It has been shown that mother-to-child human immunodeficiency virus type 1 (HIV-1) transmission can occur both during pregnancy and at delivery, but the respective frequencies in these periods are unknown. Moreover, it is difficult to determine the timing of mother-to-child HIV-1 transmission by direct sampling. The use of an elaborate statistical method is therefore necessary. The authors studied 495 consecutive infants born between May 1988 and August 1991 who were included, at birth, in the French Prospective Study on Pediatric HIV Infection. At least one blood sample was obtained from every infant during the first 14 days of life. All samples obtained within 3 months of birth were tested by at least two of the following methods: viral culture, polymerase chain reaction (PCR), and antigenemia, as well as by Western blot test. Data for the 95 infected infants (those seropositive at 18 months and those who died of HIV disease before this age), and who were exclusively bottle-fed, were analyzed in a Markov model to estimate the timing of viral transmission, the time from birth to the emergence of detectable virus, and the time from birth to seroconversion. The model indicated that one-third of the infants were infected in utero, less than 2 months before delivery (95th percentile). In the remaining 65% of cases (95% confidence interval (CI) 22-92), the date of infection was estimated as the day of birth. The estimated median period between birth and the emergence of viral markers was 10 days (95% CI 6-14) and the 95th percentile was estimated at 56 days. These results support the view that HIV infection can be diagnosed during the first 3 months of life. The authors conclude that mother-to-child HIV-1 transmission appears to occur late in pregnancy or at delivery.