Although it has long been recognized that stereotypes achieve much of their force from being shared by members of social groups, relatively little empirical work has examined the process by which such consensus is reached. This paper tests predictions derived from self-categorization theory that stereotype consensus will be enhanced (a) by factors which make the shared social identity of perceivers salient and (b) by group interaction that is premised upon that shared identity. In Experiment 1 (N=40) the consensus of ingroup stereotypes is enhanced where an ingroup is judged after (rather than before) an outgroup. In Experiment 2 (N=80) when only one group is judged, group interaction is shown to enhance the consensus of outgroup stereotypes more than those of the ingroup—an apparent ‘outgroup consensus effect’. In Experiment 3 (N=135) this asymmetry is extinguished and group interaction found to produce equally high consensus in both ingroup and outgroup stereotypes when the ingroup is explicitly contrasted from an outgroup. Implications for alternative models of consensus development are discussed. © 1998 John Wiley & Sons, Ltd.