The role of categorization and in-group norms in judgments of groups and their members

Transcription

1 See discussions, stats, and author profiles for this publication at: The role of categorization and in-group norms in judgments of groups and their members Article in Journal of Personality and Social Psychology October 1998 Impact Factor: 5.08 DOI: // CITATIONS 103 READS 29 4 authors: Jose Marques University of Porto 45 PUBLICATIONS 2,117 CITATIONS Dominic Abrams University of Kent 364 PUBLICATIONS 9,445 CITATIONS SEE PROFILE SEE PROFILE Darío Páez Universidad del País Vasco / Euskal Herriko 336 PUBLICATIONS 2,098 CITATIONS Cristina Martinez-Taboada Universidad del País Vasco / Euskal Herriko 14 PUBLICATIONS 227 CITATIONS SEE PROFILE SEE PROFILE All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately. Available from: Jose Marques Retrieved on: 10 May 2016

2 Journal uf Personality and Social Psychology I<»8, Vol. 75, No. 4, Copyright i998 by the American Psychological Association, Inc /98/$ 3.00 The Role of Categorization and In-Group Norms in Judgments of Groups and Their Members Jose M. Marques University of Porto Dominic Abrams University of Kent at Canterbury Dario Paez and Cristina Martinez-Taboada University of the Basque Country Four minima! group experiments tested the prediction that judgments of groups and then- members reflect evaluations made simultaneously but independently at the within-group and intergroup levels. On the basis of self-categorization theory and social identity theory, it was predicted that group members seek both intergroup distinctiveness and legitimization of in-group norms. In Experiments 1-3, membership (in-group, out-group), status of group members (modal, deviant), and either accountability to in-group or to out-group or salience of group norms were varied. Accountability and norm salience increased derogation of out-group normative (in-group deviant, out-group modal) and upgrading of in-group normative (in-group modal, out-group deviant) members. In Experiment 4, within-group differentiation reinforced in-group identification. These findings suggest that subjective group dynamics operate to bolster social identity when people judge modal and deviant in-group and out-group members. Historical and sociological research suggests detection and punishment of deviant group members, such as occurs in witchhunting or political purges, helps define the group and delimit the behavior and the beliefs of its members (Hamilton & Rauma, 1995; Yamagishi, 1995). Differentiation among group members as a function of their level of conformity to in-group norms has long been recognized in the context of small groups (cf. Cartwright & Zander, 1968). However, it may be less obvious how intragroup normative differentiation functions operate in the context of larger social categories. Research using the minimal group paradigm and recent research examining self-categorization theory (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987) has focused on the cognitive bases of membership of such categories. This research typically shows that social categorization is sufficient to result in perceptions of substantial intragroup similarity and intergroup difference, even when judges have no direct information about individual group members. People also Jose M. Marques, Department of Psychology and Educational Sciences, University of Porto. Porto, Portugal; Dominic Abrams, Department of Psychology, University of Kent at Canterbury, Canterbury, United Kingdom; Dario Paez and Cristina Martinez-Taboada, Department of Psychology, University of the Basque Country, San Sebastian, Spain. We are grateful to Marilyn Brewer for valuable comments on an earlier version of this article and to Lome Hulbert, Barbara Masser, Pam Maras, and David Vernon for their assistance with Experiment 3. Correspondence concerning this article should be addressed to Jose M. Marques, FPCE-UP, R. do Campo Alegre, 1055, P-4150 Porto, Portugal; to Dominic Abrams, Department of Psychology, University of Kent at Canterbury, CT2 7NP, United Kingdom; or to Dario Paez, Departamento de Psicologia Social y Metodologia, Universidad del Pais Vasco, Avda. Tolosa 70, San Sebastian 20009, Spain. Electronic mail may be sent to marques@psi.up.pt, d.abrams@ukc.ac.uk, or pspparod@ss.ehu.cs. show evaluative bias in favor of in-group members even when there is no information about the behavior of in-group and outgroup members. On balance, findings are also consistent with the conclusion that one function of social category differentiation is to maintain and to enhance social identity (Abrams & Hogg, 1990; Tajfel & Turner, 1979). There has been debate over the extent to which social identity is maintained principally through cognitive differentiation alone (e.g., Doise & Sinclair, 1973; Hogg & Abrams, 1993; McGarty, Turner, Oakes, & Haslam, 1993) and the relative importance of emotional and evaluative factors (e.g., Crocker, Blaine, & Luhtanen, 1993; Crocker & Luhtanen, 1990; Levine & Moreland, 1994). The conformity literature and the social identity literature suggest that normative differentiation among individuals (with respect to an in-group norm) or category differentiation among individuals (with respect to their category membership) can both affect evaluations of group members (Hogg, 1992, 1996). In this article, we consider the situation in which members' category membership and behavior are consistent or inconsistent with one another. We explore the idea that people subjectively exert control over the evaluative implications of their group memberships by maintaining differentiation both between and within groups. Specifically, we propose that these "subjective group dynamics" involve the simultaneous operation of category differentiation and normative differentiation. We believe that even when participants have no interpersonal relationship with any individual group members, they are motivated to sustain a psychological representation of a cohesive, well-defined, and normatively legitimated group. In doing so, they may be able to bolster their own sense of subjective reality and selfworth. In this vein, category differentiation and normative differentiation operate jointly. The former establishes the category 976

3 CATEGORY AND NORMATIVE DIFFERENTIATION 977 membership of group members (e.g., male or female), whereas the latter establishes the extent to which individual group members adhere to category norms (e.g., masculine and feminine traits or behaviors). In general, category membership and behavior tend to be positively correlated (Tajfel & Turner, 1979). For example, in Europe, Catholic parties are likely to vote against abortion. However, there are also many circumstances in which people's actions are inconsistent with the behavioral standards of their group. For instance, a member of a Catholic party may support the legalization of abortion, against the party's policy or ideology, and in line with that of an opposing party. One aim of the present research is to investigate how perceivers react to inconsistencies between category membership and normative behavior. Our general assumption is that perceivers will attempt simultaneously to sustain category differentiation and to seek legitimacy for in-group norms. As a result of these dynamics, they will make derogatory judgments of in-group deviants and positive judgments of out-group members who endorse in-group norms. That is, as Billig (1985) proposed, perceivers make particularistic judgments about individual category members that may contrast with judgments about the categories as a whole. The Development of Intergroup- and Intragroup -Level Judgments Despite the empirical reliability of in-group bias and intergroup differentiation processes (e.g., Mullen, Brown, & Smith, 1992), research on long-standing and natural groups indicates that intergroup differentiation is moderated by other factors (e.g., Brown et al., 1992; Levine & Moreland, 1994). For example, previous research has shown that individuals downgrade unlikable in-group members more strongly than outgroup members sharing the same (unlikable) characteristics. This pattern of responses arises in parallel with in-group bias, in both intergroup and intragroup comparison settings, when comparison dimensions are relevant to social identity or to intergroup differentiation (Marques, 1990; Marques, Yzerbyt, & Leyens, 1988) and when participants identify with their ingroup, but not otherwise (e.g., Branscombe, Wann, & Noel, 1994; see also Marques & Paez, 1994, for a review). The experiments reported in the present article explore the idea that judgments and evaluations of group members serve both a category differentiation and a normative function. In line with self-categorization theory, category differentiation is oriented towards clear-cut intergroup differentiation. The normative function is closer to a process of social control that is intended to establish the legitimacy of valued in-group standards. Overview of Experiments and Hypotheses We examined category and normative differentiation in four experiments using a minimal groups procedure. Participants completed a task that was ostensibly diagnostic of different judgment styles. They made decisions, one of which involved rank ordering the responsibility of six characters in a murder case. They then wrote down justifications for these decisions. In the second phase of the experiments, some days later, we orthogonally manipulated category membership and normative position of the five group members. Participants were categorized into one of two groups, ostensibly on the basis of which of two justification styles they had used in the first phase. They then received individualized feedback about the responsibility rankings (but not justifications) made by five other individuals from the in-group or the out-group. It was made explicit that all five members clearly belonged to their group on the basis of the way they had justified their decisions in the first phase of the study. The feedback indicated that four of these individuals had made normative decisions (i.e., they rank ordered the characters in line with the norm of their group) and the other individual ranked the characters in a way that deviated toward the norm of the opposing group. The analysis of evaluations of group members was conducted by means of a Group Presented (in-group vs. out-group) X Member (modal vs. deviant) design in which group presented was a between-subjects factor and member was a within-subjects factor. All participants made judgments about the in-group and out-group as a whole. These responses were analyzed according to a group rated (in-group rated vs. out-group rated) within-subjects factor. Because we were interested in the normative component of judgments about modal and deviant group members, in Experiment 2 we manipulated whether participants would be accountable to in-group or out-group members when making these judgments. Accountability (in-group accountability vs. out-group accountability) was a between-subjects factor. In Experiment 3, we varied the salience of norms directly, prior to participants' judging group members. Salience (high vs. low) was a betweensubjects factor. In Experiment 4, we examined the impact of judgments of group members on changes in group identification. Our interest in these experiments was to find evidence for distinct effects of judgments based on category differentiation and normative differentiation. We examined four hypotheses. According to social identity theory, a basic process underlying judgments of in-group and out-group categories is the establishment of positive distinctiveness for the in-group (Hogg, 1992; Hogg & Abrams, 1988, 1993). Therefore, regardless of additional manipulations, participants should evaluate the whole in-group more favorably than the whole out-group (Hypothesis 1). A complementary process is protection of positive social identity from threats from inside the in-group. This normative differentiation process, whereby people discriminate against ingroup members who threaten in-group norms, has been observed in studies on the "black sheep effect" (Marques, 1990). If people derogate in-group deviates because these members threaten in-group norms, they should, concomitantly, upgrade out-group deviant members who adhere to in-group norms, because these members legitimize in-group norms. In line with this reasoning, we predicted deviant in-group members should be derogated relative to modal in-group members, whereas deviant out-group members should be upgraded relative to modal out-group members (Hypothesis 2). This effect should be larger to the extent that attention is focused more closely on in-group norms. When attention is focused only on the intergroup comparative context, as is the case in traditional intergroup studies, normative differentiation should not emerge or, at least, should strongly decrease. Therefore, as a corollary of Hypothesis 2, we expected normative differentiation to be elevated by in-group

4 978 MARQUES, ABRAMS, PAEZ, AND MARTINEZ-TABOADA accountability (Experiment 2) and to depend on the salience of the in-group norm (Experiment 3). Also, in line with our view that normative differentiation maintains and increases participants' identification with their group, we predicted that favoring of in-group-normative over out-group-normative members (regardless of whether these members belong to the in-group or the out-group) will be positively correlated with participants' identification with the ingroup (Hypothesis 3). In Experiment 4, we examined the proposition that favoring in-group-normative over out-group-normative members actually reinforces identification with the in-group (Hypothesis 4). Method Experiment 1 Participants. Twenty female and 6 male high school students (N = 26), ages 15 to 18 years, volunteered to participate in this experiment. Sixteen participants were in the in-group presented condition and 10 were in the out-group presented condition. Design. The design consisted of one between-subjects (group presented) and one within-subjects variable (member). Group presented was manipulated by presenting participants with responses given either by in-group or out-group members. Member was varied in terms of the decision pattern of the members whose responses were presented (modal vs. deviant as compared with their group's response pattern). Procedure. We informed participants that the study was about jury decision making. In the first session, we presented participants with a murder case involving six persons and asked them to rank these persons according to their responsibility for the death of the victim. We also told participants that the goal of the study was to validate the existence of two decision-making patterns (X and Y). They were asked to justify their ranking of characters and were informed that their justification would allow the experimenters to determine their membership to one decision-making pattern. 1 Two days later, the experimenter returned participants' response sheets from Session I, together with a "personal file."' This tile provided the following information about the intergroup categorization (the criteria defining membership to each pattern): From the answers collected in the first phase of this study, we were able to identify two opposed patterns of judgment: Pattern X and Pattern Y. According to the way you justified your ranking of the characters, you undoubtedly belong to Pattern X (vs. Pattern Y). That is, you belong to the group of people who appraise this kind of event such that you tend to assign responsibility more according to the person's objective involvement in the crime than according to other factors that may have caused it indirectly. Those who, contrary to you, belong to Pattern Y (vs. Pattern X) appraise events such that they tend to assign responsibility more according to factors that may have caused the crime indirectly rather than according to the objective components of the situation. At this point, participants were provided with a first set of questions to check for their identification with their group. Pattern information and pattern membership were counterbalanced within conditions. Participants were then provided with information that defined the typical (thus normative) ranking given by in-group or out-group members. Information about the group norm was as follows: Although it is not necessary for belonging to Pattern [X or YJ, in general, people who belong to Pattern [X or Y] should order the characters of the story, from most responsible to least responsible, in the following way [example shown], whereas those who belong to the other pattern [Y or X], should rank the characters in the following way [example shown]. All participants received a printed page showing an in-group norm (pattern) corresponding exactly to their own ranking of responsibility of the six characters and an out-group norm corresponding to the opposite sequence. Finally, participants read the answers given by five group members. These were all either in-group members or out-group members (depending on group presented condition). Four members appeared as modal and normative. Their answers matched their group norm. The other member was deviant; that member's answer was constructed by reversing the order of modal members' responses and, within this order, by reversing the order of the first and second, and of the sixth and fifth characters. Dependent measures. Immediately following the description of judgmental criteria for both patterns, participants were asked four questions to check for social identification; these questions concerned agreement with in-group criteria ("To what extent do you agree with the criteria of judgment followed by people belonging to your pattern?"), agreement with out-group criteria ("To what extent do you agree with the criteria of judgment followed by people belonging to the judgment pattern opposed to yours?''), in-group belongingness (' 'To what extent do you like belonging to your pattern?"), and out-group belongingness ("To what extent would you like to belong to the pattern opposed to yours?"). Participants answered these agreement and belonging questions on 7-point scales ranging, respectively, from 1 (fully disagree, not at all) to 7 (fully agree, very much). Participants also gave their impression about the overall pattern (X or Y) by answering the following question: ' 'What is your global impression of this group?'' Ratings were given on 7-point scales (1 = very unfavorable, 1 ~ very favorable). After receiving the information about group members' rank orderings, participants were instructed to give their impressions of each group member by answering the following question: "We now ask you to evaluate each of the persons whose answers you just read. What is your global impression about each of these persons?" The evaluations of the four modal members were collapsed into a single score (Cronbach's a =.91). Results Social identification. The correlation between agreement with and belongingness to the in-group (r =.70, p <.001) and agreement with and belongingness to the out-group (r =.56, p 1 The case was as follows: "A young wife, forsaken by her husband, whose work often kept him away from home, was seduced by another man. That night, she slept with him at his house across the river. She knew that she had to be back home early in the morning, because her husband would come back from a 2-week business journey, and she wanted to conceal her affair. At dawn, she quietly left her lovers house while he was still asleep. But, when she got to the bridge, a fool sat there, gazing at her with a threatening eye. She was afraid and decided to ask the skipper of the ferryboat to shuttle her to the other side of the river. The skipper agreed and asked her for the fare money. She had no money, however. The skipper explained to her that, in order to feed his wife and children, he would not work for free. In despair, she went back to her lover's house to ask him for help, but he was angry at her for leaving without warning. He simply did not want to listen to her. She then decided to ask an old friend for help. For a long time in the past, this friend had entertained a platonic attraction to her, but she had always refused to marry him. On listening to her story, the friend was very disappointed with her behavior and refused any help. After another vain attempt with the skipper, she desperately hastened toward the bridge. The fool stabbed her with a knife and killed her."

5 CATEGORY AND NORMATIVE DIFFERENTIATION ) allowed us to collapse the scores, respectively, of agreement with and belongingness to in order to create an in-group and an out-group identification score. We performed a Group Presented (information about in-group vs. out-group members) X Group Rated (in-group vs. out-group identification) analysis of variance (ANOVA). Group presented was a between-subjects factor, and group rated was a within-subjects factor. This analysis supported the effectiveness of social categorization in generating in-group identification. As indicated by a significant main effect of group rated, participants identified significantly more with the in-group than with the out-group (Ms = 5.75 and 3.08, respectively), F(l, 24) = 40.32,/? <.001 (remaining Fs < 1). Evaluations of in-group and out-group as a whole. In the in-group presented condition, participants evaluated the ingroup as a whole, whereas in the out-group presented condition, participants evaluated the out-group as a whole. Participants rated the in-group more favorably than the out-group as a whole, consistent with Hypothesis 1 (Ms = 5.44 and 4.10, respectively), F(l, 25) = 4.08, p <.06. Evaluations of modal and deviant members. Hypothesis 2 was supported by a significant Group Presented X Member interaction on evaluations of group members, F(l, 24) = 35.3 \,p <.001 (remaining F& < 1). Simple effects analyses revealed that modal in-group members were judged more favorably than deviant in-group members (Ms = 5.44 and 3.00, respectively), F( 1, 24) = 21.93, p <.001. In turn, modal out-group members were judged more unfavorably than deviant out-group members (Ms = 2.95 and 5.50, respectively), F(l, 24) , p <.01. identification and evaluations of modal and deviant members. To examine Hypothesis 3, we constructed a normative differentiation index. For participants in the in-group presented condition, we subtracted the evaluation of the deviate from the averaged evaluations of the modal members. For participants in the out-group presented condition, we subtracted the average evaluation of modal members from the evaluation of the deviate. Thus, a higher score represents a more positive evaluation of the member(s) closer to the in-group norm. We also computed an identification index by subtracting in-group from out-group identification scores. The significant correlation between the identification and the normative differentiation index (r =.33, p <.01) supports the hypothesis that differential identification with the in-group is associated with favoring group members who endorse in-group-normative positions. Discussion The present results demonstrate that category and normative differentiation may operate simultaneously even in a minimal group context. The antagonistic interaction between group presented and member shows that deviant in-group members and modal out-group members are derogated relative to modal ingroup members and deviant out-group members. The results also uphold the idea that such derogation is based on participants' commitment to in-group norms. A simple category differentiation process might lead participants to prefer those group members who contribute to the clarification of intergroup boundaries. If this were the case, deviates in both the in-group and the out-group would have received negative evaluations relative to modal members. However, our experimental procedure held constant the extent of deviation from in-group and out-group norms. Therefore, it can be inferred that judgments of group deviates reflect the direction {toward or away from in-group norms) and not just the extent of deviation. Experiment 2 In spite of the fact that we made it clear to participants that category membership was not conditional on the responsibility rankings, the results of Experiment 1 might be interpreted as the outcome of perceived interpersonal similarity between the participants' judgments and the judgments given by the target group members. That is, out-group deviates and in-group modal members are necessarily more similar to the participant than are out-group modal members and in-group deviates. If concerns for normative differences influence such judgments over and above similarity, factors that affect normative differentiation should also affect those judgments. In Experiment 2, we checked for the role of these concerns by directly examining the impact of accountability to the in-group or the out-group on judgments of group members. After the categorization and group norm information, we informed participants that their judgments of group members would be shown to either typical in-group or typical out-group members. We assumed that accountability to the in-group would make in-group norms more salient as a basis for differentiation. Therefore, we expected that relative favorability toward the in-group as whole, as well as toward in-group-normative versus out-group-normative members, would be greater in the in-group accountability condition than in the out-group accountability condition. Judgments of the outgroup as a whole, as well as of modal and deviant out-group members, should be less strongly affected by accountability because the out-group and its members are less relevant to normative differentiation within the in-group. Method Participants. Eighteen male and 51 female (N = 69) undergraduate psychology students volunteered to participate in the experiment. Cell sizes ranged from 15 to 23, and gender was similarly distributed across conditions. Design. The design consisted of two between-subjects factors (group presented and accountability) and one within-subjects factor (member). Group presented and member were manipulated as in Experiment 1. We manipulated accountability by informing participants that in a third stage of the study, "your answer booklet will be passed to some highly typical members of the Pattern X [Pattern Y] group who will be asked to give their opinion about your answers." Also, contrary to Experiment 1, in the present experiment participants always evaluated both the in-group and the out-group as a whole. These evaluations were thus treated as a within-subjects factor, group rated (in-group rated vs. out-group rated). Procedure. The procedure was similar to that adopted in Experiment 1 with the following exceptions. The delay between the two sessions was longer (2 weeks), and participants completed two judgment tasks in the pretest session. The first task was to decide on the amount of compensation to be awarded to a victim of an industrial accident and then to provide reasons for that decision. The second was to complete the ranking task used in Experiment 1. In Session 2, we informed participants that the pattern group to which they belonged had been determined by their reasoning on both pretest tasks. The purpose of including an

6 980 MARQUES, ABRAMS, PAEZ, AND MARTINEZ-TABOADA additional task was to further strengthen the basis for category membership in within-group similarity and between-groups dissimilarity by including the compensation judgments as an additional criterion for determining judgmental style. However, as in Experiment 1, participants only received feedback about the responsibility rankings. Dependent measures. The dependent variables were the same as those used in Experiment 1 with two additions. In Experiment 1, group rated was a between-subjects factor. When group presented was the ingroup, participants judged only the in-group as a whole and in-group modal and deviant members. When group presented was the out-group, participants judged only the out-group as a whole and out-group modal and deviant members. In the present experiment, we asked the participants to provide their overall impression (1 = very unfavorable, 7 = very favorable) of both the Pattern X group and the Pattern Y group. This procedure allowed us to determine the extent of intergroup differentiation on a within-subjects basis and whether group presented would have any overall effect on the level of in-group bias. Second, we asked participants to indicate how similar or different they believed each group member was to themselves. This measure of intragroup differentiation was rated on a scale ranging from 1 (very different) to 7 (very similar). We hypothesized that parallel to the effects on evaluations, similarity would be enhanced for modal in-group members and reduced with respect to the in-group deviate and that this effect would be larger in the in-group accountability than the out-group accountability condition. Results Social identification. After being categorized as belonging to a pattern, receiving general information about the two judgmental patterns, and being informed that their judgments would be shown to other group members, participants were asked to indicate the extent of their agreement with the in-group and out-group judgmental criteria and their sense of belongingness to the in-group and the out-group. The correlations between agreement and belonging scores were.62 (p <.001) and.42 (p <.001), respectively, for the in-group and the out-group. As in Experiment 1, we collapsed scores of these items, respectively, to the in-group and the out-group and conducted a Group Presented X Accountability x Group Rated ANOVA. on this index. Group presented and accountability were between-subjects factors. Group rated was a within-subjects factor. The significant effect of group rated showed that participants identified more with the in-group than with the out-group (Ms = 5.25 and 3.09, respectively), F( 1, 67) = 90.33, p <.001 (remaining Fs < 1). Evaluations of entire in-group and out-group. Table 1 shows the means and standard deviations of evaluations of the entire in-group and the entire out-group as a function of account- Table 1 Evaluations of In-Group and Out-Group as a Whole as a Function of Accountability (Experiment 2) Group rated M Accountability SD M SD Total sample M SD ability. The effect of group rated was significant, F( 1, 65) = 67.51, p <.001. In support of Hypothesis 1, participants evaluated the in-group more favorably than the out-group. We also found a significant effect of Accountability X Group Rated, F(\, 65) = 4.02, p <.05. Participants evaluated the in-group more favorably in the in-group accountability than in the outgroup accountability condition, but no differences between these two conditions emerged on evaluations of the out-group, /(67) = 2.43, p <.02, and f(67) = 0.70, respectively. This rinding is in line with the reasoning underlying the predicted increase in normativeness associated with in-group accountability. Evaluations of modal and deviant members. Table 2 shows the means and standard deviations of evaluations of modal and deviant in-group and out-group members in the Group Presented X Accountability cells. A planned comparison on the differences between modal and deviant members 2 in each condition was used to test Hypothesis 2 (see the bottom row of Table 2). 3 This hypothesis states that derogation of deviates relative to modal members would be stronger in the in-group presented/ingroup accountability condition than in all the other conditions. Indeed, this condition presents participants with two factors (ingroup targets and in-group referees) that should reinforce the normative aspect of judgments. In support of Hypothesis 2, the difference between evaluations of modal and deviant members was more positive in the in-group presented/in-group accountability condition than in all the other conditions, (65) = 4.02, 2 The analysis of evaluations was first conducted using just judgments of modal and deviant members. Subsequently, in order to facilitate comparison with results from Experiments 1 and 3, the evaluation of the entire group presented (either in-group or out-group, depending on group presented condition) was included as a further level (target withinsubjects factor). The secondary analysis including evaluations of the entire group revealed almost identical effects. The pattern of means for the entire group closely matched, but was not quite as extreme as, those for modal members. Main effects of group presented, F( 1, 65) = 26.86, p <.001; accountability, F(l, 65) = 7.81, p <.01; and target, F(2, 130) = 12.02, p <.001, were all significant, whereas Group Presented X Accountability, F( 1,65) < 1, and Target X Accountability, F(2, 130) = 2.94, p <.06, failed to achieve significance. As in the previous analysis, there was a significant Group Presented x Target interaction, F(2, 130) = 53.55, p <.001. The significant effects of group presented were larger for judgments of the entire group, F( I, 65) = 49.00, p <.001, and modal members, F( 1,65) = 63.55, p <.001, than for deviant members, F(l, 65) = 23.80, p <.001. There was also a significant Group Presented X Accountability X Target interaction, F(2, 130) = 3.25, p <.05. Once again, the Group Presented X Target interaction effect was larger in the in-group accountability condition, F(2. 134) = 29.20, p <.001, than within the out-group accountability condition, F(2, 134) = 11.87, p <.001. In the in-group accountability condition, there were significant effects of group presented for judgments of the entire group, F(l, 67) = 18.06, p <.001; modal members, F(l, 67) , p <.001; and deviant members, F(l, 67) = 20.80, p <.001. In the in-group accountability condition, the effects remained (less) significant for the entire group, F( 1, 67) = 17.27, p <.001, and modal members, F( 1,67) = l7.84,/?<.001, but the effect was only marginally significant for deviant members, F(l, 67) = 3.66, p = We assigned contrast weighs of +3, -1, -I, and -1. respectively, to the in-group presented/in-group accountability, in-group presented/ out-group accountability, out-group presented/in-group accountability, and out-group presented/out-group accountability conditions.

7 CATEGORY AND NORMATIVE DIFFERENTIATION 981 Table 2 Favorability and Similarity Toward In-Group and Out-Group Members as a Function of Group Presented and Accountability (Experiment 2) accountability accountability presented presented presented presented Group member M SD M SD M SD M SD Modal member Deviant member Normative differentiation p < The results thus show that in-group accountability increases derogation of deviant in-group members and upgrading of modal in-group members, as well as derogation of modal out-group members and upgrading of deviant out-group members (see Figure 1). Correlation between differential identification and normative differentiation. Hypothesis 3 is that normative judgments should positively correlate with in-group identification. The normative differentiation index and identification index, constructed using the same procedure as in Experiment 1, were significantly and positively correlated (r =.36, p <.01). The more participants identified with the in-group as compared with the outgroup, the more they upgraded in-group normative members and derogated out-group normative members. Similarity between self and group members. To examine the relative impact of interpersonal similarity and group membership on judgments of group members, we computed a Group > 5 - ro 4- Presented X Accountability X Member ANOVA. The relevant effect is the significant Group Presented X Accountability X Member interaction, F{ 1, 65) = 6.79, p =.01. Perceived selfmember similarity was stronger in relation to modal in-group members than to modal out-group members, both in the ingroup accountability, F(l, 67) , p <.001, and in the out-group accountability condition, F( 1,67) = 52.57, p <.001. However, the difference between similarity to the deviant ingroup member and similarity to the deviant out-group member was stronger in the in-group accountability condition, F( 1, 67) = 23.59, p <.001, than in the out-group accountability condition, F(l, 67) = 5.03, p <.03. The impact of accountability on similarity judgments clearly suggests that these judgments were affected by normative concerns over and above the effects of actual similarity between participants' responsibility rankings and those of other group members. Discussion Consistent with Experiment 1, the results suggest that participants evaluate deviant out-group members more favorably than prototypical out-group members. In addition, when they are accountable to other in-group members rather than out-group members, participants show stronger in-group bias, consider themselves more similar to modal in-group members, and derogate deviate in-group members more strongly. These results support Hypothesis 2 and suggest that in-group accountability increases the extent to which participants engage in normative differentiation. Previous research suggests that the real or implied presence of an out-group may increase the salience of to Ingroup Outgroup Presented Presented tngroup Accountable Ingroup Presented Outgroup Presented Outgroup Accountable Figure 1. Evaluations of modal and deviant members as a function of group presented and accountability (Experiment 2). 4 A Group Presented X Accountability X Member ANOVA showed a significant second-order interaction, F(l, 65) = 4.06, p <.05. This interaction further supported Hypothesis 2. It revealed a larger Group Presented X Member effect within the in-group accountability condition, F(I, 67) = 33.98, p <.001, than within the out-group accountability condition, F(l, 67) = 13.04, p =.001. In the in-group accountability condition, participants judged modal in-group members more favorably than modal out-group members, F( 1, 67) = 24.97, p <.001, and deviant in-group members less favorably than deviant out-group members, F( 1, 67) = 20.80, p <.001. A similar pattern emerged in the out-group accountability condition. However, whereas strong in-group bias emerged for modal members, F(l, 67) = 17.84, p <.001, the bias was much reduced when participants considered the deviates, F(l, 67) = 3.66, p <.06.

8 982 MARQUES, ABRAMS, PAEZ, AND MARTINEZ-TABOADA category membership and increase category differentiation (e.g., Abrams, 1990; Abrams, Thomas, & Hogg, 1990; Skinner & Stephenson, 1981). However, the effects of in-group presence appear to be more complex. For example, people high in public self-consciousness are typically more concerned with being seen to conform to and comply with in-group norms (Scheier & Carver, 1981; cf. also Cialdini, Kallgren, & Reno, 1991). This effect could be described as a positive orientation to these norms, and the present results indicate that such orientation has a significant impact on appraisals of in-group as well as out-group members, depending on their contribution to these norms. Experiment 3 In Experiment 3, we investigated the effects of norm salience on the derogation of in-group deviates. The procedure was similar to Experiment 2, except that instead of varying accountability, we manipulated norm salience more directly. After the minimal group categorization, participants reported their agreement with the judgmental criteria, their sense of belongingness to each pattern, and the perceived difference between both patterns. Then, they were informed about group members' responsibility rankings. Participants either were not informed about the group norm at all (low salience) or were informed of the group norm prior to receiving information about responses made by individual members (high salience). Notice that the high-salience condition was essentially identical to what was done for all participants in the preceding experiments. The crucial condition here was thus the low-salience condition. In Experiment 2, we found that participants reported stronger interpersonal similarity between themselves and in-group normative targets when they were accountable to the in-group rather than the out-group. However, our research paradigm explicitly rendered modal in-group members and deviant out-group members more similar to participants than deviant in-group and modal out-group members. Our hypotheses assume that similarity does not provide a sufficient account of evaluations of group members, a conclusion that is well supported by other research in the social identity literature (e.g., Abrams & Brown, 1989; Brown & Abrams, 1986; Hogg, 1992). Instead, we propose that judgments of group members are driven by an interaction between category membership and the normative direction of group members' behavior. If this proposition is correct, derogation of deviant in-group members and upgrading of deviant outgroup members should diminish when similarity ceases to be related to in-group norms. Thus, in the low-salience condition, we expect participants to focus on category differentiation and to be less concerned with normative differentiation. Specifically, the critical prediction in this experiment was that we should replicate the general pattern of results from the preceding experiments in the high-salience condition but find only simple ingroup bias in the low-salience condition; normativeness should have reduced impact on evaluations. In contrast, the similarityattraction interpretation would hold that self-target similarity should have similar effects regardless of the salience of the norm. Method Participants. Thirty-seven female and 9 male (A' = 46) undergraduate psychology students volunteered to participate in this experiment. Four participants were discarded from the analysis because of missing responses, leaving 42 participants, ranging from 10 to II within each cell of the design. Gender was similarly distributed across conditions. Design, The design consisted of two between-subjects factors (group presented and norm) and one within-subjects factor (member). Group presented and member were manipulated as in Experiment 1. We manipulated norm by varying whether participants were directly provided (high salience) or not (low salience) with information about existing in-group and out-group norms. Group rated was manipulated as a within-subjects factor with the same purpose as in Experiment 2. Procedure. Session 1 was identical to Experiment 1. However, in Session 2, participants were divided into two conditions according to whether they were provided with norm information. In the high-salience condition, participants learned the norm about the group members' responses, read the members' responses, and then evaluated each member. In the low-salience condition, participants simply read the members' responses and then evaluated each member. Finally, all participants evaluated the whole group. Information about the group norm was constructed in the same way as in Experiments 1 and 2. Dependent measures were identical to those of Experiment 1. Results Social identification. Agreement and belongingness ratings were positively correlated both for the in-group (r =.70, p <.001) and the out-group (r =.53,p<.001). As in the preceding experiments, we collapsed these scores to, respectively, an ingroup and an out-group identification score. We then submitted these scores to a Group Presented (information about in-group vs. about out-group members) X Norm (low vs. high salience) x Group Rated (identification with in-group vs. identification with out-group) ANO\A. Group presented and norm were between-subjects factors, and group rated was a within-subjects factor. The only significant effect was that of group rated. F( 1, 38) = 13.48, p =.001. Participants expressed higher identification with the in-group (M = 4.76) than with the out-group (M ). Evaluations of modal and deviant members. For evaluations of modal and deviant members, a Group Presented x Norm x Member ANO\5\ supported Hypothesis 2. 5 The Group Presented X Member interaction was not significant, F(l, 39) < 1. In s In order to facilitate comparison results with Experiments 1 and 2, the evaluation of the entire group presented (either in-group or outgroup, depending on group presented condition) was included as a third level (target within-subjects factor). This analysis yielded significant effects of group presented, F( 1,38) = 32.52,p <.001; Group Presented X Norm, F(l, 38) = 7.72, p <.01; Group Presented X Target, F(l, 38) = 9.22, p <.001; and Group Presented X Norm X Target, F(l, 38) = 7.65, p =.001. Secondary analyses run separately for each level of the target factor revealed only a significant effect of group presented on evaluations of the entire group, F(l. 38) = 37.82, p <.001. Similarly, modal member evaluations were significantly affected only by group presented, F(l, 38) = 25.45,p <.001. In turn, deviant member evaluations were not affected by group presented, F(l, 38) = 1.72, n.v, or by norm, F{ 1, 38) = 2.74, ns, for both, but the Group Presented X Norm interaction was significant, F(l, 38) = 20.36, p <.001, remaining Fs(l, 38) < 1.

9 CATEGORY AND NORMATIVE DIFFERENTIATION 983 support of Hypothesis 2, we found a significant Group Presented x Norm x Member interaction, F(l, 38) = 9.72, p <.01. Within the high-salience condition, group presented was not significant, but the Group Presented X Member effect was significant, F( 1,39) < l,andf(l,39) = r p =.001, respectively. Modal in-group members were judged more favorably than modal out-group members, and deviant in-group members were judged more unfavorably than deviant out-group members, F(l, 39) = 10.59, p <.01, and f(l, 39) = 4.69, p <.04, respectively (see Table 3 and Figure 2). Within the low-salience condition, we found a significant in-group bias, as shown by the effect of group presented, F( 1, 39) = 33.40, p <.001. Ingroup members were judged more favorably than out-group members regardless of whether they were modal or deviant. The results fully support the main hypothesis of the present experiment. In the low-salience condition, participants judged deviant out-group members less favorably than modal in-group members, even though modal in-group members' and deviant out-group members' responsibility rankings were both similar to the participants' rankings. Participants in this condition judged modal out-group members significantly less favorably than deviant in-group members, even though the responsibility rankings made by modal out-group members and by deviant in-group members were both dissimilar to the participants' rankings. These results rule out an alternative interpretation of the findings from our previous experiments based on interpersonal similarity between participants and target group members. The critical element appears to be closeness to in-group norms. When these norms are not salient, the effects of interpersonal similarity on judgments of group members is outweighed by simple category differentiation. We also examined whether derogation of in-group deviates and upgrading of out-group deviates were stronger in the high-salience condition. The planned comparison on the differences between modal and deviant members in each condition was significant, f(38) = 3.15, p < As expected, the difference between the Table 3 Ratings of Entire Group, Modal Members, and Deviant Members as a Function of Group Presented and Norm Salience (Experiment 3) Group presented M High Entirti group Norm salience SD Modal members M Low SD i 6- >- 5- A cc 4 o > Member: Modal Deviant Ingroup Outgroup Ingroup Outgroup Presented Presented Presented Presented High Salience Low Salience Figure 2. Evaluations of modal and deviant members as a function of group presented and norm salience (Experiment 3). evaluations of modal and deviant members was the highest in the high-salience conditions (Ms = 1.25 and in the in-group and out-group conditions, respectively). In the low-salience conditions, the difference was lower (Ms = 0.80 and 0.43 in the in-group and out-group condition, respectively). Correlational analysis between differential identification and normative differentiation. Contrary to findings in the preceding experiments, the identification index, constructed as the difference between the in-group and the out-group identification scores, was not associated with the normative differentiation index (r =.18, ns). The attenuation in this correlation is attributable to the reduction in normative differentiation in the low-salience condition. Consistent with this interpretation, the correlation between the identification and the normative differentiation index was positive, albeit nonsignificant, in the highsalience condition (r -.22) but negative in the low-salience condition (r = -.50, p <.05). This is due to the fact that in the low-salience condition participants evaluated the deviant ingroup member more favorably than the modal in-group member. These correlations differ significantly (z = 2.44, p <.01, onetailed). Discussion The results of Experiment 3 add further support to our interpretation of Experiment 1. In Experiment 1, norms were available to all participants and affected judgments of in-group and Deviant member We assigned +2, +1, 2, and -1, respectively, to the in-group presented/high-salience, in-group presented/low-salience, out-group presented/high-salience, and out-group presented /low-salience condition.

10 984 MARQUES, ABRAMS, PAEZ, AND MARTINEZ-TABOADA out-group deviates. Specifically, in Experiment 1 and in the high-salience condition of Experiment 3, participants were more positive toward members who adhered to the in-group norm regardless of their membership category. However, in the lowsalience condition of Experiment 3, participants' responses reflect a simple use of category differentiation, although the similarity between participants' responses and target group members' responses did not vary. In this condition, participants were positively biased toward deviant in-group members even though these individuals were less similar to themselves than were deviant out-group members. Thus, interpersonal similarity in the absence of in-group norms was insufficient to produce derogation of deviant in-group members. The results are also consistent with those obtained in Experiment 2. They show that in-group and out-group judgments can be pervaded by concerns about normative differentiation to support in-group norms. The results suggest that situational factors can vary the focus on category membership or in-group norms when people make appraisals of group members (cf. Framing & Carver, 1981). This is consistent with Abrams and Brown's (1989) finding that focus on in-group norms increases positivity toward out-groups that endorse those norms. Abrams and Brown found that participants high in public self-consciousness were more positive toward out-groups when those out-groups appeared to share in-group norms than when they did not. In our view, the most important conclusion to be drawn from the present results is that interpersonal similarity produces shifts in evaluations of individual category members only if similarity corresponds to group norms. This finding appears to be consistent with attraction based on group prototypicality rather than interpersonal similarity per se (Hogg & Hains, 1996). However, attraction to the in-group prototype may not provide a full account of the results. In the high-salience condition, participants rated deviant out-group members almost as favorably as they judged modal in-group members. Judgments driven by attraction to the in-group prototype alone would not have been favorable toward out-group atypical members. Indeed, by definition, outgroup atypical members cannot be prototypical of the in-group. A parsimonious explanation seems to be, in line with our reasoning, that the attractiveness of out-group deviates increases to the extent that they legitimize norms held by the in-group. Experiment 4 Experiments 1 to 3 support the idea that differentiation between in-group and out-group categories can occur independently of evaluations of category members. They also confirm that judgments of members reflect their closeness to in-group norms when those norms are salient, but not otherwise. Our model proposes that normative differentiation between targets who do and do not conform to in-group norms enhances identification with the in-group. We reason that this reflects participants' desire to impose subjective control over group members and to legitimize in-group norms. In Experiment 4, we examined the relationship between normative differentiation and group identification. If our reasoning is correct, it should be the case that those who differentiate more strongly will subsequently feel more strongly attached to the in-group. Moreover, because legitimization does not depend on the group membership of the target, differentiation in favor of in-group-normative targets should increase the in-group identification regardless of the group membership of the target. As in Experiments 1 to 3, we predicted that participants should derogate deviant in-group members more than normative in-group members. We also predicted that in-group-normative out-group members would be judged more favorably than out-group-normative out-group members. However, the present experiment also tested Hypothesis 3, that favoring of in-group-normative over out-group-normative individuals will be positively correlated with participants' in-group identification. In order to test this hypothesis, we asked participants first to indicate their belongingness to the in-group and out-group, then to evaluate modal and deviant ingroup or out-group members, and finally to indicate their group belongingness again. To decrease demand characteristics due to the fact that participants responded twice to the group belongingness measure, we included the second measure in a more general questionnaire purportedly dealing with the "validation of a group membership scale." We expected that the relationship between the first and second measures of in-group identification would be partially mediated by normative differentiation, such that more differentiation in favor of in-group-normative members would result in increased in-group identification. Method Participants. Ninety-eight male and 34 female undergraduate psychology students (N - 132), ages 18 to 22 years, volunteered to participate in this experiment. Gender was similarly distributed across conditions. Design. In Experiment 4, we employed a 2 (in-group presented vs. out-group presented) x 2 (in-group rated vs. out-group rated) X 2 (pre- vs. postidentification) X 2 (modal members vs. deviant members) design. The first factor (group presented) was a between-subjects factor, and the remaining (group rated, pre-post, and member) were withinsubjects factors. We manipulated group presented (information about ingroup members' vs. information about out-group members 1 responses), group rated (identification with the in-group, identification with the outgroup), and member (modal, deviant) in the same way as in Experiments 1 to 3. Pre-post was manipulated by asking participants to report their identification with the in-group and the out-group both before and after evaluating modal and deviant group members. Procedure. The general procedure for Session 1 was identical to Experiments 1 and 3. Three days later, in Session 2, participants were categorized and completed a first measure of identification, both to the in-group (pre-in-group identification) and to the out-group (pre-outgroup identification). They then received information about the modal and deviant members of either the in-group or the out-group. After evaluating these members, participants completed a second measure of identification with the in-group (post-in-group identification) and the out-group (post-out-group identification). The dependent measures were identical to those used in Experiments 1 and 3. Results Social identification. We submitted the pre- and postidentification measures to a Group Presented X Group Rated X Pre- Post ANOVA in which group presented was a between-subjects factor and group rated and preidentification-postidentification were within-subjects factors. There was a significant effect of group rated. Participants reported stronger identification with

11 CATEGORY AND NORMATIVE DIFFERENTIATION 985 the in-group than with the out-group, (Ms = 4.96 and 3.12, respectively), F(\, 130) = , p <.001. There was also a significant main effect for pre- versus postidentification, (Ms = 3.96 and 4.12, respectively), F(l, 130) = 6.88, p =.01 (remaining Fs < 1). Participants identified more strongly after they had evaluated the groups and group members. Evaluations of modal and deviant members. For evaluations of modal and deviant members, a Group Presented x Member ANOVA revealed significant effects of group presented, F(l, 130) = 29.86,/? <.001, and Group Presented X Member, F(\, 130) = , p <.001. The modal in-group member was evaluated more positively than the modal out-group member (Ms = 5.35 and 2.13, respectively), F(l, 130) = , p <.001. In turn, the deviant in-group member was evaluated more negatively than the deviant out-group member (Afs = 2.56 and 4.34, respectively), F(l, 130) = 80.89, p <.001. Correlational and mediation analysis. In the in-group presented condition, we subtracted the evaluation of the deviate from the average evaluation of the modal members. In the outgroup presented condition, we subtracted the modal evaluations from the deviate evaluation. Thus, for all participants, a higher score represents a more positive evaluation of member(s) that are closer to the in-group norm. This corresponds to a normative differentiation index, which stands for differences in evaluations of in-group normative and out-group normative members. The correlations between pre-in-group identification and post-ingroup identification with the normative differentiation index were significant (r =.17, p <.03, and r -.35, p <.001, respectively, one-tailed; see also Table 4). To check for the strengthening effect of evaluations of modal and deviant members on in-group identification, we conducted a regression analysis to test the prediction that evaluative differentiation affects change in identification. In line with the preceding experiments, we subtracted pre-in-group identification from pre-out-group identification (preidentification index), and postin-group identification from post-out-group identification (postidentification index). The larger these indexes, the more participants identified with the in-group as compared with the outgroup. The normative differentiation index and preidentification were taken as predictors, and postidentification was the criterion variable. We expected that, with preidentification accounted for, postidentification would be predicted by the normative differentiation index. The significant effects of preidentification (ft =.46, t = 5.42, p <.001) and of the normative differentiation index (/3 =.27, t = 3.36, p <.001) supported this prediction, R =.53, R 2 =.30, F(2, 124) = 25.12, p <.001. In order to examine the possibility that normative differentiation mediated the change in identification, we followed the procedure outlined by Baron and Kenny (1986). For evidence of mediation, it is necessary that the total effect of preidentification on postidentification be reduced when its indirect effect (through normative differentiation) is accounted for. The data confirm that mediation did occur, as the total effect (/? =.46) is reduced to /? -.42 once the effect of differentiation is accounted for. 7 In sum, the data reveal that prior in-group identification resulted in greater normative differentiation, that normative differentiation resulted in greater in-group identification, and that a small part of the effect of prior identification on subsequent identification was mediated by normative differentiation. Discussion The results of Experiment 4 replicate our basic finding that participants favor in-group modal members over in-group deviates and out-group deviant members over out-group modal members, consistent with the results of Experiments 1 to 3. We also found that relative identification to the in-group is associated with favoring members that endorse in-group normative positions over those endorsing non-normative positions. More important, we found that favoring members who are in-group normative reinforces relative identification to the in-group when prior identification is accounted for. This result is consistent with our hypothesis that identification to the in-group motivates endorsement of norms and that endorsement of these norms bolsters identification. This fits with our view that social categorization fosters a process akin to our conception of subjective group dynamics. Participants based their evaluations of group members on their desire to acquire a legitimized sense of positive social identity by committing themselves to norms and exerting subjective pressures on other group members to conform to those norms. As a result of reinforcing those norms, participants became still more strongly attached to the in-group. Table 4 Correlations Between Normative Differentiation Index (NDl) and Differential In-Group Identification Before (Pre-ID) and After (Post-ID) Judgments of Modal and Deviant Group Members (Experiment 4) Measure Measure/group presented Pre-ID Post-ID NDl Pre-ID *.10 <.10 (marginally significant). * p <.05. ** p < **.24t 31**.61** General Discussion The four experiments reported in this article support our hypotheses consistently. Experiment 1 demonstrates that categorical and normative differentiation can be observed simultaneously even using a minimal group procedure. Participants upgraded the in-group as a whole, as well as in-group and out-group members when these members behaved according to in-group norms. Experiment 2 illustrates how situational features that highlight in-group norms can moderate the effect of categorization on judgments of in-group attributes. Accountability to the in-group elevated expressions of agreement with the in-group, judged dissimilarity between groups, positive evaluations of the in-group, positive evaluations 7 Using EQS, we found the indirect effect of preidentification (P =.48) is marginally significant (t = 1.71, p <.05, one-tailed).

12 986 MARQUES, ABRAMS, PAEZ, AND MARTINEZ-TABOADA of group members who endorsed in-group norms, and negative evaluations of members who opposed these norms. Experiment 3 provides a direct check on the effects of salience of group norms. When participants were unaware of the relationship between group members' behavior and prescriptive group norms, their judgments were simply driven by a tendency to evaluate the ingroup more favorably than the out-group. This applied equally to evaluations of the group as a whole, evaluations of modal members, and evaluations of deviant members. However, once prescriptive in-group norms became explicit, deviant in-group members were strongly derogated, although evaluations of the entire group and modal members remained unchanged. Finally, Experiment 4 reveals the mediating effect of judgments about in-group-normative and out-group-normative members (regardless of these members' category membership) on group identification. These results show that in-group identification fostered normative differentiation and that normative differentiation reinforced identification to the in-group. Category Differentiation and Normative Differentiation in Group Judgments Our findings highlight two interrelated processes. The first process relates to category differentiation and in-group favoritism and is consistent with self-categorization theory. This theory proposes that judgments of group members are determined by a metacontrast, according to which individuals subjectively assess the ratio between average intergroup differences and average intragroup differences (Hogg & Abrams, 1988; Hogg & McGarty, 1990; Turner et al., 1987). The larger the metacontrast ratio, the more perceivers disregard intragroup differences and view members of distinct categories as representatives of contrasting prototypes. In this vein, group prototypes function as group norms, because perceivers expect group members to adhere to them (Oakes, Haslam, & Turner, 1994). As a result, in intergroup contexts, individuals strive to maximize the clarity of the distinction between in-group and out-group features (e.g., Brown & Abrams, 1986; Hogg, 1992). Our results indicate that participants engaged in this process. Following social categorization, participants engaged in general intergroup differentiation and in-group bias across conditions. These responses remained relatively unaffected by the intragroup focus generated by the salience of in-group or out-group accountability and salience of in-group norms. It would seem that intergroup differentiation and in-group favoritism emerged as a simple and direct response to judgmental settings that induced in-group/out-group comparisons. Social identity and self-categorization theory also suggest that individuals prefer those in-group or out-group members who maintain or increase intergroup differences (e.g., Hogg, 1996; Hogg & Abrams, 1993; McGarty et al., 1993) and that in-group members should typically be regarded as more attractive than out-group members (e.g., Tajfel, 1978; Tajfel & Turner, 1979). However, this suggestion may be problematic under circumstances in which the deviation arises on generally valued dimensions such as likability. Research by Marques and colleagues (see Marques & Paez, 1994, for a review) reveals that people are more favorable toward likable (but atypical) out-group members than toward unlikable (but atypical) in-group members, although both reduce intergroup contrasts. These findings imply that, within a general framework of intergroup differentiation, people may try to establish normative distinctions among individuals. The present results directly support this idea. The information we provided to participants emphasized that all group members did conform to a group membership criterion (Pattern X or Pattern Y reasoning). If participants were motivated only to maintain clear-cut intergroup boundaries, their judgments should have been more favorable towards prototypical or modal and less favorable towards atypical or deviant members, regardless of whether they were in-group or out-group members. The fact that this phenomenon was not observed is consistent with the operation of a second process, that of normative differentiation among group members. Group members that deviated from in-group norms were derogated more strongly when accountability or norm salience heightened participants' awareness of in-group norms. Under these conditions, judgments of individual in-group and outgroup members reflected the extent of their deviation from ingroup norms. However, variations in these within-group judgments did not affect the levels of intergroup differentiation. These findings are consistent with the idea that an important concern for people in intergroup situations is to establish the legitimacy of in-group norms. Our results are not necessarily in contradiction with selfcategorization theory. In our experiments, intergroup boundaries were clear-cut from the outset. Hence, it is possible that participants did not see in-group-normative out-group members as a strong threat to intergroup distinctiveness. It may have been subjectively unproblematic for them to accept counternormative out-group members as a source of legitimacy for their own group norms. This raises the interesting question of haw downgrading of out-group-normative members and upgrading of ingroup-normative members might vary as a function of the clarity of intergroup boundaries. Self-categorization theory might predict that normative intragroup differentiation emerges primarily when category differentiation is not at risk. An interesting avenue for future research would be to identify the conditions under which the tension between validation of in-group norms and threats to in-group distinctiveness is resolved in different directions (cf. Brewer, 1993; Triandis, Bontempo, Villareal, Asai, & Lucca, 1988). Social Identity and the Operation of Subjective Group Dynamics A further possibility, in line with more traditional research on group cohesiveness (see Cartwright & Zander, 1968; Levine, 1989, for a review), is that normative differentiation supports the perceived legitimacy of intergroup differences and, consequently, emerges primarily when social identity is insecure (see also Tajfel, 1978). For instance, Moreland, Levine, and Cini (1993) proposed that one core process in judgments of group members is the evaluation of the members' acceptance of and motivation to fulfill group normative expectations. The group takes corrective action to reduce discrepancies for instance, by redefining the status of individual members. This action requires a committed, ego-involved attempt to harmonize between the actual characteristics or behavior of group members and

13 CATEGORY AND NORMATIVE DIFFERENTIATION 987 perceived group goals (Levine & Moreland, 1994). Group members who threaten group cohesiveness are negatively evaluated and ultimately may be rejected from the group. Assuming that the motivation underlying this process applies to larger social categories, our findings seem consistent with this latter view. Together, our findings are consistent with the operation of "subjective group dynamics" in judgments of group members (Marques & Paez, 1994; Marques, Paez, & Abrams, 1998). When social category memberships are salient, individuals may assign themselves characteristics they recognize as most typical (i.e., shared by most members) of the group (Hogg, 1992, 1996; Hogg & Abrams, 1988; Turner et al., 1987). In the absence of other cues, "social reality" (Festinger, 1950) is primarily defined by group membership, and a meaningful goal individuals can strive for is clarification and confirmation of their selfdefinition (cf. Abrams, 1994; Hogg & Abrams, 1993; Marques & Paez, 1994). In support of the above idea, previous research has demonstrated that openness to social influence is affected by common self-categorization between the source and target (Abrams & Hogg, 1990; Hogg & Turner, 1987; Turner et al., 1987). For example, Abrams, Wetherell, Cochrane, Hogg, and Turner (1990, Experiments 1-3) found that group membership is an important moderator of conformity in the Sherif autokinetic effect paradigm, the Asch conformity paradigm, and the group polarization paradigm. In the present experiments, we showed that evaluations of group members are strongly affected by the extent to which they support subjectively valid in-group norms. These evaluations may reflect privately (Experiments 1, 2, and 4) or publicly implicit (Experiment 2) social pressures on others to reinforce those norms. Perceived "interdependence" or "common fate" with other in-group members is absolute, because the value and coherence assigned to the cognitive representation of the in-group fully affects the clarity of the definition of self and its value as a group member (Marques, Paez, & Abrams, 1998). In particular, threats to social identity arising within in-groups lead participants to attempt to restore the in-group's overall relative positivity. With this intragroup focus, people may actively seek to constrain other group members to reinforce in-group norms for instance, by derogating in-group members who represent a threat to in-group cohesiveness. Similarly, validation for in-group norms, even from outgroup members, helps to verity the social reality implied by those norms. References Abrams, D. (1990). How do group members regulate their behaviour? An integration of social identity and self-awareness theories. In D. Abrams & M. A. Hogg (Eds.), Social identity theory: Constructive and critical advances (pp ). London: Harvester Wheatsheaf; and New Itbrk: Springer Verlag. Abrams, D. (1994). Social self-regulation. Personality and Social Psychology Bulletin, 20, Abrams, D., & Brown, R. (1989). Self-consciousness and social identity: Self-regulation as a group member. Social Psychology Quarterly, 52, Abrams, D., & Hogg, M. A. (1990). Social categorisation, social identity and social influence. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 1, pp ). New >fork: Wiley. Abrams, D., Thomas, J., & Hogg, M. A. (1990). Numerical distinctiveness, social identity, and gender salience. British Journal of Social Psychology, 29, Abrams, D., Wetherell, M. S., Cochrane, S., Hogg, M. A., & Turner, J. C. (1990). Knowing what to think by knowing who you are: Selfcategorisation and the nature of norm formation, conformity and group polarisation. British Journal of Social Psychology, 29, Baron, R. M., & Kenny, D. A. (1986). The moderator mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, Billig, M. (1985). Prejudice, categorization and particularization: From a perceptual to a rhetorical approach. European Journal of Social Psychology, 15, Branscombe, N. R., Wann, D. L., & Noel, J. G. (1994). or out-group extremity: Importance of the threatened social identity. Personality and Social Psychology Bulletin, 19, Brewer, M. B. (1993). The role of distinctiveness in social identity and group behaviour. In M. A. Hogg & D. Abrams (Eds.), Group motivation: Social psychological perspectives (pp. 1 16). London: Harvester Wheatsheaf. Brown, R., & Abrams, D. (1986). The effects of intergroup similarity and cooperative vs. competitive orientation on intergroup discrimination. Journal of Experimental Social Psychology, 22, Brown, R., Hinkle, S., Ely, P. G., Fox-Cardamone, L., Maras, P., & Taylor, L. A. (1992). Recognizing group diversity: Individual! st-collectivist and autonomous-relational social orientations and their implications for intergroup processes. British Journal of Social Psychology, Cartwright, D., & Zander, A. (1968). Group dynamics. London: Tavistock. Cialdini, R. B., Kallgren, C. A., & Reno, R. R. (1991). A focus theory of normative conduct: A theoretical refinement and revaluation of the role of norms in human behavior. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 24, pp ). New\brk: Academic Press. Crocker, J., Blainc, B., & Luhtanen, R. (1993). Prejudice, intergroup behaviour and self-esteem: Enhancement and protection motives. In M. A. Hogg & D. Abrams (Eds.), Group motivation: Social psychological perspectives (pp ). London: Harvester Wheatsheaf. Crocker, J., & Luhtanen, R. (1990). Collective self-esteem and ingroup bias. Journal of Personality and Social Psychology, 58, Doise, W., & Sinclair, A. (1973). The categorisation process in intergroup relations. European Journal of Social Psychology, 3, Festinger, L. (1950). Informal social communication. Psychological Review, 57, Framing, W. J., & Carver, C. S. (1981). Divergent influences of private and public self-consciousness in a compliance paradigm. Journal of Research in Personality, 15, , Hamilton, V. L., & Rauma, D. (1995). Social psychology of deviance and law. In K. S. Cook, G. A. Fine, & J. S. House (Eds.), Sociological perspectives on social psychology (pp ). Boston: Allyn & Bacon. Hogg, M. A. (1992). The social psychology of group cohesiveness: From attraction to social identity. New York: New York University Press. Hogg, M. A. (1996). Social identity, self-categorization, and the small group. In E. H. Witte & J. H. Davis (Eds.), Understanding group behavior: Vol. 2. Small group processes and interpersonal relations (pp ). Mahwah, NJ: Erlbaum. Hogg, M. A., & Abrams, D. (1988). Social identifications: A social psychology of intergroup relations. New York: Routledge, Chapman & Hall.

14 988 MARQUES, ABRAMS, PAEZ, AND MARTINEZ-TABOADA Hogg, M. A., & Abrams, D. (1993). Towards a single process uncertainty-reduction model of social motivation in groups. In M. A. Hogg & D. Abranis (Eds.), Group motivation: Social psychological perspectives (pp ). London: Harvester Wheatsheaf. Hogg, M. A., & Hains, C. S. (1996). Intergroup relations and group solidarity: Effects of group identification and social beliefs on depersonalized attraction. Journal of Personality and Social Psychology, 70, Hogg, M. A., & McGarty, C. (1990). Self-categorisation and social identity. In D. Abrams & M. A. Hogg (Eds.), Social identity theory; Constructive and critical advances (pp ). London: Harvester Wheatsheaf; and New "fork: Springer-Verlag. Hogg, M. A., & Turner, J. C. (1987). Social identity and conformity: A theory of referent information influence. In W. Doise & S. Moscovici (Eds.), Current issues in European social psychology (Vol. 2, pp ). Cambridge, England: Cambridge University Press; and Paris: Maison des Sciences de 1'Homme. Levine, J. M. (1989). Reactions to opinion deviance in small groups. In P. B. Paulus (Ed.), Psychology of group influence (pp ). Hillsdale, NJ: Erlbaum. Levine, J. M., & Moreland, R. L. (1994). Group socialization: Theory and research. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 5, pp ). New "fork: Wiley. Marques. J. M. (1990). The black sheep effect: homogeneity in social comparison settings. In D. Abrams & M. A. Hogg (Eds.), Social identity theory: Constructive and critical advances (pp ). London: Harvester Wheatsheaf; and New \brk: Springer-Verlag. Marques, J. M., & Paez, D. (1994). The black sheep effect: Social categorization, rejection of ingroup deviates, and perception of group variability. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 5, pp ). New "\brk: Wiley. Marques, J. M., Paez, D., & Abrams, D. (1998). Social identity and intragroup differentiation as subjective social control. In S. Worchel, J. F. Morales, D. Paez, & J.-C. Deschamps (Eds.), Social identity: International perspectives (pp ). New "York: Sage. Marques, J. M., Yzerbyt, V. Y., & Leyens, J.-P. (1988). The black sheep effect: Judgmental extremity towards ingroup members as a function of group identification. European Journal of Social Psychology, 18, McGarty, C, Turner, J. C, Oakes, P. J., & Haslam, S. A. (1993). The creation of uncertainty in the influence process: The roles of stimulus information and disagreement with similar others. European Journal of Social Psychology, 23, Moreland, R. L., Levine, J. M., & Cini, M. A. (1993). Group socialization: The role of commitment. Tn M. A. Hogg & D. Abrams (Eds.), Group motivation: Social psychological perspectives (pp ). London: Harvester Wheatsheaf. Mullen, B., Brown, R., & Smith, C. (1992). Ingroup bias as a function of salience, relevance, and status: An integration. European Journal of Social Psychology, 22, Oakes, P. J., Haslam, S. A., & Turner, J. C. (1994). Stereotyping and social reality. Oxford, England: Blackwell. Scheier, M. F., & Carver, C. S. (1981). Private and public aspects of self. In L. Wheeler (Ed.), Review of personality and social psychology (Vol. 2, pp ). Beverly Hills, CA: Sage. Skinner, M. E, & Stephenson, G. M. (1981). The effects of intergroup comparison on the polarization of opinions. Current Psychological Research, 1, Tajfel, H. (1978). Differentiation between social groups: Studies in the social psychology of intergroup relations. London: Academic Press. Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. Austin & S. Worchel (Eds.), The social psychology of intergroup relations (pp ). Monterey, CA: Brooks/Cole. Triandis, H. C, Bontempo, R., Villareal, M. J., Asai, M., & Lucca, N. (1988). Individualism and collectivism: Cross-cultural perspectives on self-ingroup relationships. Journal of Personality and Social Psychology, 54, Turner, J. C, Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). Rediscovering the social group: A self-categorisation theory. London: Blackwell. Yamagishi, T. (1995). Social dilemmas. In K. S. Cook, G. A. Fine, & J. S. House (Eds.), Sociological perspectives on social psychology (pp ). Boston: Allyn & Bacon. Received March 17, 1998 Revision received May 27, 1998 Accepted May 27, 1998