The present study examines the impact of non-diagnostic individualized information on judgments concerning stereotypic and non-stereotypic behaviours performed by ingroup versus outgroup members. Ninety-five Law School students were confronted with either a lawyer or a teacher in mathematics and were given either the category label alone, the category label plus written individualized informations or the category label plus a videotaped interview of the target. Subjects indicated the likelihood that the target would perform each of four behaviours, two of which were stereotypic and two of which were counter-stereotypic of the target's professional category and this in four different situations. In accordance with classical dilution results (Nisbett, Zukier and Lemley, 1981), individualized information weakened the influence of stereotypes. Also, in line with Park and Rothbart's (1982) work on the outgroup homogeneity hypothesis, judgments were more extreme for the outgroup than for the ingroup target. However, a derivation of Linville's (1982) complexity-extremity hypothesis was not supported: the dilution effect was not more effective for the outgroup than for the ingroup target. Lastly, our data argue against Nisbett et al.'s explanation for the lack of dilution these authors found for stereotype-irrelevant behaviours. Taken together, our results may be seen from an optimistic as well as from a pessimistic point of view. The latter perspective points out the pervasive polarization of judgments about outgroup targets across individualization levels whereas the former stresses upon the dilution taking place for both in-and outgroup targets.