A community-based longitudinal study of acute and persistent diarrhoea in 705 children less than five years old was carried out for a year in a rural area of Bangladesh. Diarrhoea morbidity data were collected from each study child every fourth day by home visit. Clinical features of diarrhoeal episodes and diarrhoeal management information were documented. The overall diarrhoeal incidence rate in the study children was 4.6 episodes per child per year. The incidence of persistent diarrhoea was 34/100 child-years. Persistent diarrhoea was positively associated with young age and more severe illness, characterized by the presence of clinical dehydration or blood in the stool in the first week. Use of ORT in the first week was positively associated and use of an antibiotic was negatively associated with the occurrence of persistent diarrhoea. Reduced breast-feeding and consumption of cow's milk at some time during the episode were also positively associated with persistence. This would suggest that appropriate fluid and dietary management for all episodes should be the goal. Children with more severe initial illness characterized by the presence of blood in the stool or clinical dehydration should have more careful follow-up to identify persistent episodes and adverse nutritional effects. Breastfeeding should be continued during acute diarrhoea, but the role of ORT, antibiotics and cow's milk deserves further investigation.