A case-control study of environmental and behavioural risk factors for childhood diarrhoea was conducted in Kurunegala district, Sri Lanka. From five hospitals, 2458 children aged less than 5 years and suffering from diarrhoea were recruited as clinic cases, and a further 4140 reporting with complaints other than diarrhoea were recruited as clinic controls. Community-based cross-sectional surveys were also conducted in three of the five areas served by these hospitals, and from these a further 1659 children were recruited as community controls. Children from households where excreta were reported to be disposed of in a latrine were less likely to have diarrhoea than children whose families improperly disposed of excreta. The results obtained from comparisons of cases with clinic controls (adjusted odds ratio [OR] 1.42, 95% confidence interval [CI] : 1.01-1.98), and of cases with community controls (OR 1.35, 95% CI : 0.85-2.13) were in agreement, suggesting that no important selection bias operated on this association. If the observed proportion (91%) of improper excreta disposal among the population could be reduced to 50%, 12% of childhood diarrhoea episodes would be prevented. Although latrine ownership may be a necessary condition for safe excreta disposal behaviour, diarrhoeal morbidity may only be reduced in Sri Lanka if behavioural changes take place concomitant with the construction of sanitation facilities.