Past research on member-to-group generalization has failed to distinguish the effect of member descriptive deviance from the effect of member evaluative deviance. In addition, researchers have used group judgments that confound stereotypicality with prejudice. Three experiments resolved these methodological problems and provided the first systematic test of HamburgerÕs (1994) model of stereotype change. In Experiment 1 (N ¼ 60), consistent with HamburgerÕs predictions, exposure to one deviant group member increased perceived group dispersion, but did not affect judgments of group stereotypicality and prejudice. Experiment 2 (N ¼ 120) replicated these results in an interpersonal setting, but not in an intergroup setting. Experiment 3 (N ¼ 125) replicated the results of Experiment 1 when a memberÕs profile conveyed information about eight stereotype-relevant dimensions, but not when it conveyed information about only four stereotype-relevant dimensions. We discuss the results in the light of past evidence and future strategies for stereotype change.