If two-parent care has different consequences for the reproductive success of sons and daughters, then natural selection may favor adjustment of the sex ratio at birth according to circumstances that forecast later family structure. In humans, this partnership status hypothesis predicts fewer sons among extra-pair conceptions, but the rival "attractiveness" hypothesis predicts more sons among extra-pair conceptions, and the "fixed phenotype" hypothesis predicts a constant probability of having a son, regardless of partnership status. In a sample of 86,436 human births pooled from five US population-based surveys, I find 51.5% male births reported by respondents who were living with a spouse or partner before the child's conception or birth, and 49.9% male births reported by respondents who were not (X2=16.77, d.f. = 1, p<.0001). The effect was not explained by paternal bias against daughters, by parental age, education, income, ethnicity, or by year of observation, and was larger when comparisons were made between siblings. To my knowledge, this is the first direct evidence for conditional adjustment of the sex ratio at birth in humans, and could explain the recent decline in the sex ratio at birth in some developed countries.