In a population-based cohort study in Southern Brazil, 87.3% of 5914 liveborn infants were followed for over 12 months. The 215 infant deaths occurring in this cohort were studied in relation to birthweight, gestational age and socioeconomic status. Causes of death were ascertained through the review of case notes and interviews with the parents. As predicted, there was a strong inverse association between birthweight and neonatal and postneonatal mortality. Preterm infants with an adequate weight for their gestational age, despite being slightly heavier than small-for-dates, showed an IMR which was twice as large as the latter. The relative risk associated with low birthweight for deaths due to respiratory infections was over twice as large as that for diarrhoea deaths, but there were only 25 deaths in each category and the estimates are therefore not very precise. There was an interaction between birthweight and socioeconomic status, with the relative risk of mortality associated with low birthweight being much larger among rich than among poor infants. Estimates of the magnitude of the reduction in infant mortality which would accompany a given improvement in the birthweight distribution, which have been mostly based on data from developed countries, may prove to be overoptimistic.