In the context of the Demographic and Health Surveys program (DHS), data were collected on diarrhoeal diseases in childhood and related treatment patterns. In this paper we assess the accuracy and completeness of mothers' recall of diarrhoea in 19 national DHS surveys and discuss the implications for health interview surveys in developing countries. It is concluded that there is under-reporting of diarrhoea if the recall period is longer than 2-3 days, whereas there may be over-reporting of very recent or current diarrhoea in most DHS surveys. Reporting errors appear to vary considerably between countries, which affects the comparability of survey results. A second and related issue, that is addressed in this paper, is the reporting of treatment practices by duration of diarrhoeal episode. There were no major differences in reported treatment patterns between children with diarrhoea that terminated in the last two weeks and children with current diarrhoea of at least two days' duration. The implications of the findings for retrospective surveys on childhood morbidity and treatment patterns are discussed.