OBJECTIVES: This study examines, in the context of the human immunodeficiency virus (HIV) epidemic, the effects of optimal breast-feeding, complete avoidance of breast-feeding, and early cessation of breast-feeding. METHODS: The three categories of breast-feeding were weighed in terms of HIV transmission and infant mortality. Estimates of the frequency of adverse outcomes were obtained by simulation. RESULTS: Avoidance of all breast-feeding by the whole population always produces the worst outcome. The lowest frequency of adverse outcomes occurs if no HIV-seropositive women breast-feed and all seronegative women breast-feed optimally, given infant mortality rates below 100 per 1000 and relative risks of dying set at 2.5 for non-breast-fed compared with optimally breast-fed infants. For known HIV-seropositive mothers, fewer adverse outcomes result from early cessation than from prolonged breast-feeding if the hazard of HIV transmission through breast-feeding after 3 months is 7% or more, even at high mortality rates, given relative risks of dying set at 1.5 for early cessation compared with optimal duration of breast-feeding. CONCLUSIONS: The risk of HIV transmission through breast-feeding at various ages needs to be more precisely quantified. The grave issues that may accompany a possible decline in breast-feeding in the less developed world demand evaluation.