Abstract: We investigate whether prenatal testosterone can explain gender differences in educational performance. Prenatal testosterone is elevated for individuals with a male co-twin due to testosterone transfers in utero. Given that the sex of the co-twin is determined randomly, this provides exogenous variation in prenatal testosterone which can be used to study a potential biological determinant of gender gaps in educational performance. Administrative data on a large sample of Dutch twins is used. After controlling for the effects of growing up with a same sex or opposite sex sibling, we find no improved performance for boys with a male co-twin. Girls with a male co-twin perform 0.07 of a standard deviation worse on math. No effects are found for language and an aggregate ability score. An elevated level of prenatal testosterone decreases math performance for girls.