In this study, sex ratios at birth (male/female births) were found to vary according to family composition. Using Demographic and Health Survey (DHS) maternity histories from sub-Saharan Africa, the study shows that the sex ratio at birth increases with the number of previous male births and decreases with the number of previous female births. For families with only males, the sex ratio increases from 1·046 for the first birth to 1·083 for the 8th birth. For families with only females, the sex ratio decreases from 1·046 for the first birth to 0·942 for the 8th birth. The differences were highly significant when tested with a linear logistic model (p=0·018 for males; p=1·85✕10−11 for females). The effect was not symmetrical, and was found to be significantly stronger for females. These effects could be reproduced assuming a strong heterogeneity between couples. The distribution of sex ratios was fitted with an asymmetrical log-gamma function, which revealed a wide range of variation in the sex ratio between 0·50 and 1·30, and a peak around 1·14. The results and their implications are discussed in the light of former findings in France and in the United States of America.