A review of the published literature on the impact of water supply and/or excreta disposal facilities on diarrhoeal diseases, or on infections related to diarrhoea, reveals several methodological problems that hamper the drawing of definitive conclusions from these studies. This paper examines eight of these methodological problems: lack of adequate control, the one to one comparison, confounding variables, health indicator recall, health indicator definition, failure to analyse by age, failure to record usage, and the seasonality of impact variables. It is suggested that an evaluation of the impact on health of environmental interventions may best be undertaken by the combined efforts of engineers, social scientists and epidemiologists in 'opportunistic' settings and that the intervening behavioural processes so necessary for health impact to occur should be a primary focus of such evaluations.