The finding that minorities may tend to produce conversion whereas majorities may tend to produce compliance is an extremely important one for the study of social influence. Most research into minority conversion has been based on conversion theory (Moscovici, 1980) which, we would claim, rests on the notion that the 'true' influence exerted by minorities is possible only because they are essentially different from self. This paper reports two studies in a programme of research testing an alternative explanation, based on self-categorization theory, which assumes that perceived similarity to a potential source of influence will be the key to its deep and lasting success. Both studies employ a full majoritylminority X in-group/out-group design, socially significant reallife in-groupout-group memberships and measure attitudes directly relevant to these social identities. Study 1 uses immediate and delayed and Study 2 , public and private, responses as the measures of, respectively, compliance and conversion. The results support self-categorization theory in that, when exposed to both majority and minority out-group sources, subjects exhibited an immediatdpublic polarization away from the sources, towards a more extreme in-group position, and there was no diminution of the extremity of their position on delayed/private measures. The classic pattern of majorities bringing about greater compliance than conversion and minorities greater conversion than compliance was evinced in the in-group conditions of both studies, although this tended to be comparative rather than absolute. The implications of the results for the conflict between self-categorization theory and conversion theory are discussed and the future direction of our research indicated.

Self-categorization theory hypothesizes that group polarization arises from an intragroup process ofconformity to apolarizedin-group norm, i.e. a norm more extreme than the mean of members' initial opinions but in the same direction. I t argues that the ingroup norm is embodied in the prototypical response of members, which is not necessarily their average position. Polarization occurs when the prototype is more extreme than the mean in the same direction and when mutual influence leads members to conform to the norm. Convergence on the mean occurs when the prototype coincides with the mean position. Whether the prototype and the mean coincide or differ depends on whether group responses are moderate or extreme in terms of the comparative social context. A computer simulation (Study 1, N = 1000 groups) and two semi-naturalistic group discussion studies (Studies 2 and 3, Ns = 32 and 44 groups respectively over five items) were conducted to test this analysis. All three studies found significant correlational support for the hypothesis (HI) that in-group prototypes become more polarized as the group becomes more extreme in the social context and Studies 2 and 3 also found a significant empirical correlation (€12) between polarization and the extent to which group members conformed to a polarized prototype.

The aim of the reported studies was to provide a self‐categorization‐based explanation of the unique temporal patterning of minority influence (David, 1994; David & Turner, 1992, 1996; Turner, 1991) as being based on changes in social context. Moderate feminist participants were exposed to an in‐group minority (separatist feminist) message in no explicit context, an intragroup, or an intergroup context, and immediate and delayed measures of influence were taken. The explicit contexts were either stated once or repeated prior to the delayed measure of influence. Participants rejected the minority message when it was presented in a purely feminist context and accepted it in a context which included the salient outgroup. When no explicit context was stated and when an intragroup context was stated prior to the immediate measure only, participants displayed the classic conversion pattern of no immediate influence followed by evidence of delayed influence. Discussed are implications of the results for the self‐categorization claim that influence can only proceed from those categorized as similar to self, and the conflict of this proposal with conversion theory (Moscovici, 1980, 1985; Moscovici & Mugny, 1983; Moscovici & Personnaz, 1980) and conflict elaboration theory (Mugny, Butera, Sanchez‐Mazas & Perez, 1994; Mugny & Perez, 1987, 1991), which we understand to explain minority conversion as a product of the essential difference from self of prospective sources of influence.

Theorists suggest that gender differences in moral reasoning are due to differences in the self-concept, with women feeling connected to others and using a care approach, whereas men feel separate from others and adopt a justice approach. Using a self-categorization analysis, the current research suggests that the nature of the self-other relationship, rather than gender, predicts moral reasoning. Study 1 found moral reasoning to be dependent upon the social distance between the self and others, with a care-based approach more likely when interacting with a friend than a stranger. Study 2 suggests that when individuals see others as ingroup members they are more likely to utilize care-based moral reasoning than when others are seen as outgroup members. Further, traditional gender differences in moral reasoning were found only when gender was made salient. These studies suggest that both the self and moral reasoning are better conceptualized as fluid and context dependent.

Role congruity theory predicts prejudice towards women who meet the agentic requirements of the leader role. In line with recent findings indicating greater acceptance of agentic behaviour from women, we find evidence for a more subtle form of prejudice towards women who fail to display agency in leader roles. Using a classic methodology, the agency of male and female leaders was manipulated using assertive or tentative speech, presented through written (Study 1, N = 167) or verbal (Study 2, N = 66) communications. Consistent with predictions, assertive women were as likeable and influential as assertive men, while being tentative in leadership reduced the likeability and influence of women, but not of men. Although approval of agentic behaviour from women in leadership reflects progress, evidence that women are quickly singled out for disapproval if they fail to show agency is important for understanding how they continue to be at a distinct disadvantage to men in leader roles.