To review evidence for the impact of HIV on fertility from empirical sources pertaining to African populations and to discuss the implications for surveillance based on antenatal clinic populations. The theoretical equivalence between the fertility rate ratio for HIV-positive to HIV-negative women and the relative odds of being infected for pregnant women compared with the general female population is demonstrated. This equivalence is used to compare fertility differentials measured in cohort studies with those calculated indirectly from antenatal clinic data. Data from case-control studies and theoretical predictions from a model of the proximate determinants of fertility and HIV incidence are used to obtain plausible explanations of the fertility differences. Estimates of population attributable change are made. Fertility of HIV-positive women is lower than that of HIV-negative women, in all but the youngest age-group, and the differential increases with women's age and epidemic duration. Selection for early start of sexual activity explains the reverse pattern at younger ages. Lower fertility amongst HIV-positive women causes a population attributable decline in total fertility of the order of 0.4% for each percentage point HIV prevalence in the general female population. In populations that do not use contraceptives, HIV-positive women have lower fertility principally as a result of foetal losses consequent to infection with HIV and coinfection with other sexually transmitted diseases; behavioural factors tend to enhance this differential. Other factors being equal, HIV prevalence estimates based on antenatal surveillance underestimate true prevalence in women in the childbearing years.