Based on free descriptions of men and women by male and female respondents, a language approach to outgroup homogeneity and discrimination between gender groups is presented. Verbalizations elicited by 10 topic cues (e.g., sexuality, education of children, and housework) were coded by valence (positive vs. negative), group reference (ingroup vs. outgroup), and the abstractness of the predicate (descriptive action verb, interpretive action verb, state verb, and adjective). Outgroup discrimination was confined to female respondents, as manifested in their tendency to describe men in more negative terms than women. Outgroup‐related statements in general, and outgroup‐derogative statements in particular, were more abstract than statements about the ingroup. However, outgroup discrimination was due more to repetition of biased statements than to differential abstraction. The distribution of discrimination across topics is neither related to conflict proneness nor to the empathic experience associated with intimate (e.g., sexuality) as opposed to other topics (e.g., driving). Outgroup homogeneity is apparent in the consistency of language profiles characterizing outgroup descriptions for different topics but not in generalization over outgroup members. Discrimination and homogeneity are systematically related across topics, and this relation is not due to evaluative constraints. Rather, discrimination is strongest for topics for which the most distinctive and reliable language profiles are available. These results are discussed in the context of a language‐based approach to intergroup af