We took the opportunity provided by a large outbreak of cryptosporidiosis in the North West of England to investigate the potential impact of recall bias on strength of association and on estimates of outbreak size. We conducted a community-based survey of 4 towns within the outbreak area and 4 control towns. A postal questionnaire was sent to 120 homes, chosen at random from the local telephone directory, in each of the 8 towns. Although not statistically significant, the prevalence of self-reported diarrhoeal disease was marginally higher in the control towns than in the outbreak towns. There was a very strong association between self-reported diarrhoea and drinking water consumption in both control and outbreak areas. The impact of recall bias in outbreak investigations is much greater than previously thought. Identification of the cause of outbreaks should not be based solely on case-control studies conducted after the press has reported the outbreak. Such evidence is likely to be unreliable and give falsely significant associations between water consumption and disease.