If individual abilities are imperfectly observable, statistical discrimination may affect hiring decisions. In our lab experiment, pairs of subjects solve simple mathematical problems. Subjects then hire others to perform similar tasks. Before choosing whom to hire, they receive information about the past scores of pairs, not of individuals. We vary the observability of individuals' abilities by ordering pair members either according to performance, or alphabetically by nickname. We find no evidence of gender discrimination in either treatment, however, possibly indicating that gender stereotypes are of limited importance in the context of our study.