Abstract: Variations in birth sex ratios and sex differences in juvenile mortality occur in a number of mammalian species, and in many cases have been linked to resource availability. Most of these biases in offspring sex ratios concern polygynous species with pronounced sexual dimorphism, and where females only are philopatric. Data on species with unusual life-history strategies, such as slight sexual dimorphism or dispersal by both sexes, are of particular interest. In this study of a natural herd of horses (Equus caballus) which experienced an eruptive cycle, and therefore a period of nutritional stress, male offspring had higher neonatal mortality rates in nutritionally poor years than in good ones, whereas “year quality” had no effect on the mortality of female offspring; year quality could therefore be used by mares as predictor of sex-specific offspring survival. We show that the environmental conditions that predicted lower survival of males were negatively related to their production: the birth sex ratio the following year was female-biased; and mares were less likely to produce a son when they had produced a son the preceding year. There was no significant effect of mother's parity, age or rank, or the timing of conception or birth on offspring sex ratios. The mechanism leading to biases in the birth sex ratio could have been the loss of male embryos by mares that did not foal. As there was no evidence for selective abortion of male foetuses in females that did foal the next year, it is not necessary to invoke maternal adjustment, though this remains a possibility. Finally, there was a suggestion that male offspring were more costly to raise than females, since mothers that reared a son in poor years tended to experience an increase in the interbirth interval between their two subsequent offspring.