Waterborne gastroenteritis outbreaks have often gone undetected or been incompletely defined in terms of source and extent. Methods which allow detection or clarification of such events are therefore useful. We describe the methods used to detect and investigate such an outbreak. In autumn 1996 high school absence rates and the rate of parents absent from work to care for sick children suggested a health problem in a Swedish town which had a history of unexplained outbreaks of gastrointestinal disease. A systematic sample of 300 households was surveyed by post. Respondents represented 10% of the total population of the town. Questions concerning symptoms and exposures were included. The same questionnaire was used in a nearby town as a control. Sixty four percent of respondents reported an acute gastrointestinal illness during a two month period. Diarrhoea (90%) and abdominal pain (88%) were the most frequent symptoms among the sick. Two percent of those sick sought medical care. Exposures associated with disease were being a member of a large household, young age, and consumption of water from the community water supply. Attack rate showed a dose response relationship with increasing frequency of water consumption. The peak incidence of gastrointestinal illness occurred shortly after raw water quality control data had shown a rise in indicator bacteria. Further analysis, dividing those infected into groups according to when they became ill and whether they were the first member of their household to fall ill, supported the hypothesis of primary cases being infected from the water supply with some secondary person to person spread.