A computer model was developed to assess the impact on under-5 child mortality of breastfeeding practices in developing countries in the context of HIV infection. The model was used to estimate the effect on mortality of cessation of breast-feeding among mothers HIV-positive and mothers HIV-negative at birth, for both urban and rural settings. Using parameter values for a hypothetical East African country, cessation of breast-feeding in urban areas was predicted to result in increases in under-5 mortality of 108% for children of mothers HIV negative at birth, and 27% for those HIV-positive at birth, with slightly larger increases in rural areas, suggesting that breast-feeding should continue to be promoted. A sensitivity analysis was conducted to identify critical values of key variables for which a review of policies encouraging breast-feeding is indicated. This showed that, even under extreme assumptions, cessation of breast-feeding among mothers HIV-negative at birth (but at risk of acquiring HIV during the lactation period) would increase under-5 mortality. For mothers HIV-positive at birth, the key variables are the additional risk of vertical transmission attributable to breastfeeding, the under-5 mortality rate (U5MR) in breast-fed children, and the relative risk of mortality in non-breast-fed compared to breast-fed children. Depending on the values of these key variables, there may be some urban populations with low U5MR in which the positive and negative effects on under-5 mortality of a policy change are finely balanced. However, no change in policy should be made in these areas until more precise information is available on the key variables, and the many adverse consequences of such a change have been fully explored.