A community-based prospective study of diarrhoeal morbidity of weaning-age children in 18 geographically separate village clusters was conducted as the baseline phase of a controlled trial of a hygiene education intervention to reduce diarrhoeal diseases in rural Zaire. For 12 weeks trained interviewers collected information at weekly home visits about the diarrhoeal morbidity of 2082 children aged 3-35 months. Included in the analyses were 1914 children (92%) with 9 or more complete weeks of data. Mothers' reporting of the existence or otherwise of episodes appeared reliable, and more than 70% of children had at least one episode of diarrhoea during the 12-week study period, the mean number of episodes being 1.9 per child. Reporting of the start and termination of diarrhoeal episodes was, however, irregularly distributed among the 7 days between successive home interviews, with 36% of all episodes reported as starting on interview days, and 29% reported as ending the day before an interview: in each case, only 14% would have been expected. After adjusting for these reporting biases, 61% of the episodes lasted 2-4 days, but a number of very long duration were also reported, and the mean duration of the episodes was 4.8 days. Children, had, on average, 9.2 days of diarrhoea during the study period. Contrasts with previous characterizations of reporting error in diarrhoeal studies are noted.