This paper studies how gender representation affects collective decision-making in expert committees. I exploit quasi-random assignment of judges to panels in the Greek Supreme Court using newly digitized data on 3,700 criminal appeals. I find that panels with more female judges are more likely to reject appeals and less likely to delegate cases. Effects are nonlinear and emerge primarily once at least three of five judges are female; below this level, representation has no detectable effect. The mechanism appears to operate at the panel rather than the individual level - panels with a higher share of female judges take significantly longer to decide, especially in complex cases and in familiar panel compositions, consistent with more thorough deliberation rather than coordination costs. These findings suggest that diversity policies targeting modest increases in female representation will have limited impact unless they shift the deliberative composition of the group itself.