We examine the importance of group membership in stigma and its role in the eectiveness of self-protective cognitions in three experiments. In Experiment 1, men are asked to interact with an attractive female who will judge their value as a potential date, and either eat a mint or a clove of raw garlic prior to the interview. Although the stigmatized-by-garlic men discounted negative feedback and attributed it to their garlic breath, discounting and attributions were negatively correlated with self-esteem. In Experiment 2, White participants were evaluated positively or negatively by a bogus partner who the participants believed had been told that the participant was either White or Black. Although participants receiving negative feedback engaged in several self-protective cognitions, including attributing their negative feedback to racism, the strategies were uncorrelated with self-esteem. In Experiment 3, women prepared to interact via computer with a partner who expressed sexist or non-sexist beliefs. In the absence of feedback, self-esteem increased when their partner was sexist. In contrast with the ®rst two experiments, perceiving the partner as prejudiced was signi®cantly and positively correlated with self-esteem. Together, these experiments suggest that selfprotective cognitions ®nd their eectiveness when stigma has a basis in group membership. Copyright # 2000 John Wiley & Sons, Ltd.Early self theories suggested that self-esteem was based on experience and re¯ected information from the social world about social acceptance, moral value, and personal accomplishments (see Shrauger & Schoeneman, 1979, for a review). Over the past several decades, the early theories have been shown to be over-simpli®ed, limited, and, in some cases, simply wrong. One way in which these theories' predictions have