This paper explores the contribution of biological factors in explaining gender differences in educational performance, with a particular focus on the role of prenatal testosterone. We exploit the fact that prenatal testosterone is hypothesized to transfer in-utero from a male twin to his twin sibling causing exogenous variation in exposure to prenatal testosterone in twins. By using Dutch administrative data and controlling for potential socialization effects, we find that girls with a twin brother score 7% of a standard deviation lower on math compared to girls with a twin sister. Adherence to traditional gender norms can explain this finding, implying that our results are not just driven by biology but materialize depending on environmental factors.