The objective of this cohort study was to assess risk factors for child dysentery and watery diarrhoea. The study participants consisted of 254 children aged 12-24 months in rural South Africa and Zimbabwe in households where drinking water was collected from communal sources. The main outcome measure was the most severe diarrhoea episode: dysentery, watery diarrhoea or none. For dysentery, drinking water from sources other than standpipes had a relative risk ratio of 3.8 (95% CI 1.5-9.8). Poor source water quality, as indicated by Escherichia coli counts of 10 or more cfu 100 ml(-1), increased risk by 2.9 (1.5-5.7). There were no other significant risk factors for dysentery and none for watery diarrhoea. In this study, endemic dysentery is associated only with faecal contamination of source water. Sources other than standpipes, including improved groundwater, are of greater risk. Remediation of water quality by treatment at source or in the household will be required to achieve access to safe drinking water in accordance with the 7th Millennium Development Goal.