In a population-based cohort of approximately 6000 Brazilian children, the associations between maternal education and a number of child health outcomes were studied while controlling for potentially confounding variables such as family income and education of the husband. In the crude analyses, maternal education was associated with perinatal and infant mortality, hospital admissions in the first 20 months of life and the three nutritional indicators (length-for-age, weight-for-age and weight-for-length) at mean age 20 months. After adjustment for confounding, the apparent associations with outcomes in early infancy--birthweight and perinatal mortality--were no longer present, while that with infant mortality persisted despite being reduced. Strong associations remained with later outcomes including hospital admissions, length-for-age and weight-for-age at mean age 20 months. Among infants born to women with little or no schooling, deaths due to diarrhoea, pneumonia and other infectious diseases were particularly common. These findings support the hypothesis that maternal education has an effect on child health which is partly independent from that of other socioeconomic factors; they also suggest that maternal care is more important than the biological characteristics of the mothers since stronger effects were observed for the late (postneonatal mortality, hospital admissions and nutritional status) than for the early (birthweight, perinatal mortality) outcomes.