salience and/or women’s enhanced social identity salience.

Yzerbyt-HB-184690619-ch-11.p65

208

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

Third, it is possible that the interaction reflects a general tendency for people to perceive differences in intragender variability: People may tend to perceive men to be more variable than women. Following Lorenzi-Cioldi (1998), this general tendency may represent a “low status group homogeneity effect.” Finally, it is possible that the interaction reflects real differences in intragender variability: Men may actually be more variable than women on the given dimensions (Brauer, 2001). For example, Brown and Smith (1989) suggested this possibility when discussing their findings that male academic staff showed OH and female academic staff showed IH on dimensions related to academic ability. They proposed that sexist employment procedures may have led to standards of selection being higher for female staff than for male staff. These differential selection standards may have led to less actual variability in ability among female staff compared to male staff. Hence, men perceiving OH and women perceiving IH could be explained as accurate perceptions of objective reality rather than as social psychological biases (for similar interpretations, see Brown & Wootton-Millward, 1993; Guinote, Judd, & Brauer, 2002; Simon, Glässner-Bayerl, & Stratenwerth, 1991). As Brauer and Judd (2000; see also Bartsch & Judd, 1993; Brauer, 2001) pointed out, the last three interpretations are confounded because the interaction between participants’ gender (male/female) and target group (ingroup/ outgroup) is statistically identical to the main effect of target group when target group is defined as “men/women” rather than as “ingroup/outgroup”. Although measures of actual variability may be used to rule out the last interpretation, the second and third interpretations are particularly difficult to disentangle because no gender-neutral control group is available to provide objective (i.e., unaffiliated) judgments of male and female variability. Since participants are themselves always either male or female, they always have the potential to categorize male and female target groups subjectively (i.e., as “ingroup” and “outgroup”) as well as objectively (i.e., as “men” and “women”). Researchers need to consider other ways of determining the extent to which the interaction between participants’ gender and “ingroup/outgroup” target group represents differences between participants, perceived differences between “men/women” target groups, or a combination of both (see Brauer & Judd, 2000). The Effects of Group Status on OH. Lorenzi-Cioldi’s (1998) predictions are applicable to all groups that exist within a status hierarchy, not just gender groups. Consequently, it is instructive to consider the effects of group status on OH in order to evaluate these predictions. Lorenzi-Cioldi (1998, p. 42) acknowledged that OH often occurs between groups of equal status, and he took this “default” OH effect into account when he made his empirical predictions concerning the degree of OH shown by unequal status groups. Specifically, he assumed that the processes associated with high status groups are liable to combine with default OH processes to produce a relatively large OH effect for members

Yzerbyt-HB-184690619-ch-11.p65

209

8/29/03, 8:45 AM

209

210

THE PSYCHOLOGY OF GROUP PERCEPTION

of high status groups. In contrast, the processes associated with low status groups are liable to counteract the default OH processes to produce no effect or an IH effect for members of low status groups. Interestingly, Simon and Hamilton (1994, p. 704) arrived at the opposite prediction to Lorenzi-Cioldi (1998) based on the principles of social identity theory (Tajfel & Turner, 1979). They argued that people are motivated to perceive themselves in terms of a positive social identity and because high status groups provide a more positive social identity than low status groups, members of high status groups should be more motivated to perceive themselves in terms of their social identity than members of low status groups. It follows that members of high status groups should show less OH than members of low status groups.3 The available evidence provides more support for Lorenzi-Cioldi’s (1998) prediction than for Simon and Hamilton’s (1994) prediction. Members of high status groups tend to perceive greater OH than members of low status groups (Brewer & Weber, 1994; Boldry & Kashy, 1999; Cabecinhas & Amâncio, 1999, Experiment 1; Lorenzi-Cioldi, 1995b, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi et al., 1998, Experiments 1, 2, & 3; Sedikides, 1997, Experiments 1 & 2; van Knippenberg, Blaauw, & Vermunt, 1996, as cited in Young et al., 1999; Voci & Capozza, 1999), although some studies show no difference in OH between members of high and low status groups (Kashy, 1988, as cited in Boldry & Kashy, 1999), and other studies have found the reverse effect (Brauer, 2001; Judd, Park, Ryan, Brauer, & Kraus, 1995; Pickett & Brewer, 2001). In summary, consistent with Lorenzi-Cioldi (1998), the evidence shows that members of high status groups tend to perceive greater OH than members of low status groups. Consequently, the finding that men show greater gender OH than women may be attributable to perceived intergender differences in status.

THE EFFECTS OF GROUP SIZE ON OH Group status and group size are often confounded in the real world such that majority groups are perceived to have a higher status than minority groups (Farley, 1982; Sachdev & Bourhis, 1984, 1991). However, the relationship between group size and group status is not always a positive one. In some cases, a minority group may be perceived as an elite group that possesses higher status than a majority group (e.g., white people in apartheid South Africa). Furthermore, there is some evidence that people may automatically assume high status minimal groups to be numerically smaller than low status minimal groups (Lorenzi-Cioldi, 1995b, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi & Dafflon, 1996, as cited in Lorenzi-Cioldi, 1998). Consequently, it is appropriate to look at the effects of group size independent of group status as well as to consider the possibility of an interaction between these two factors.

Yzerbyt-HB-184690619-ch-11.p65

210

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

Theorists agree on the prediction that members of majority groups should show greater OH than members of minority groups, who should show IH (Brewer, 1993; Brown & Smith, 1989; Mullen & Hu, 1989; Simon, 1992). Certainly, the evidence supports the prediction that majority group members show greater OH than minority group members (Badea & Askevis-Leherpeux, 2002; Brewer & Weber, 1994, Experiment 1; Brown & Smith, 1989; Guinote, 2001; Hewstone et al., 1993, Experiments 1 & 2; Simon & Brown, 1987; Simon & Hamilton, 1994, Experiments 1 & 2; Simon & Mummendey, 1990; Simon & Pettigrew, 1990) and that minority group members tend to show IH (Brown & Smith, 1989; Guinote, 2001; Simon & Brown, 1987; Simon, Glässner-Bayerl, & Stratenwerth, 1991; Simon & Hamilton, 1994, Experiment 2; Stephan, 1977; Verkuyten & Kwa, 1996; for a meta-analytic review, see Mullen & Hu, 1989). However, there is little consensus about why this group size effect occurs. Simon and colleagues (Simon, 1992b; Simon & Brown, 1987; Simon & Pettigrew, 1990) proposed that minority group membership poses a threat to self-esteem, and that this threat motivates minority group members to enhance the positivity of their social identity. IH is thought to be one means of enhancing social identity because it indicates that the ingroup provides more solidarity and social support than the outgroup. Brewer (1991, 1993) proposed that social identity represents a compromise between two opposing needs: the need for assimilation and the need for differentiation. Majority group members have a greater need for intraingroup differentiation than minority group members because majority groups are, by definition, more inclusive. Consequently, majority group members should be motivated to perceive ingroup heterogeneity (i.e., OH) more than minority group members. In contrast, minority group membership should arouse the need for intra-in-group assimilation, leading to IH. Several theorists have proposed that category salience is greater for members of minority groups than for members of majority groups because minority groups are more visible in the social context (Kanter, 1977; W. J. McGuire & McGuire, 1982; Mullen, 1991; Pichevin & Hurtig, 1996; Simon & Hamilton, 1994; Taylor et al., 1978; van Twuyver & van Knippenberg, 1999). Consequently, social identity salience should be higher for minority group members than for majority group members, leading to less OH for minority group members. Finally, some theorists have suggested that majority groups are generally perceived to be more variable than minority groups (Bartsch & Judd, 1993; Judd, Ryan, & Park, 1991; Messick & Mackie, 1989; Park & Rothbart, 1982). Like the “low status group homogeneity effect,” this “minority group homogeneity effect” (see Park & Rothbart, 1982, p. 1052) is thought to occur independent of the perceiver’s own group membership. The “minority group homogeneity effect” may occur because, on purely statistical grounds, greater variability is observed in large samples than in small samples, and the effect may account for OH because people tend to perceive their ingroup as being numerically larger than their outgroup (Lorenzi-Cioldi, 1997, as cited in

Yzerbyt-HB-184690619-ch-11.p65

211

8/29/03, 8:45 AM

211

212

THE PSYCHOLOGY OF GROUP PERCEPTION

Lorenzi-Cioldi, 1998; Lorenzi-Cioldi, 1995b, as cited in Lorenzi-Cioldi, 1998; Simon & Mummendey, 1990; Willis et al., 2000, Experiment 1).

The Interactive Effects of Group Status and Group Size on OH As we have seen, the evidence indicates that members of high status groups and members of majority groups show the greatest OH effects. But how do group status and group size interact with one another to influence OH? As mentioned previously, Simon and Hamilton (1994) proposed that members of high status groups are more motivated to think of themselves in terms of their social identity than are members of low status groups. However, Simon and Hamilton also proposed that this difference in social identity salience is qualified by group size. Work by Ditto and Jemmott (1989) demonstrated that evaluations of characteristics become more polarized when the characteristics are perceived to be scarce. On the basis of this “scarcity principle”, Simon and Hamilton proposed that group status differences in social identity salience are enhanced for minority group members. Specifically, members of the high status minority should show the greatest IH and members of the low status minority should show the greatest OH. Simon and Hamilton (1994, Experiment 2) manipulated group status and group size orthogonally in a minimal groups experiment and measured OH in terms of perceived intragroup similarity. Contrary to their predictions, they found that members of the high status majority group showed OH, members of the high status minority group showed IH, and members of the low status majority and minority groups showed no significant effects. A subsequent minimal group experiment using a name-matching measure of OH failed to replicate these results (Lorenzi-Cioldi & Dafflon, 1996, as cited in Lorenzi-Cioldi, 1998; see also Ellemers & van Rijswijk, 1997). Consequently, firm conclusions cannot be drawn with respect to interactions between group status and group size.

SOME CONTRIBUTIONS OF OUR OWN RESEARCH In our own research, we were interested in determining the roles of differential familiarity, participants’ gender, and group size in determining gender OH on dispersion measures. Previous research has demonstrated gender differences in gender OH using name-matching measures of OH (e.g., Cabecinhas & Amâncio, 1999, Experiment 2; Lorenzi-Cioldi, 1993; Lorenzi-Cioldi et al., 1995; Young et al., 1999). However, to our knowledge, only R. Brown and Smith (1989) found clear evidence for this gender difference using a dispersion measure of OH (the “range” measure). One problem with Brown and Smith’s study is that gender and group size are confounded in its design. Brown and Smith found that majority males perceived greater OH than minority females. This result

Yzerbyt-HB-184690619-ch-11.p65

212

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

may be interpreted as evidence for either a gender difference in gender OH (men show greater OH than women) or as a group size difference in OH (majority group members show greater OH than minority group members). In our own research, we used the range measure to investigate gender OH, but we also unconfounded gender and group size in order to analyze each variable separately as well as in interaction with one another (Hewstone & Crisp, 2000; Hewstone, Crisp, Contarello, Conway, Voci, Marletta, et al., 2000; Hewstone, Crisp, Richards, Voci, & Rubin, 2000; Rubin et al., 2001; Rubin, Richards, & Hewstone, 2000; see also Hewstone, Martin, Hammer-Hewstone, Crisp, & Voci, 2001). The first part of our research program provided a competitive test of the gender and group size explanations of Brown and Smith’s (1989) findings. First, we attempted to replicate Brown and Smith’s findings using the same gender-group size relationship that they used. In this experiment (Hewstone & Crisp, 2000, Experiment 2), we asked male and female police officers to judge each group’s dispersion using the range measure. Although we found no overall OH effect (i.e., a main effect of target group), we found a two-way interaction between participants’ gender (male/female) and target group (ingroup/ outgroup). Consistent with Brown and Smith’s results, majority group members (male participants) showed OH and minority group members (female participants) showed IH. Hence, we replicated Brown and Smith’s findings using male majority and female minority groups. Measures of familiarity revealed that majority group members were more familiar with the ingroup than the outgroup and that the reverse was true for minority group members. This pattern of findings makes sense because, all other things being equal, people have a higher probability of interacting with majority group members than with minority group members. Importantly, however, although the familiarity results followed the same pattern as the OH results, familiarity did not mediate OH. Like Brown and Smith’s (1989) study, our first experiment confounded group size and gender in its design. Consequently, the effects that we found could have been due to either gender or group size. In order to disentangle these variables, we conducted a second experiment (Hewstone & Crisp, 2000, Experiment 1) using a population in which women comprised the majority and men comprised the minority. Specifically, we sampled student nurses instead of police officers in order to reverse the gender-group size relationship. Again, although we found no overall OH effect, we found a two-way interaction between participants’ gender and target group. Majority group members (female participants) showed OH and minority group members (male participants) showed IH. Hence, in this second experiment, we replicated Brown and Smith’s group size finding using female majority and male minority groups. Once again, although the familiarity results followed the same pattern as the OH results, familiarity did not mediate OH. Our first two experiments (Hewstone & Crisp, 2000) showed that Brown

Yzerbyt-HB-184690619-ch-11.p65

213

8/29/03, 8:45 AM

213

214

THE PSYCHOLOGY OF GROUP PERCEPTION

and Smith’s (1989) findings were most likely to be due to the influence of group size rather than group gender. However, we were also interested in investigating how these two factors interacted with one another. Consequently, we continued our program of research by unconfounding group gender and group size within the same experimental design. In the first of these experiments (Hewstone, Crisp, Richards, et al., 2000, Experiment 1), we sampled male and female university students who belonged to departments that contained either a majority of male students (earth sciences, engineering), a majority of female students (education, psychology), or an approximately equal number of male and female students (biochemistry, mathematics). Participants judged the dispersion of male and female students in their own department. We found a three-way interaction between department (male majority/female majority/parity), participants’ gender (male/female), and target group (ingroup/outgroup). In the male majority and parity departments male participants showed OH and female participants showed no significant effect. In the female-majority departments participants showed no significant effects. Hence, we confirmed Lorenzi-Cioldi’s (1998) prediction of a gender difference in gender OH, but only when participants belonged to male majority or parity departments. Although we found similar results in an additional experiment (Hewstone, Crisp, Richards, et al., 2000, Experiment 3), we failed to replicate the interaction between participants’ gender and group size in a third experiment (Rubin et al., 2001, Experiment 3). In this latter experiment, the typical gender difference in gender OH was not qualified by group size. Contrary to previous research (Brown & Smith, 1989; Linville et al., 1989, Experiment 3), participants in our first experiment (Hewstone, Crisp, Richards, et al., 2000, Experiment 1) rated themselves as being more familiar with their gender ingroup than with their gender outgroup. Again however, we found no evidence of a significant relationship between familiarity and OH. In the experiments reported above (Hewstone, Crisp, Richards, et al., 2000, Experiments 1 & 3; Rubin et al., 2001, Experiment 3), we collected data from students individually. In a further experiment (Hewstone, Crisp, Richards, et al., 2000, Experiment 2), we reasoned that the group size variable would have a greater impact if we collected data from participants in lecture theaters in which members from both gender groups were present. We thought that this experimental setting would enhance the salience of the group size variable leading to stronger group size effects. The results for this experiment were consistent with our predictions. As in our previous experiments, male participants in male majority departments showed an OH effect. However, we also found that female participants in male majority departments showed a significant IH effect. Hence, increasing the salience of the group size variable led to IH for minority female participants (cf. Hewstone & Crisp, 2000, Experiments 1 & 2). Once again, differential familiarity was unrelated to OH.

Yzerbyt-HB-184690619-ch-11.p65

214

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

For the sake of completeness, we should also mention two experiments in which we simply asked students to rate the dispersion of male and female university students in general, as opposed to male and female students in their department (Rubin et al., 2001, Experiment 2; Rubin et al., 2000). Although we found overall OH effects, we failed to find a gender difference in either of these experiments. We shall discuss a possible explanation for these null findings a little latter. The last stage of our research program investigated the effects of group size in more detail. Hewstone, Crisp, Contarello, et al. (2000, Experiment 1) sampled male and female staff members at a British university who belonged to departments that contained either a skewed majority of male staff (92% male) or a tilted majority of male staff (73% male; Kanter, 1977). Participants judged the dispersion of other male and female staff members in their department. We found a three-way interaction between department (male-tilted/male-skewed), participants’ gender (male/female), and target group (ingroup/outgroup). Men showed OH and women showed IH, and these effects were more pronounced in male-skewed departments than in male-tilted departments. In other words, the gender difference in gender OH increased as a function of group size. Hewstone, Crisp, Contarello, et al. (2000, Experiment 2) extended this research by sampling male and female staff members at an Italian university. Some participants belonged to departments that contained either a skewed majority of male staff (92% male) or a tilted majority of male staff (80% male) as in the first experiment. Other participants belonged to a department that contained either a relatively balanced proportion of male staff (60% male) or a tilted majority of female staff (31% male). Again, we found a three-way interaction between department, participants’ gender, and target group. In the male-skewed departments, men showed OH and women showed IH. No significant effects were found in the other three types of department. However, we were able to show that perceived variability for men and women varied as a linear function of group proportions: The OH shown by men tended to disappear as they lost their ingroup majority status, and the IH shown by women tended to disappear as they lost their ingroup minority status. Hence, these final two experiments showed that gender differences in gender OH are most likely to occur under skewed group size conditions. To summarize our research findings, we replicated Brown and Smith’s (1989) finding that male majorities show OH and female minorities show IH on the dispersion measure (Hewstone & Crisp, 2000, Experiment 2). However, we expanded on Brown and Smith’s research by showing that this result was primarily due to the group size factor rather than the gender factor (Hewstone & Crisp, 2000, Experiment 1). In a second series of experiments, we unconfounded gender and group size in order to analyze these factors in interaction with one another. We found that gender differences in gender OH only occurred in male majority contexts. Specifically, only men in male majorities showed OH

Yzerbyt-HB-184690619-ch-11.p65

215

8/29/03, 8:45 AM

215

216

THE PSYCHOLOGY OF GROUP PERCEPTION

(Hewstone, Crisp, Richards, et al., 2000, Experiments 1, 2, & 3). Women in male majorities were more likely to show IH, especially when group size was particularly salient (Hewstone, Crisp, Richards, et al., 2000, Expt. 2) or skewed (Hewstone, Crisp, Contarello, et al., 2000, Experiments 1 & 2). Hence, we qualified Lorenzi-Cioldi et al.’s (1995) predictions by showing that group size moderated gender differences in gender OH. Taken together, our research findings suggest that group size is a more powerful moderator of gender OH than participants’ gender, and that gender differences in gender OH are most likely to occur when men are in the majority and women are in the minority.

CONCLUSIONS Explaining Null Gender OH Effects OH has been found in a wide variety of natural groups. However, there has been some debate over the existence of OH in gender groups (Linville, 1998; Linville et al., 1989; Lorenzi-Cioldi et al., 1995; Park & Judd, 1990). Linville et al. (1989) proposed that gender OH is unlikely to occur because people are equally familiar with members of both gender groups (Linville et al., 1989). However, the extant evidence shows that differential intergroup familiarity is not necessary for OH to occur. Moreover, our own research suggests that even when differences in intergender familiarity do occur, they are unrelated to gender OH effects (Hewstone & Crisp, 2000, Experiments 1 & 2; Hewstone, Crisp, Richards, et al., 2000, Experiments 1 & 2). Consequently, the differential intergroup familiarity hypothesis does not provide a convincing explanation for null gender OH effects. Lorenzi-Cioldi et al. (1995) proposed that overall gender OH effects are unlikely to occur because OH and IH effects shown by men and women counteract one another in the overall analysis. Contrary to Lorenzi-Cioldi et al., there is evidence for both overall gender OH and gender differences in gender OH. Hence, although the proposed gender difference in gender OH appears to be a reliable phenomenon, it cannot be used to explain null gender OH effects. In summary, neither Linville et al. (1989) nor Lorenzi-Cioldi et al. (1995) provide satisfactory theoretical explanations for null gender OH effects. Indeed, one may question whether theoretical explanations are even necessary. Given that null gender OH effects only occur in a relatively small number of studies, and given that the OH effect tends to be a small effect (Mullen & Hu, 1989; Park & Judd, 1990), it is perhaps most parsimonious to explain null gender OH effects as Type II statistical errors.

Moderators of the Gender Difference in Gender OH Our research supports Lorenzi-Cioldi et al.’s (1995) prediction of a gender difference in gender OH. However, our findings suggest that this gender differ-

Yzerbyt-HB-184690619-ch-11.p65

216

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

ence is moderated by three factors: group size, type of dependent measure, and target group description. First, we found that gender differences in gender OH tended to be moderated by group size. Specifically, we found that men only tended to show OH when their gender was in the majority (Hewstone, Crisp, Contarello, et al., 2000, Expts. 1 & 2; Hewstone, Crisp, Richards, et al., 2000, Experiments 1, 2, & 3) and women only tended to show IH when their gender was in a salient minority (Hewstone, Crisp, Contarello et al., 2000; Hewstone, Crisp, Richards et al., 2000, Experiment 2). We believe that membership in male majority groups enhances personal identity salience for men and social identity salience for women (Voci, 2000). Second, with respect to type of dependent measure, gender differences in gender OH tend to have been found using name-matching measures (Cabecinhas & Amâncio, 1999, Experiment 2; Lorenzi-Cioldi, 1988; Lorenzi-Cioldi, 1992, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi, 1993; Lorenzi-Cioldi et al., 1995; Young et al., 1999; see also Frable & Bem, 1985, footnote 4). However, together with Brown and Smith (1989), our work has shown that this gender difference is equally apparent using the range measure of dispersion. Interestingly, studies using measures of intragroup similarity (Mackie et al., 1993, Experiment 1) or stereotypicality (Park & Judd, 1990, Experiment 1; Park & Rothbart, 1982, Experiments 1 & 2) have failed to find gender differences in gender OH. The obvious factor that links the name-matching and dispersion measures and that separates them from the similarity and stereotypicality measures is that the former two measures require participants to respond in terms of single group members whereas the latter two measures require participants to respond in terms of the group conceived as a whole. We would argue that measures specifying single group members enhance personal identity salience (Voci, 2000), and that men accept this enhancement more than women because they wish to legitimize their high status (Lorenzi-Cioldi, 1998). Finally, the way in which target groups are described also seems to be a potentially important moderator. Consistent with previous research (Brown & Smith, 1989), we found that men showed greater gender OH than women when targets were described as individuals whom participants could relate to on a personal basis (e.g., “male and female students in your department”; Hewstone, Crisp, Richards, et al., 2000, Experiments 1, 2, & 3; Rubin et al., 2001, Experiment 3). However, this gender difference disappeared when targets were described as more abstract social categories that were unlikely to elicit interpersonal comparisons (e.g., “male and female students”; Rubin et al., 2001, Experiment 2; Rubin et al., 2000). Again, we would argue that the potential for interpersonal comparisons enhances personal identity salience, and that men are more prone to this enhancement than women (Lorenzi-Cioldi, 1998). It is interesting to note that some previous research has investigated the influence of target group descriptions on OH. Park and Rothbart (1982, Experiment 1) manipulated target group descriptions (“women [men] as a whole”

Yzerbyt-HB-184690619-ch-11.p65

217

8/29/03, 8:45 AM

217

218

THE PSYCHOLOGY OF GROUP PERCEPTION

versus “women [men] in your class”) but found no significant difference between these conditions using a stereotypicality measure of OH. Lorenzi-Cioldi et al. (1995) described male and female targets as either students at the participants’ own university (superordinate ingroup), employees (superordinate outgroup), or persons (control condition). They found the typical gender difference in gender OH whereby male participants showed OH and female participants showed no significant effect. However, this gender difference was qualified by condition such that it was greatest when targets were described as students at the participants’ own university and nonsignificant when targets were described as persons. Hence, consistent with our own results, the gender difference was accentuated when target groups were described in ways that triggered interpersonal as well as intergroup comparisons.

Summary Overall gender OH is a relatively reliable phenomenon. However, there are gender differences such that men tend to show OH and women tend to show no effect or IH. Although we cannot rule out the possibility of a “low status group homogeneity effect,” we believe that this gender difference in gender OH may arise because men have a higher status than women, and members of high status groups tend to have higher personal identity salience and/or lower social identity salience than members of low status groups. Given a certain degree of social identity salience, enhanced personal identity salience should lead to greater OH because the presence of the personal self in the ingroup increases perceived ingroup variability (Park & Judd, 1990; Simon, 1993). In contrast, enhanced social identity salience should lead either to no difference in perceived ingroup/outgroup variability or, under conditions of ingroup threat, to IH (Mullen & Hu, 1989; Simon, 1992b; Voci, 2000). Notably, the gender difference in gender OH is more pronounced under conditions that enhance personal identity salience, namely (1) membership in a majority group, (2) measures that make reference to single group members, and (3) target group descriptions that encourage interpersonal comparisons. We believe that the enhancing personal identity may exacerbate the gender difference in gender OH because men are more motivated to personalize their ingroup’s status.

A MODEL OF OH We conclude by outlining a tentative model of OH and discussing its theoretical implications. Our model of OH assumes that both personal and social identity must be salient in order for OH to occur. Social identity must be salient because perceptions of “outgroup homogeneity” require that people categorize them-

Yzerbyt-HB-184690619-ch-11.p65

218

8/29/03, 8:45 AM

GENDER OUTGROUP HOMOGENEITY

selves as ingroup members and categorize others as outgroup members. Personal identity must be salient in order to create greater variability in the ingroup than the outgroup (Park & Judd, 1990; Simon, 1993). It is important to note that, consistent with traditional theorizing (e.g., Tajfel & Turner, 1979; Turner, 1987a), our model assumes that personal identity and social identity cannot be salient at the same time. However, we note three points that support the possibility of an interactive relationship between personal and social identity across time. First, situational factors may have independent influences on personal and social identity salience such that the potential for the salience of each type of identity varies orthogonally from situation to situation. Second, people may be able to switch relatively rapidly between personal identity and social identity within the same situation (Spears, Doosje, & Ellemers, 1999, p. 78). Third, both personal and social identity concerns may influence intergroup perception and behavior (R. Brown & Turner, 1981, p. 65). Taking these three points together, we believe that it is appropriate to consider personal and social identity as having independent and interactive influences on intergroup perception (and behavior) in any given situation. In particular, we predict that OH will occur when situational factors increase the potential salience of both personal and social identity. Our model also attempts to resolve the apparent contradiction between Lorenzi-Cioldi’s (1998) predictions and those based on social identity theory (e.g., Simon & Hamilton, 1994). Consistent with Lorenzi-Cioldi (1998), we predict that members of high status groups will use a personalizing strategy and show greater OH than members of low status groups. However, we also predict that a second set of strategies comes into play when group members make judgements of group dispersion along dimensions that are directly related to the intergroup status hierarchy. Specifically, members of high status groups will be tempted towards IH as a strategic means of consolidating their in-group’s position along the status dimension. In contrast, members of low status groups will perceive greater in-group variability in order to distance their personal self from the in-group’s relatively negative position (Doosje, Spears et al., 1995). This personal dissociation strategy should lead to OH (Rubin et al., 2001). Hence, our model assumes that members of high status groups show OH on status-unrelated dimensions as the consequence of a “personalizing strategy”, and that members of low status groups show OH on status-related dimensions as the consequence of a “dissociation strategy”. We believe that an important avenue for future research will be to investigate dimension type as a potential moderator of the effects of group status on OH.

AUTHOR NOTE Part of this chapter was prepared while Miles Hewstone was a Fellow at the Center for Advanced Study in the Behavioral Sciences, Stanford. He gratefully acknowledges the

Yzerbyt-HB-184690619-ch-11.p65

219

8/29/03, 8:45 AM

219

220

THE PSYCHOLOGY OF GROUP PERCEPTION

financial support provided by the William and Flora Hewlett Foundation. Correspondence concerning this chapter should be addressed to Mark Rubin at the School of Behavioral Sciences, University of Newcastle, Callaghan, NSW 2308, Australia. Tel: +61 (0)2 4921 6706. Fax: +61 (0)2 4921 6980. E-mail: [email protected]

NOTES 1. Linville et al. (1989, p. 185) proposed that even when there is little or no explicit difference in intergroup familiarity there is still likely to be an implicit difference in intergroup familiarity because people always have privileged knowledge of at least one in-group member: the self. However, in order to explain OH, Linville et al. rely on the additional assumption that the self is represented by more exemplars than other individuals. Consequently, we will treat this “self hypothesis” separately from the “differential familiarity with others” hypothesis. 2. Some of these gender studies found an OH effect that is restricted to male participants (e.g., Cabecinhas & Amâncio, 1999, Experiment 2; Lorenzi-Cioldi, 1992, as cited in Lorenzi-Cioldi, 1998; Lorenzi-Cioldi et al., 1995). However, unless it is assumed that for some reason differences in intergroup familiarity are restricted to male participants, even these limited OH effects provide evidence against the necessity of differential intergroup familiarity. Also, note that Brown and Smith’s (1989) study is not cited in this list of evidence because this study confounded gender with group size, and although differential intergroup familiarity is unlikely to have covaried with gender it is possible that it covaried with group size. 3. Simon and Hamilton (1994, p. 704) originally made their predictions with respect to self-stereotyping but later (p. 709) related these predictions to OH.

Yzerbyt-HB-184690619-ch-11.p65

220

8/29/03, 8:45 AM