Studies of gender wage differential decompositions based on the human capital model are legion. What makes this exercise an interesting one to undertake for Sri Lanka is that, unlike many developing countries, Sri Lankan females possess an educational advantage over their male counterparts. Yet, returns to education favour women. Thus, the Sri Lankan case is in some sense a natural experiment, providing an indication to countries struggling to close the gender gap in education: if they succeed, what is the magnitude of the wage gap that would persist, and why? Existing studies of the gender wage gap in Sri Lanka show that “discrimination” in the labour market, rather than differences in productive characteristics, account for a large fraction of the gender wage differential. The present study contributes to the literature by (a) analysing changes in and determinants of the gender wage differential during a period when employment opportunities for females in Sri Lanka were expanding rapidly along with their labour force participation, and (b) using a superior methodology than has been used so far in estimates of earnings functions for Sri Lanka. The study reveals that the gender wage gap did narrow between 1985 and 1991, and that returns to education increased over the period for both men and women. Returns to education are greater for women than for men, while “returns” to experience which were initially higher for men in 1985/86 were higher for women in 1990/91. Correcting for omitted variable bias revealed that OLS overestimates the extent of discrimination, but that even so, in the absence of discrimination, women would earn more than men. The study indicated that although returns to education are higher for women than men, initial disadvantages are so great that favourable returns are insufficient to eliminate the gender gap.