The reported study compared change in stereotypic perceptions of homogeneous and heterogeneous groups, when subjects were presented with a pattern of stereotype-inconsistent information that was either concentrated in two extreme group members or dispersed across six members. Results pro vided some support for the ‘conversion’ model (in which stereotypes change in response to salient instances) in the case of a homogeneous group, where stereotypical responding was lower in concentrated than dispersed conditions. In the heterogeneous-group conditions, there was no effect of pattern. In addition, subjects' estimates of stereotype-consistent information were higher, and of inconsistent information were lower, and they perceived more members as typical, and fewer as atypical, when the target group was heterogeneous versus homogeneous. There was also support for the ‘subtyping’ model (in which disconfirming individuals are isolated from other group members) in the concentrated conditions. A theoretical account of these findings is given in terms of stereotype change via salience for homogeneous groups, and the need to integrate research on cognitive models of stereotype change and perceived group variability.