The incidence and some determinants of low birth weight (LBW) were studied in two population-based cohorts of singletons born live to families in Ribeirão Preto, São Paulo State, Brazil. The first cohort comprised infants born between June 1978 and May 1979 (6750 births--population survey) and the second, infants born between May and August 1994 (2990 births--sample survey). The incidence of LBW was 7.2% in 1978-79 and 10.6% in 1994. After adjustment for confounding factors, the following determinants remained significant in 1978-79: female sex, maternal age > or = 35 years, preterm delivery, < four antenatal health visits, maternal smoking, lower level of maternal education, and manual work/unemployment. In 1994, the significant determinants were preterm delivery, maternal smoking and caesarean section. The adjusted percentage population attributable risk (PAR%) fell for the majority of risk factors but increased for caesarean section, preterm birth, multiparity (> or = 5), primiparity and non-cohabitation. The increase in the rate of LBW from 1978-79 to 1994 was higher for families with more qualified occupations, and occurred only for infants delivered at 36-40 weeks' gestational age and weighing 1500-2499 g, i.e. those most likely to be born by elective caesarean section. The caesarean section rate rose from 30.3% in 1978-79 to 51.1% in 1994. The increase in LBW was probably due to iatrogenic practices associated with elective caesarean section.