We conducted a prospective cohort study to determine the post-natal incidence of and possible transmission routes for HIV-I infection in rural Ugandan children. The cohort consisted of the population of a cluster of 15 villages in Masaka District, south-west Uganda, and was enrolled in 1989-1990 through a demographic and serological survey. During the period 1991-1993 the population was resurveyed annually. A total of 5492 children aged 0-12 years were enrolled; of these, 41 (0.7%) were seropositive infants. A total of 3941 (72%) children were HIV-negative on enrolment and had at least one follow-up specimen. During 8596 person-years of observation only I seroconversion was observed, an incidence rate of 0.12 (95% CI 0.00-0.35) per 1000 years of observation. The transmission of HIV was most probably through breast milk. The case corresponds to a rate of 1.1 per 1000 in households with one or more HIV-positive adults (874 years of observation); no incident case was observed in households with only seronegative adults (6423 years of observation). Thus, HIV infection among children aged 0-12 years in this population is virtually exclusively the result of mother-to-child transmission. No infections were observed attributable to parenteral exposure, non-sexual casual or household contact, or insects.