Policy interventions to increase women's presence in the workforce and leadership positions vary in their intensity, with some including a lone or token woman and others setting higher quotas. However, little is known about how the resulting group gender compositions influence individuals' experiences and broader workplace dynamics. In this paper, we investigate whether token women are disadvantaged compared to women on majority-women mixed-gender teams. We conducted a multi-year field experiment with a top-10 undergraduate accounting program that randomized the gender composition of semester-long teams. Using laboratory, survey, and administrative data, we find that even after accounting for their proportion of the group, token women are seen as less influential by their peers and are less likely to be chosen to represent the group than women on majority-women teams. Token women also participate slightly less in group discussions and receive less credit when they do. Women's increased authority in majority-women teams is driven primarily by men's behavior, not homophily or self-assessment. We find that over time, the gap in general assessments of influence between token and other women shrinks, but this improvement does not carry over to task-specific assessments. Finally, predictors of future grades are different for token women than for other participants, and regardless of treatment condition, women's task expertise is incorporated into group decisions less often than mens. Our findings have implications for team assignments in male-dominated settings and cast significant doubt on the idea that token women can solve influence gaps by "leaning in."