Using data on all 1,901 U.S. Supreme Court oral arguments between 1999 and 2013, we document that voice-based snap judgments based solely on the introductory sentences of lawyers predict Justices votes. The connection between vocal characteristics and court outcomes is specific to perceived masculinity even when judgment of masculinity is based only on less than three seconds of exposure to a lawyer’s speech sample. Although previous studies suggest a significant role for vocal characteristics on real world behavior, prior to our work none has identified a definitive connection using identical phrases. Roughly 30% of the association between voice-based masculinity and court outcomes comes from within-male lawyer variation, whereas 70% comes from between-male lawyer variation. Moreover, voice-based first impressions predict both male and female lawyers’ court outcomes: less masculine males and more feminine females are more likely to win. A de-biasing experiment separately identifies statistical discrimination and prejudice by showing that information reduces 40% of the correlation between perceived masculinity and perceived win and incentives reduces another 20% of the correlation. The negative correlations between perceived masculinity and win rates were stronger in private firms and in petitioner classes with more masculine voices. Perceived masculinity explains an additional 10% of variance relative to best existing prediction models of Supreme Court justice votes. Sincere and strategic voting considerations may explain why liberal justices were more likely to vote against male lawyers perceived as more masculine and conservative justices were more likely to vote for female lawyers perceived as more feminine.