During 1991-1994, anonymous screening of newborn infants for maternal antibody to human immunodeficiency virus (HIV) was carried out in three regions of Spain: Valencia, Galicia and Sevilla. The newborn infants whose heel-stick blood eluates were satisfactory for HIV antibody tests were a consecutive series of 104 876, representing 99.3% of all newborn infants undergoing routine metabolic screening and estimated as comprising at least 98% of all births in the three regions. Enzyme immunoassay (EIA) positives were confirmed by immunoblot, yielding 246 confirmations: a rate of 2.3 per 1000. Seropositivity rates ranged from 1.4 per 1000 in Galicia to 2.1 in Sevilla and 3.1 in Valencia, and remained relatively stable in each region during the years of the study. Within socioeconomically defined subgroups of birth hospitals in Valencia and Galicia, all subgroups contained seropositives, even though there was a twofold to fivefold over-representation in the "inner city" public hospitals. To estimate the proportion of HIV-1-seropositive newborn infants who were positive for HIV-1 DNA, polymerase chain reaction (PCR) assays were performed on 165 dried blood spots that had been retained following positive immunoblot assays. Fifteen (9%) were PCR positive, and when this proportion is adjusted for the age-specific sensitivity of the method, it translates into an estimated HIV-1 transmission rate of 24% (range 18-36%). For 94,906 of the 104,876 newborn infants screened, the EIA used could detect antibodies that react with epitopes of HIV-1 and HIV-2. There were 30 newborn infants whose blood eluate was positive by this combined HIV-1/HIV-2 antibody screen and whose secondary screening with monovalent HIV-2 and HIV-1 EIA indicated that the HIV-2 reactivity was above the cut-off whereas the HIV-1 was not. Ranking these 30 results according to absolute HIV-2 reactivity and relative reactivity with respect to HIV-1 indicated that four infants were probable true HIV-2 seropositives and a total of 12 were possible HIV-2 seropositives, a prevalence of the order of 1:10000 to 1:20000 newborn infants. These anonymous population-based serological studies provide "leading-indicator" data to complement traditional AIDS surveillance for epidemiological and planning purposes.