Two experiments examined the role of memory for behavioral episodes in judgments about in-groups and out-groups. Using a minimal group paradigm, participants read either positive or negative trait-relevant behaviors performed by group members. They then were asked to make judgments about the group's trait characteristics. Results demonstrated that, for groups described positively, judgments about the out-group but not the in-group were accomplished by retrieving from memory specific behaviors performed by group members. By contrast, for groups described negatively, judgments about the in-group but not the out-group were accomplished by retrieving specific behaviors performed by group members. These results suggest that basic differences in the way judgments about in-groups and out-groups are made contribute to the establishment and perpetuation of intergroup bias by decreasing the stability of negative in-group and positive out-group impressions and increasing the stability of positive in-group and negative out-group impressions.