HIV prevalence estimates for pregnant women from unlinked anonymous surveys are becoming increasingly available and can be used to determine the prevalence of HIV in women in the same population. The ratio of prevalence in pregnant women to that in all women is influenced by HIV-related risk behaviours that are different for pregnant and nonpregnant women and also by differences in fertility level among infected and uninfected women. This ratio is affected by biases that are likely to be culturally and socially specific. A model is proposed for the qualification and quantification of these biases and hence the estimation of general female population prevalence from serosurveillance data on pregnant women.