This paper reports the results from an experiment where men and women are paired toform a two-member team and asked to execute a counting task. An individual’s payoff isproportional to the joint production of right answers. Participants who perform better thantheir partner in the task can punish him or her by imposing a fine. We manipulate the pairs’gender compositions to analyze whether an individual’s performance and sanctioningbehavior depend on his or her gender and the gender of his or her partner, which is revealedto the subjects at the beginning of the experiment. The data show that, conditional on underperformance,women are sanctioned more often and more heavily than men; however, ifthey are sanctioned, men tend to improve their performances, while women’s performancesdo not change.