First impressions from faces are consequential, shaping important social outcomes. In this thesis, based on converging evidence from empirical studies and computational models, I show that women are at a disadvantage because of gender biases in impressions. First, impressions of women are less differentiated than impressions of men and more highly valence-laden (Part I). Specifically, impressions of social traits (e.g., trustworthiness, dominance) are more highly intercorrelated for women than for men. Second, computational models of first impressions show that although people use similar facial information when forming impressions of women and men, they interpret this information differently (Part II). For instance, while masculinity cues tend to contribute to positive impressions of men, they tend to contribute to negative impressions of women. Third, this is particularly salient in the case of impressions of competence (Part III). Specifically, when faces are manipulated to look competent but not attractive, competence impressions correlate highly with multiple measures of masculinity. The current work shows that (1) impressions of women are more simplified than those of men; with positive impressions of women being contingent on the typicality of their looks and (2) women are perceived as less competent than men; with competence impressions of women being contingent on their attractiveness. The current work highlights the importance of social categorization in first impressions and reveals a major obstacle to gender equality and social justice at large.