The intergroup contact hypothesis is tested with self-reports of 3,806 survey respondents in seven 1988 national probability samples of France, Great Britain, the Netherlands, and West Germany. With seven key variables controlled, the hypothesis is confirmed-especially for intergroup friendships and affective prejudice. Nonrecursive models indicate that the predicted friends-to-less-prejudice causal path is larger than the prejudice to-fewer-intergroup friends path. These effects generalize in two ways: to immigration polity preferences and a wide variety of out-groups. To explain this generalization, three mediating processes are proposed that can override the many cognitive barriers to generalization: empathy and identification with the out-group and reappraisal of the in-group (deprovincialization). A situation's "friendship potential" is hence indicated as an essential condition for optimal intergroup contact.