The emergence of gender as a major area of interest in medical sociology in the 1970s set an exciting agenda for future research. However, despite a growing literature, our understanding of the mechanisms linking gender and health has advanced little in recent years. This may, in part, be due to problems in conceptualising gender. In particular, despite theoretical advances, biological sex and sociological gender continue to be conflated in empirical research. In this paper we argue that sex and gender are not necessarily coincidental and that gender roles and gender role orientations may cross-cut sex. We draw upon a recent tradition of research on gender role orientation in psychology which assesses the level of ‘bmasculinity’ and ‘femininity’ in an individual's self-concept through identification with traditionally ‘masculine’ and ‘feminine’ characteristics. Considering a range of health measures we conclude that the sex (male-female) differences that are traditionally found in research on gender and health may mask an association of ‘femininity’ with relatively poor health and ‘masculinity’ with relatively good health in both men and women.