A longitudinal prospective study of the effect of drainage and sewerage systems on diarrhoea in children aged < 5 years was conducted in 9 poor urban areas of the city of Salvador (population 2.44 million) in north-east Brazil in 1989-90. Due to complex political and administrative reasons, 3 areas had benefited from drainage improvements, 3 from both drainage and sewerage improvements, and 3 from neither. An extensive questionnaire was applied to collect information on each child and on the conditions of the household, and mothers recorded diarrhoea episodes in their children aged < 5 years daily for 1 year, using calendars. Fortnightly home visits were made to collect the data. The incidence of diarrhoea in children in neighbourhoods with drainage was less than two-thirds, and in neighbourhoods with drainage and sewerage less than one-third, of the incidence in neighbourhoods with neither. After controlling for potential confounders, the proportion of children with 'frequent diarrhoea' showed the same significant trend across the study groups. Though the groups were not exactly comparable, more than one child was monitored per household, and it was not possible to rotate fieldworkers between study groups, the study provides evidence that community sanitation can have an impact on diarrhoeal disease, even without measures to promote hygiene behaviour.