Described are the findings of a multicentre cohort study to test an algorithm for the treatment of persistent diarrhoea relying on the use of locally available, inexpensive foods, vitamin and mineral supplementation, and the selective use of antibiotics to treat associated infections. The initial diet (A) contained cereals, vegetable oil, and animal milk or yoghurt. The diet (B) offered when the patient did not improve with the initial regimen was lactose free, and the energy from cereals was partially replaced by simple sugars. A total of 460 children with persistent diarrhoea, aged 4-36 months, were enrolled at study centres in Bangladesh, India, Mexico, Pakistan, Peru, and Viet Nam. The study population was young (11.5 +/- 5.7 months) and malnourished (mean weight-for-age Z-score, -3.03 +/- 0.86), and severe associated conditions were common (45% required rehydration or treatment of severe infections on admission). The overall success rate of the treatment algorithm was 80% (95% CI, 76-84%). The recovery rate among all children with only diet A was 65% (95% CI, 61-70%), and was 71% (95% CI, 62-81%) for those evaluated after receiving diet B. The children at the greatest risk for treatment failure were those who had acute associated illnesses (including cholera, septicaemia, and urinary tract infections), required intravenous antibiotics, and had the highest initial purging rates. Our results indicate that the short-term treatment of persistent diarrhoea can be accomplished safely and effectively, in the majority of patients, using an algorithm relying primarily on locally available foods and simple clinical guidelines. This study should help establish rational and effective treatment for persistent diarrhoea.