Abstract
Two studies examined participants’ evaluations of ingroup or outgroup normative and deviant members and changes in agreement with a prescriptive norm. In Experiment 1 (N = 51), the normative target was either a full or marginal ingroup or outgroup member, and the deviant was a full member. In Experiment 2 (N = 113), both targets were full or marginal members, or one was a full member and the other was marginal. As predicted, maximal upgrading of normative members and downgrading of deviant members, as well as endorsement of the norm, occurred when both targets were full ingroup members. In contrast, the deviant was derogated least and the deviant’s position was endorsed most when the deviant target was a full ingroup member and the normative target was a marginal ingroup member. Evaluations of normative and deviant ingroup members mediated the effects of their role on participants’ agreement with the norm.
All groups have norms that prescribe how members should behave, and people who violate these norms (deviants) can be problematical for several reasons. For example, deviants can undermine other members’ confidence in their perceptions of reality, they can inhibit the group’s movement toward collective goals, and they can reduce other members’ perception of the group’s overall value and thereby lower their self-esteem (e.g., Abrams, Randsley de Moura, Hutchison, & Viki, 2005; Festinger, 1950). It is not surprising, therefore, that social psychologists have long been interested in how groups respond to deviants (for reviews, see Jetten & Hornsey, 2014; Levine & Kerr, 2007; Marques & Paez, 1994). Of the various frameworks used to explain reaction to deviance, subjective group dynamics theory (SGDT) has proven to be useful for explaining the functions fulfilled by group members’ negative reactions to deviants and the intragroup and intergroup contexts that influence such reactions (e.g., Abrams, Marques, Bown, & Henson, 2000; Marques, Abrams, Paez, & Martinez-Taboada, 1998; Marques, Abrams, & Serôdio, 2001; Pinto, Marques, Levine, & Abrams, 2010; Pinto, Marques, & Paez, 2015).
The basic premise of SGDT is that ingroup members who oppose generic, prescriptive norms (deviant members) threaten others’ positive social identity, whereas members who uphold these norms (normative members) bolster this identity (Marques, Yzerbyt, & Leyens, 1988). Examples of such norms include speaking respectfully to a religious leader, regardless of whether one is a member of that person’s faith, or saluting an athlete who performed well, regardless of whether one is a member of that person’s team (cf. Marques et al., 2001). Evidence shows that when both such a norm and the actor’s ingroup membership are salient, people evaluate normative ingroup members more favorably and deviant ingroup members more unfavorably than normative and deviant outgroup members holding the same positions. This phenomenon, called the black sheep effect (BSE; Marques et al., 1988), occurs because people seeking a positive social identity are more motivated to differentiate between ingroup members who support versus refute a prescriptive norm than between outgroup members taking the same positions (Marques & Paez, 1994). The BSE is well established across a variety of intergroup and intragroup contexts (e.g., Abrams, Palmer, Rutland, Cameron, & van de Vyver, 2013; Bègue, 2001; Castano, Paladino, Coull, & Yzerbyt, 2002; Coull, Yzerbyt, Castano, Paladino, & Leemans, 2001; Doosje, 2003; Frings, Abrams, Randsley de Moura, & Marques, 2010; Shin, 1999).
Recently, Pinto et al. (2010) used Levine and Moreland’s group socialization model (Levine & Moreland, 1994; Levine, Moreland, & Hausmann, 2005; Moreland & Levine, 1982) to derive a novel set of hypotheses regarding conditions under which the BSE would and would not occur. According to the group socialization model, the roles that members occupy in a group are a function of the length and quality of their relationship to it. The most central role, full member, is occupied by people who have completed the socialization phase of group membership and hence are responsible for carrying out the group’s most important tasks. By virtue of their tenure in the group and contributions to it, full members have more power and status than do members occupying other roles. One such role is that of marginal member, which is occupied by people who once had full member status but have since lost it, for example because they no longer feel strong commitment to the group. According to the group socialization model, full members are perceived as adhering to and reinforcing the group’s normative position more than do other members. Therefore, they are especially capable of confirming or disconfirming this norm (Levine & Moreland, 1994).
This analysis suggests that whether normative and deviant ingroup members occupy full or marginal roles should be critical to their impact on social identity, which in turn should affect the evaluations they receive. In support of this idea, Pinto et al. (2010, Experiments 1–3) found the BSE when normative and deviant members were full members of the group, but not when they were either new or marginal members (for other work on reactions to highly representative group members, see Abrams, Randsley de Moura, Marques, & Hutchison, 2008; Jetten, Hornsey, Spears, Haslam, & Cowell, 2010; Levine & Moreland, 2002; McGarty, Turner, Oakes, & Haslam, 1993).
A potentially important constraint on the generalizability of Pinto et al.’s (2010) findings, as well as of BSE research in general, is the fact that they only studied cases in which both deviant and normative members occupied the same role in the group. For this reason, this research does not answer the question of whether the BSE depends on both the normative and deviant targets or only one of these targets having full member status. Thus, a major goal of the present studies was to compare the impact of the copresence of normative and deviant members playing the same roles in the group with those playing different roles.
The present studies extended Pinto et al.’s (2010) work in another important way, namely by considering participants’ opinion change toward or away from the group’s modal position as well as their evaluation of the normative and deviant targets. The possibility of deviant-induced opinion change was identified many years ago in Festinger’s (1950) classic analysis of reaction to opinion deviance and received attention in studies stimulated by this perspective (e.g., Levine, Saxe, & Harris, 1976; Levine, Sroka, & Snyder, 1977) as well as those stimulated by Moscovici’s (1976, 1980) analysis of minority influence (for a review, see Levine & Tindale, 2014). However, deviant-induced opinion change has not been assessed in prior research on the BSE.
Effects of Balance Between Normative and Deviant Ingroup Members’ Roles
Prior BSE research does not account for the fact that intragroup conflict can occur in contexts in which normative and deviant members occupy different roles as well as the same role in the group. By simultaneously comparing the impact of the copresence of normative and deviant members playing the same roles with those playing different roles, we aim to extend prior work on the BSE in an important new direction. More specifically, we aim to investigate how the balance of the normative and deviant ingroup members’ roles affects both participants’ evaluations of the normative and deviant targets and their opinion change toward or away from the group’s modal position.
We propose that ingroup members’ undermining or reinforcing potential should depend on the balance between their role and the roles of other members who either agree or disagree with the group norm. That is, evaluations of deviant and normative members and agreement/disagreement with deviant and normative opinions should be jointly determined by the target members’ group affiliation (i.e., whether they belong to the ingroup or outgroup) and by the balance between the roles that the deviant and normative members occupy (e.g., full vs. marginal member).
To clarify these ideas, consider the four role balance variations that could occur when perceivers observe two ingroup members, one who disagrees with the group’s position (deviant) and the other who agrees with this position (normative). In one variation, the deviant and normative are both full members of the group. Here, because challenge to the ingroup’s normative position is espoused by a full member, it poses a real threat to that position. However, the fact that another full member espouses the normative position provides assurance about its validity. Therefore, the challenge to the normative position by the deviant full member should increase participants’ motivation to defend that position. As a result, participants should derogate the deviant, upgrade the normative member, and shift their own opinion further toward the normative position (i.e., reinforce their allegiance to the prescriptive norm).
In the second variation, the deviant is a full member while the normative is a marginal member. Here, the available confirmation of the normative position does not provide convincing evidence for its validity, and the deviant member should undermine certainty about the normative position and increase the perceived validity of the deviant position (cf. Randsley de Moura, Abrams, Marques, & Hutchison, 2010). As a result, participants should upgrade the deviant ingroup member, derogate the normative ingroup member, and shift their opinion toward the deviant position.
In the third and fourth variations, the deviant is a marginal member while the normative member is either a full or a marginal member. In both cases, and particularly in the former, deviance should pose little threat to the ingroup’s normative position. As a result, participants should have relatively mild evaluative reactions to both ingroup members and show little opinion shift toward or away from the ingroup’s position.
In summary, we propose that participants’ evaluations of deviant and normative ingroup members and their movement toward or away from the ingroup’s normative position will depend on the balance between the membership roles (full or marginal) that these members occupy. Moreover, we propose that participants’ evaluations will influence the amount and direction of their opinion shift, such that stronger derogation of the deviant relative to the normative ingroup member will produce stronger agreement with the norm. Therefore, we expect that differential evaluations of deviant and normative ingroup targets will mediate the effect of role (full vs. marginal) and position (normative vs. deviant) on opinion shift. By contrast, outgroup members’ positions are less relevant for participants’ social identity. As a result, outgroup members’ statuses will have little effect either on how participants evaluate them or on participants’ level of agreement with the prescriptive norm.
To address these questions, we conducted two experiments. The first manipulated the role of the normative member (full vs. marginal) while holding constant the role of the deviant member (full). The second experiment manipulated the roles of both the normative member (full vs. marginal) and deviant member (full vs. marginal). In both experiments, we also manipulated the group memberships of the two targets (ingroup vs. outgroup).
Experiment 1
Our goal in Experiment 1 was to examine the effect of varying the role of the normative member (full vs. marginal) when the deviant was a full member. In addition, we manipulated the group memberships of the two targets (ingroup or outgroup) to provide a test of the BSE in these two situations. Participants indicated how much they agreed or disagreed with the normative and the deviant opinions before they received information about targets’ roles and positions, then evaluated the targets, and finally gave their own opinions again. 1
Consistent with Pinto et al. (2010), we expected evaluations of the normative target to be most favorable and evaluations of the deviant target to be least favorable when both the deviant and normative targets were ingroup full members than in all other (ingroup and outgroup) conditions. We also expected most shift toward the ingroup’s normative position in this condition and most shift away from this position when the deviant target was an ingroup full member and the normative target was an ingroup marginal member than in the remaining conditions. Finally, we predicted that differential evaluations of normative and deviant targets would mediate the joint impact of membership role (full or marginal) and group membership (ingroup or outgroup) on opinion shift toward or away from the normative position.
Method
Participants and design
Participants were 51 (43 female and eight male) university students attending two different Portuguese universities. Sex, university membership, and age were similarly distributed across experimental conditions. 2
We used a 2 (targets’ group: Ingroup vs. Outgroup) x 2 (normative target’s role: Full Member vs. Marginal Member) x 2 (target’s opinion: Normative vs. Deviant) mixed design, in which targets’ group and normative target’s role were between-participants factors, and target’s opinion was a within-participants factor. 3
Procedure
In the first of two sessions, we informed participants that they were taking part in an interuniversity program designed to help students “reach a consensus concerning their views about recent changes in the university systems in the European Union (the ‘Bologna Process’) and the various consequences of these changes.” We asked participants to indicate how much they agreed or disagreed with two statements representing, respectively, a normative and a deviant opinion about the Bologna Process. In the second session (1 week later), we informed participants that student representatives of their university and of a neighboring university would soon meet to discuss students’ goals and strategies regarding the Bologna Process. Allegedly, participants’ task was to help choose, from a pool of students of their own (or the other) university (“for the sake of objectivity”), those who would represent them in the forthcoming debates. Following the procedure employed by Pinto et al. (2010), participants were given information about two (fictitious) students “chosen at random from the pool obtained in the first session.” This information concerned their university affiliation, the length of their enrollment, how much they felt integrated into and enjoyed being students of that university, and the opinion (normative or deviant) they agreed with the most. This procedure allowed us to manipulate three independent variables.
Target’s opinion manipulation
Information indicated that one target advocated the normative opinion (normative target), whereas the other advocated the deviant opinion (deviant target). The normative opinion was “University students should coalesce in order to be able to negotiate their best options regarding the Bologna Process.” The deviant opinion was “University students are too immature to participate in decision-making, and should abide by the authorities’ decisions regarding the Bologna Process.” These opinions corresponded, respectively, to a generally accepted and a generally rejected opinion among students at the time we conducted the study (cf. Pinto et al., 2010).
Targets’ group and normative target’s role manipulations
The two target students were presented as belonging either to the participant’s university (ingroup) or the other university (outgroup). Additional information indicated that the target students either had been enrolled in the university for 3 years, felt integrated into it, identified with it, and wished to remain in it (full member) or had been enrolled in the university for 3 years, but did not feel integrated into it, did not identify with it, and wanted to leave it (marginal member). The deviant target was always presented as a full member, and the normative target was presented either as a full member or a marginal member.
Measures and manipulation checks
Ingroup identification
In the first session, participants answered three questions designed to assess their identification with their university: “How much do you feel you are a student of your university?”; “How much do you consider yourself to be similar to the other students of your university?”; “How much do you identify yourself with the other students of your university?” (1 = not at all; 7 = very much so). We averaged participants’ responses to these questions to create an ingroup identification score (Cronbach’s α = .83). This measure was included to determine if ingroup identification was equivalent across experimental conditions and to assess the overall level of identification.
Agreement with the normative and deviant opinions
In the first session, before any experimental manipulations, participants indicated how much they agreed with eight statements ostensibly designed “to measure the opinions of the student population about the Bologna Process.” Six statements were fillers aimed to increase the credibility of the procedure. The remaining two statements corresponded to the normative and the deviant opinions (1 = I totally disagree; 7 = I totally agree).
Evaluations of targets
In the second session, participants evaluated each target on seven bipolar scales. Endpoints were selfish, bad friend, ill-mannered, senseless, dull, envious, and disloyal (= 1), and altruistic, good friend, considerate, sensible, interesting, generous, and loyal (= 7). For each participant, we averaged the evaluations of each target on these scales to create a normative target score and a deviant target score (Cronbach’s α = .95 and .91, respectively). For several analyses, we employed an evaluative differentiation score, computed as the difference between the evaluations of the normative and deviant targets.
Opinion change
After evaluating the targets, participants were again asked to state their agreement with the normative and deviant opinions (1 = I totally disagree; 7 = I totally agree) and filler statements. We reversed the response scale for the deviant opinion and computed a normative opinion change and a deviant opinion change score for each participant. Positive normative (deviant) opinion change scores reflected increased agreement with the normative (deviant) position from the first to the second session. We then computed an opinion change score by subtracting the deviant opinion change score from the normative opinion change score, such that positive values indicate increased agreement with the normative opinion (and disagreement with the deviant opinion) and negative values indicate increased agreement with the deviant opinion (and disagreement with the normative opinion).
Manipulation checks
We assessed the accuracy of participants’ recall of the targets’ membership group (“Which university do student A and student B attend?”) and role (“Does student A[B] feel well integrated in the university?”; “Is student A[B] motivated to participate in the discussion program?”). All participants whose responses were considered (cf. Participants section) perceived the target’s membership group and targets’s role as intended.
Results and Discussion
Ingroup identification
A Targets’ Group x Normative Target’s Role ANOVA computed on the ingroup identification scores showed no significant differences in ingroup identification across experimental conditions, all Fs < 1. Furthermore, a one-sample t test comparing the mean ingroup identification score of all participants to the midpoint of the response scale (4) indicated that participants significantly identified with their ingroup university (M = 5.08, SD = 1.08), t(50) = 7.10, p < .001, 95% CI [0.77, 1.38].
Prior agreement with the normative and deviant opinions
We conducted a Targets’ Group x Normative Target’s Role x Opinion (agreement with normative opinion, agreement with deviant opinion) repeated-measures ANOVA on participants’ agreement with the normative and deviant opinions in the first session. We found only a significant effect of opinion, F(1, 47) = 276.68, p < .001, ηp2 = .855. Participants agreed more with the normative (M = 6.24, SD = 0.84) than with the deviant (M = 2.29, SD = 1.59) opinion. Thus, prior to the manipulation of the normative target’s role, participants did not differ across conditions in their agreement with the normative and deviant statements.
Evaluations of targets
We conducted a Targets’ Group x Normative Target’s Role x Target’s Opinion repeated-measures ANOVA on target evaluation scores. Directly relevant to our predictions, we found a significant Targets’ Group x Normative Target’s Role x Target’s Opinion interaction, F(1, 47) = 15.45, p < .001, ηp2 = .247. 4
We decomposed the three-way interaction according to normative target’s role (see Table 1). In line with our hypothesis, we found a significant Targets’ Group x Target’s Opinion interaction when the normative target was a full member, F(1, 48) = 14.95, p < .001, ηp2 = .238, but not when the normative target was a marginal member, F(1, 48) < 1. As expected, in the former condition, participants evaluated the normative ingroup target more favorably than the normative outgroup target, F(1, 48) = 6.06, p = .018, ηp2 = .112, and the deviant ingroup target more unfavorably than the deviant outgroup target, F(1, 48) = 20.58, p < .001, ηp2 = .300.
Evaluations of normative and deviant targets as a function of the normative target’s role and targets’ group (Experiment 1).
Note. 7 = favorable evaluation; 1 = unfavorable evaluation. Standard deviations are presented in brackets.
p ⩽ .10. *p = .001. **p < .001.
Additionally, we conducted separate contrast analyses on the normative and deviant target scores. For the normative target score, we assigned the values of +3 to the condition in which the normative target was an ingroup full member and −1 to the other conditions. For the deviant target score, we assigned the values of −3 to the condition in which the normative target was an ingroup full member and +1 to each of the remaining conditions. Both contrasts were significant, t(47) = 7.91, p < .001, and t(47) = 7.61, p < .001, respectively. As predicted, participants judged the normative target more favorably and deviant target less favorably when those targets were ingroup full members compared to targets in all other conditions.
Opinion change
We expected participants’ opinion shift toward the normative position to be greatest when the normative target was an ingroup full member and smallest (or most negative) when the normative target was an ingroup marginal member compared to the remaining conditions. A Targets’ Group x Normative Target’s Role ANOVA on the opinion change scores yielded a significant Targets’ Group x Normative Target’s Role interaction, F(1, 47) = 5.46, p = .024, ηp2 = .104 (see Table 2). 5
Agreement with the normative opinion and agreement with the deviant opinion in the first and second sessions across conditions (Experiment 1).
Note. 7 = I fully agree; 1 = I fully disagree. Standard deviations are presented in brackets.
To test our hypothesis, we compared the opinion change score in each condition with the value of 0 (no change). We found that (a) increased agreement with the normative position was significant only when the normative target was an ingroup full member (M = 1.33, SD = 1.07), t(11) = 4.93, p < .001, 95% CI [0.65, 2.02] and (b) decreased agreement with the normative position was significant only when the normative target was an ingroup marginal member (M = −2.57, SD = 3.84), t(13) = −2.51, p = .026, 95% CI [–4.79, –0.36]. Within the outgroup conditions, we found no significant differences between the opinion change scores and 0 (t always < 1.71, ns; see Figure 1). These results support our opinion change hypothesis.
Opinion change as a function of targets’ group and normative target’s role (Experiment 1).
Mediation model
We predicted that differential evaluations of normative and deviant targets would mediate the joint impact of membership role (full or marginal) and group membership (ingroup or outgroup) on opinion shift toward or away from the normative position. We first created an evaluative differentiation score by subtracting evaluation of the deviant member from evaluation of the normative member (Abrams, Marques, Bown, & Dougill, 2002; Abrams et al., 2000; Marques et al., 1998). Evaluative differentiation was significantly correlated with opinion change, r = .489, p < .001, indicating the potential for mediation involving these variables.
We then created a new independent variable, labeled “normative conflict” to test our focal prediction. The code values we assigned to this variable reflect our predictions about the expected evaluative differentiation between the normative and deviant targets, and the associated opinion change in each experimental condition. Specifically, we predicted the strongest evaluative differentiation and the strongest opinion shift towards the normative position when both the normative and deviant targets were ingroup full members (normative conflict = 1). In contrast, we expected the lowest evaluative differentiation and strongest opinion shift towards the deviant opinion when the normative target was ingroup marginal and the deviant was ingroup full member (normative conflict = −1). Finally, in both outgroup conditions we expected moderate evaluative differentiation and no opinion change (normative conflict = 0). Therefore, the former two conditions should reflect respectively, the strongest and the weakest prescriptive focus on the ingroup normative position. Opinion change and evaluative differentiation were the dependent and mediator variables, respectively.
We tested mediation effects using PROCESS analysis with 10,000 bootstrap samples (Hayes, 2013, Model 4; see Figure 2). The overall model was significant, F(2, 48) = 9.25, p < .001, R2 = .28. The joint effect of normative conflict and evaluative differentiation significantly predicted opinion change (indirect effect: b = 0.85, SE = 0.47, 95% CI [0.12, 1.98]). Moreover, the effect of normative conflict on opinion change became nonsignificant when we entered evaluative differentiation in the equation (b = 1.11, SE = 0.69, t = 1.61, p = .115, 95% CI [−0.28, 2.50], indicating that evaluative differentiation fully mediated the association between normative conflict and opinion change.
Mediation model predicting opinion change as a function of normative conflict through evaluative differentiation (Experiment 1).
Our mediation test was based both on theoretical and temporal considerations (given that opinion change was measured after evaluations). However, given that other research suggests that opinion change might precede evaluations of group members in some instances (cf. Frings & Abrams, 2010), we tested a reverse mediation model that treated opinion change as a mediator of differential evaluation. The model was also significant, F(2, 48) = 24.88, p < .001, R2 = .51. The joint effect of normative conflict and opinion change significantly predicted evaluative differentiation (indirect effect: b = 0.23, SE = 0.10, 95% CI [0.06, 0.48]). Importantly, however, the effect of normative conflict on evaluative differentiation remained significant when we entered opinion change in the equation (b = 1.39, SE = 0.27, t = 5.13, p < .001, 95% CI [0.84, 1.93]), indicating that opinion change only partially mediated the association between normative conflict and evaluative differentiation. To summarize, taken together, the two mediation analyses provided relatively strong evidence that evaluative differentiation mediated the effect of normative conflict on opinion change and relatively weak evidence for the reverse pattern of mediation.
Discussion
The results of this study support our predictions. When participants were presented with a normative ingroup full member together with a deviant ingroup full member, they upgraded the former member and derogated the latter, and they increased their agreement with the normative position. This did not occur when targets were outgroup members or when the normative target was a marginal ingroup member and the deviant target was an ingroup full member. In the latter case, participants evaluated the deviant ingroup target more favorably, and they decreased their agreement with the normative position. We also found support for the hypothesis that differential evaluations of normative and deviant targets would mediate the joint impact of membership role (full or marginal) and group membership (ingroup or outgroup) on opinion shift toward or away from the normative position. 6 Our results thus support the idea that the joint enhancement of the normative member and derogation of the deviant member influence participants’ commitment to the ingroup normative position.
These findings are consistent with the idea that when people favor normative ingroup members and derogate deviant ingroup members they do so to sustain a valued normative position (e.g., Pinto et al., 2010). They suggest that substantial evaluative differentiation only occurs when there is strong support for the norm (from a full ingroup member), which empowers others to resist the threat represented by a full member who is deviant. In contrast, in the presence of weak normative ingroup support (from a marginal member), individuals decrease their allegiance to the normative position and shift toward the opinion of the (deviant) full member. Therefore, unlike normative marginal members, normative full members can counteract the negative effects of other full members’ deviant opinions on the subjective validity of the ingroup’s normative position. Consistent with this interpretation, only when observing a normative ingroup full member did participants’ evaluative differentiation between the normative and the deviant ingroup targets predict their subsequent (greater) endorsement of the normative position. These results are in line with SGDT’s assumption that evaluations of normative and deviant ingroup (full) members reflect individuals’ motivation to uphold the normative position that sustains their social identity.
The findings also partially help disambiguate the interpretation of Pinto et al.’s (2010) results. Specifically, whereas Pinto et al. (2010) showed that joint full membership was sufficient for the BSE, their research did not demonstrate that it is necessary. The present study showed that, in the presence of a deviant full member, the normative target must also be a full member in order for the BSE to occur. However, the question remains whether, in the presence of a normative full member, the deviant target must also be a full member. Besides addressing this question, Experiment 2 also replicated Experiment 1. Such a replication is particularly desirable given that the power of the statistical analyses in Experiment 1 was compromised by the relatively small cell sizes in this study.
Experiment 2
Earlier we differentiated four situations involving two ingroup members, one who disagrees with the group’s position (deviant member) and one who agrees (normative member). Experiment 1 investigated situations in which the deviant is a full member and the normative member is either a full or marginal member. Experiment 2 replicates these two situations and adds two more in which the deviant member is a marginal member and the normative member is either a full or marginal member. This design allows us to compare all four combinations of normative and deviant target members’ roles (full vs. marginal) when targets are either members of an ingroup or an outgroup. In addition, it allows us to test alternative interpretations of our findings in Experiment 1.
One alternative interpretation is that more extreme differential evaluation and opinion change occurred when both targets were full members not because they both occupied this particular role but simply because they both occupied the same role. This interpretation, which is contrary to our theory, implies that there should also be extreme judgments and increased agreement with the normative position when the two ingroup targets are both marginal members.
Another alternative interpretation, which is also contrary to our theory, is that opinion shift depended exclusively on the role held by the normative ingroup target rather than on the balance between the threat posed by the deviant member and the support provided by the normative member. This interpretation implies that, regardless of the deviant member’s role, opinion shift toward the normative position would increase in the presence of a normative ingroup full member and remain unchanged or decrease in the presence of a normative ingroup marginal member.
To recapitulate our earlier argument, we assume that participants’ evaluations of deviant ingroup members and subsequent opinion change are based on the degree to which these members threaten the ingroup normative position and on the strength of normative support provided by other members. We assume that such threat is much stronger when deviants are ingroup full members than when they are ingroup marginal members. Conversely, normative support is stronger when provided by an ingroup full member than by an ingroup marginal member. Therefore, consistent with the results of Experiment 1, we expected that the role of the normative ingroup target (full or marginal) should strongly affect evaluations of both the normative and the deviant ingroup targets. Differential evaluations of normative and deviant targets, in addition, should mediate the joint impact of membership role (full or marginal) and group membership (ingroup or outgroup) on opinion shift toward or away from the normative position.
In the two conditions of the present study in which the deviant is a full member, we expected to replicate the pattern obtained in Experiment 1. In contrast, in the two new conditions in which the deviant is a marginal member, we expected a different pattern of results. In these latter conditions, the deviant should not threaten the normative position, regardless of whether the normative member is full or marginal. Therefore, when the deviant is a marginal member, we did not expect polarized evaluations of the two targets or substantial change in participants’ position toward or away from the group norm. As a corollary to this hypothesis, we expected the deviant ingroup full member to be judged less unfavorably than all other deviant targets when accompanied by a normative ingroup marginal member. This is because when the accompanying normative ingroup member is marginal, the full membership of the deviant makes it harder to derogate the deviant without also implicitly derogating the ingroup.
As regards participants’ change in agreement with the normative position, consistent with Experiment 1, we expected (a) a stronger shift towards the normative opinion when both normative and deviant ingroup targets are full members than in all other conditions and (b) a stronger shift towards the deviant opinion when the normative ingroup target is a marginal member and the deviant ingroup target is a full member than in all other conditions. Because outgroup members’ roles and opinions should be largely irrelevant, this should not occur when the deviant and normative members belong to the outgroup. Thus, in line with Experiment 1 and our theory, we again expected that differential evaluations of normative and deviant targets should mediate the joint impact of membership role (full or marginal) and group membership (ingroup or outgroup) on opinion shift toward or away from the normative position.
Method
Participants and design
Participants were 26 male and 87 female students (N = 113) recruited in the campus of two Portuguese universities. Sex, university membership, and age were equally distributed across experimental conditions. 7 A 2 (targets’ group: Ingroup vs. Outgroup) x 2 (deviant target’s role: Full Member vs. Marginal Member) x 2 (normative target’s role: Full Member vs. Marginal Member) x 2 (target’s opinion: Normative vs. Deviant) mixed design was used. Targets’ group, deviant target’s role, and normative target’s role were between-participants factors; target’s opinion was a within-participants factor.
Procedure
The procedure was the same as that used in Experiment 1. As before, we collapsed the items measuring identification with the ingroup university to create an ingroup identification score (Cronbach’s α = .74). Moreover, we computed a normative target and a deviant target score from the items used to evaluate each of these targets (Cronbach’s alphas, respectively, were .89 and .83). Finally, we computed an opinion change score.
Results and Discussion
Ingroup identification
A Targets’ Group x Deviant Target’s Role x Normative Target’s Role ANOVA on the ingroup identification score yielded no significant effects, F(1, 105) always ⩽ 2.41, ns. Participants identified with their group equally across experimental conditions, and they identified with their ingroup university as shown by the significant difference between their mean ingroup identification score and the midpoint of the response scale, M = 4.64, SD = 0.88; t(112) = 7.78, p < .001, 95% CI [0.48, 0.80].
Prior agreement with the normative and deviant opinions
We conducted a Targets’ Group x Deviant Target’s Role x Normative Target’s Role x Opinion (agreement with normative opinion, agreement with deviant opinion) repeated-measures ANOVA on participants’ agreement with the normative and the deviant opinions in the first session of the study. We obtained only a significant effect of opinion, F(1, 105) = 1152.05, p < .001, ηp2 = .916. Participants agreed more with the normative (M = 6.35; SD = 0.76) than with the deviant (M = 2.01; SD = 0.94) opinion. Thus, prior to the manipulation of members’ roles, participants did not differ across conditions in their agreement with the normative and deviant opinions.
Evaluations of target members
We conducted a Targets’ Group x Deviant Target’s Role x Normative Target’s Role x Target’s Opinion repeated-measures ANOVA on the normative target and deviant target scores. Directly relevant to our predictions, we found a significant effect of Targets’ Group x Deviant Target’s Role x Normative Target’s Role x Target’s Opinion, F(1, 105) = 10.57, p = .002, ηp2 = .091. 8
We decomposed the four-way interaction according to deviant target’s role. We found the expected significant Targets’ Group x Normative Target’s Role x Target’s Opinion interaction only in the two deviant full member conditions, F(1, 110) = 10.00, p = .002, ηp2 = .083 (in the two deviant marginal member conditions: F(1, 110 < 1). Therefore, in the deviant full member conditions, we decomposed the Targets’ Group x Target’s Opinion within the conditions defined by the normative target’s role factor. Results yielded a significant Targets’ Group x Target’s Opinion interaction only when both the normative and deviant targets were full members, F(1, 110) = 14.22, p < .001, ηp2 = .114. When the normative target was marginal member and the deviant target was a full member, F(1, 110) < 1. As predicted, we found a BSE in the former condition. The normative ingroup target was judged more favorably than the normative outgroup target, F(1, 110) = 6.79, p = .010, ηp2 = .058, and the deviant ingroup target was judged more unfavorably than the deviant outgroup target, F(1, 110) = 12.45, p = .001, ηp2 = .101 (see Table 3).
Evaluations of normative and deviant targets as a function of the deviant target’s role, normative target’s role and targets’ group (Experiment 2).
Note. 7 = favorable evaluation; 1 = unfavorable evaluation. Standard deviations are presented in brackets.
p = .001. **p < .001.
We also decomposed the full interaction by targets’ group. Results were significant within the ingroup condition, but not within the outgroup condition, respectively, F(1, 110) = 14.13, p < .001, ηp2 = .114, and F(1, 110) < 1. In the ingroup condition, we found a marginally significant Deviant Target’s Role x Normative Target’s Role interaction for normative target evaluations, F(1, 110) = 3.22, p = .075, ηp2 = .028. Concomitantly, the Deviant Target’s Role x Normative Target’s Role interaction was significant for deviant target evaluations F(1, 110) = 21.55, p < .001, ηp2 = .164.
We expected that, when presented with ingroup targets, participants (a) would judge the normative target more favorably when both targets were full members than in all other conditions and (b) would judge the deviant target more unfavorably in that condition and less unfavorably in the condition where the normative target was a marginal member and the deviant target was a full member than in the two conditions in which the deviant target was a marginal member. To test these predictions, we conducted contrast analyses for the normative and deviant target evaluation scores within the ingroup condition. To examine normative target scores, we assigned the values of +3 to the condition in which both the normative and deviant targets were full members and −1 to all the other conditions. To examine the deviant target scores, we conducted two contrast analyses. The first analysis tested the prediction that the deviant full member should trigger the most negative evaluation in the presence of a normative full member. Therefore, we assigned the value of −3 to the condition in which both the normative and deviant targets were full members and +1 to the remaining conditions. The second analysis tested the prediction that, when accompanied by the normative marginal ingroup member, the deviant ingroup full member should be the least negatively evaluated of all deviants. Therefore, we assigned the value of +3 to the condition in which the normative target was a marginal member and the deviant target was a full member and −1 to the remaining conditions. In support of our predictions, all three contrasts were significant, t(54) = 7.20, p < .001, t(54) = 4.28, p < .001, and t(54) = 4.76, p < .001, respectively.
Opinion change
We predicted opinion change (a) in the direction of the normative opinion when both the normative and deviant targets were ingroup full members and (b) in the direction of the deviant opinion when the normative target was an ingroup marginal member and the deviant target was a full member. We first conducted a Targets’ Group x Deviant Target’s Role x Normative Target’s Role ANOVA on the opinion change score. Directly relevant to our predictions, we found a marginally significant Targets’ Group x Deviant Target’s Role x Normative Target’s Role interaction, F(1, 105) = 3.09, p = .082, ηp2 = .029 (see Table 4). 9 We decomposed the interaction by targets’ group. As predicted, results were significant within the ingroup conditions but not within the outgroup conditions, respectively, F(1, 110) = 6.68, p = .011, ηp2 = .057, and F(1, 110) < 1. 10
Agreement with the normative opinion and agreement with the deviant opinion in the first and second sessions across conditions (Experiment 2).
Note. 7 = I fully agree; 1 = I fully disagree. Standard deviations presented in brackets.
Finally, we compared the opinion change score in each condition to the value of 0 (no change). In support of our hypothesis, when both the normative and deviant targets were ingroup full members, participants significantly reinforced their adherence to the normative position from the first to the second session of the experiment, t(15) = 2.80, p = .014, 95% CI [0.22, 1.65]. Concomitantly, when the normative target was an ingroup marginal member and the deviant was a full member, participants significantly shifted toward the deviant position between the two sessions, t(13) = −3.51, p = .004, 95% CI [–3.112, –0.74]. No significant changes occurred in the remaining conditions, t always < 1.70, ns (see Table 4 and Figure 3). Together, these results support our opinion change hypothesis.
Opinion change as a function of targets’ group, deviant target’s role, and normative target’s role (Experiment 2).
Mediation model
We predicted that normative and deviant member’s role should affect opinion change through evaluative differentiation especially when both the normative and deviant targets were ingroup full members (predicting strong evaluative differentiation and increased support for the normative opinion) and when the normative target was an ingroup marginal member and the deviant target was a full member (predicting weaker evaluative differentiation and increased support for the deviant opinion). Thus, we coded the normative conflict variable (which was calculated as in Experiment 1) based on this prediction. We assigned the value of +1 to the condition in which both targets were ingroup full members (where we expected stronger evaluative differentiation and stronger shift towards the normative opinion than in all the other conditions); the value of 0 to the conditions in which (a) the normative target was an ingroup full member and the deviant target was a marginal member, (b) both the normative and deviant targets were ingroup marginal members, and (c) all outgroup conditions (where we expected moderate evaluative differentiation and did not expect any opinion change); and the value of −1 to the condition in which the normative target was an ingroup marginal member and the deviant target was a full member (where we expected weaker evaluative differentiation and stronger shift towards the deviant opinion than in all other conditions).
We conducted a mediation analysis, in which normative conflict was the independent variable, evaluative differentiation was the mediator, and opinion change was the dependent variable. Evaluative differentiation and opinion change were significantly related, providing a basis for testing mediation (r = .394, p < .001), for which we used PROCESS Model 4 with 10,000 bootstrap samples (Hayes, 2013). The overall model was significant, F(2, 110) = 10.42, p < .001, R2 = .40. In line with our predictions, the joint effect of normative conflict and evaluative differentiation significantly predicted opinion change (indirect effect: b = 0.58, SE = 0.24, 95% CI [0.14, 1.10]). Moreover, the initial effect of normative conflict on opinion change (b = 1.42, SE = 0.32, t = 4.43, p < .001) was significantly reduced by the inclusion of evaluative differentiation in the equation (b = 0.84, SE = 0.41, t = 2.05, p = .042, 95% CI [0.03, 1.65]; see Figure 4), indicating that evaluative differentiation partially mediated the association between normative conflict and opinion change.
Mediation model predicting opinion change as a function of normative conflict through evaluative differentiation (Experiment 2).
We also tested the reverse mediation model. This model was also significant, F(2, 110) = 41.82, p < .001, R2 = .41. Interestingly, here opinion change operated as a partial suppressor of the impact of normative conflict on evaluative differentiation. That is, the effect of normative conflict on evaluative differentiation (b = 1.42, SE = 0.32, t = 4.43, p < .001) significantly increased when we included opinion change in the equation (direct effect: b = 1.52, SE = 0.21, t = 7.32, p < .001, 95% CI [1.11, 1.94]; indirect effect: b = 0.18, SE = 0.09, 95% CI [0.05, 0.40]). Thus, the reverse mediation analysis is consistent with our hypothesis that evaluative differentiation mediated opinion change rather than vice versa. Indeed, in this case, on the contrary, opinion change suppressed rather than mediated the direct effect of normative conflict on evaluative differentiation.
Discussion
In Experiment 2, we further disambiguated the interpretation of Pinto et al. (2010) by demonstrating that, in order for the BSE to occur: (a) in the presence of a deviant full member, the normative target must also be a full member; and (b) in the presence of a normative full member, the deviant target must also be a full member. Therefore, the simultaneous presence of a full normative and a full deviant member is both necessary and sufficient to produce the BSE. In line with our predictions, we also found that normative and deviant member’s role affected opinion change through evaluative differentiation and that this occurred especially when both the normative and deviant targets were ingroup full members and when the normative target was an ingroup marginal member and the deviant target was a full member.
Furthermore, this experiment tested predictions that contrasted with our theoretical position. The first was that extreme evaluative judgments of normative and deviant targets and increased agreement with the normative position would occur when the two targets are both marginal members as well as when they are both full members. The second was that opinion shift toward the normative position would increase in the presence of a normative ingroup full member and remain unchanged or decrease in the presence of a normative ingroup marginal member regardless of the deviant member’s role. As predicted by our theory, the answer to both questions was “no.”
The present results support our hypotheses and replicate and extend the findings of Experiment 1. Participants upgraded the normative ingroup full member and derogated the deviant ingroup full member (as compared to all other targets) only when they were presented simultaneously. Concomitantly, participants evaluated the deviant ingroup full member less negatively when a normative ingroup marginal member accompanied this target than in all other conditions. In addition, participants reinforced their agreement with the normative opinion only when both targets were ingroup full members. In contrast, when participants were presented with a normative ingroup marginal member and a deviant ingroup full member, they shifted their position toward the deviant opinion. 11
According to SGDT, individuals upgrade normative ingroup members and derogate deviant ingroup members in order to sustain positive ingroup differentiation (e.g., Marques et al., 2001; Pinto et al., 2010). However, we have now shown that this process operates most strongly when both the deviant and the normative members occupy an important role (full member) in the group. Simply being faced with both an ingroup deviant member and an ingroup normative member may not be sufficient to generate extreme evaluative differentiation between these members or to increase participants’ agreement with a normative ingroup position. These outcomes depend, not on the roles of salient ingroup members per se, but rather on the balance of their role relationships. Only the full member role endows the deviant with a threatening potential, which turns support provided by the normative full member into a particularly useful resource to sustain adherence to the group’s normative position. On the other hand, if the normative member is marginal, a deviant full member actually has the potential to undermine or change the norm itself.
General Discussion
Our results have interesting implications for understanding the processes that lead groups to change their norms or to resist such change. Responses to ingroup deviance were strongly affected by the marginal versus full member role not merely of the deviants but also of the normative members in the immediate context of judgment. Indeed, participants were mildly tolerant of deviants who were ingroup marginal members, as they were of deviant outgroup members in general. However, they were either very tolerant or very rejecting of deviants who were ingroup full members, depending on the role normative members occupied in the group. Deviant ingroup full members were tolerated when they were accompanied by normative marginal members, but they were rejected when they were accompanied by normative full members.
These findings demonstrate that group members are quite vigilant to the potential implications of deviance—deviants who are marginal members, and thus may be viewed as prototypically peripheral have little chance of influencing the group and they can be tolerated, perhaps as a way of showing that the group is respectful of diversity (cf. Hogg & van Knippenberg, 2003; Hogg, van Knippenberg, & Rast, 2012). In contrast, deviants who are full members and thus may be viewed as prototypically central pose a serious challenge to the group’s position. In this case, the presence of a normative full member is a catalyst for rejection of the deviant. Similarly, the presence of the deviant full member stimulates praise of the normative member. The evaluative reactions engendered by this dynamic balance influence group members’ adherence to one of the positions at stake. Only through derogating the deviants or being more accepting of them can group members mitigate the threat they pose to their social identity. Importantly, the normative position espoused by the ingroup may begin to be transformed if the deviant is a full member and the normative member is a marginal one (see Chan, Louis, & Jetten, 2010).
The present findings suggest that group members can effectively resist deviant opinions espoused by ingroup full members when their normative beliefs are backed by other ingroup full members. However, when support for such normative beliefs is fragile (espoused only by marginal ingroup members) and deviance is powerful (espoused by deviant full members), group members shift away from the normative position. This strongly suggests that the absence of normative full members together with the presence of deviant full members undermines members’ certainty about the prescriptive normative position. As a result, they may perceive that the true norm is actually closer to the deviant’s position than they had formerly believed, precisely because they expect full members to espouse the group norm (see Abrams et al., 2008; Levine & Moreland, 1994; Randsley de Moura et al., 2010). It may be that deviants are particularly likely to succeed in shifting group norms when they can gain the ear of other members in the absence of the contrasting voice of a normative full member.
Interestingly, these ideas are compatible with minority influence research showing that groups often resist change by attributing dissenters’ behavior to internal dispositions, thus discounting the existence of objective grounds for change (Levine, 1989; Moscovici & Personnaz, 1986; Papastamou & Mugny, 1985). Levine and colleagues’ group socialization model (e.g., Levine & Moreland, 1994; Levine et al., 2005; Moreland & Levine, 1982) suggests that attributing deviance to idiosyncrasy may be difficult if the deviant person has already proved capable of eliciting group acceptance by attaining the full member role. The potential conflict generated by such a situation (see Mugny, 1980; Pérez & Mugny, 1996) should be even stronger when the contrasting (normative) opinion is espoused by another member who is in the process of leaving the group (i.e., a marginal member).
The increased influence of the deviant full member may correspond to the initial stage of the conversion effect posited by minority influence researchers (e.g., Martin, 1998; Prislin & Filson, 2009; cf. Moscovici, 1980, 1985). When ingroup normative members are marginal in the group, it seems plausible that deviant ingroup full members acquire the ability to elicit divergent thinking whereby group members consider alternative arguments that validate the deviant position (Martin & Hewstone, 2008; Nemeth, 1986; see also Goodman, Alexander, Chizhik, Chizhik, & Eidelman, 2010). As a result, if a deviant member occupies a high-status role in the group others become more tolerant of the person (Abrams et al., 2008) and more accepting of his or her position (see Levine & Moreland, 1985; Levine et al., 2005). Deviants who want to change their group might be well advised to first become established as full members and then attempt to persuade other members (Hollander, 1958), particularly through creating uncertainty about the ingroup’s norms and beliefs.
Footnotes
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.