We frequently encounter members of stereotyped groups that do not fit the stereotype, and one might think at these encounters would eventually bring about stereotype change. Intuitively, this should be particularly true when the atypical group members all contradict the stereotype in the same way. However, search indicates that stereotypes are very difficult to change and, in particular, that subgroups of highly typical group members are subtyped, or mentally segregated from the rest of the group such that they are ineffective in bringing about stereotype change. The studies in this proposal investigate variables that might moderate the effectiveness of highly atypical group members in encouraging stereotype change. In addition, the studies reported here investigate the abilities of connectionist networks to model learning bout atypical group members. Connectionist models of memory are receiving substantial attention in the cognitive literature, and it is probable that incorporating them into our thinking about social aspects of behavior and thought will serve to 1) encourage thinking about social memory from a new perspective that ill potentially lead to new predictions about social phenomena, and 2) allow social phenomena to inform he development of these models that are becoming increasingly prevalent throughout psychology.
