What processes best explain women’s underrepresentation in science, math, and engineering fields in the U.S.? Do they also explain men’s underrepresentation in the humanities? Two survey studies across two U.S. West Coast universities (N = 62; N = 614) addressed these questions in the context of two fields: one male-dominated (computer science) and the other female-dominated (English). Among a set of social predictors—including perceived similarity to the people in the field, social identity threats, and expectations of success—the best mediator of women’s lower interest in computer science and men’s lower interest in English was perceived similarity. Thus, changing students’ social perceptions of how they relate to those in the field may help to diversify academic fields.