To estimate the proportion of cases missed in a passive surveillance study of diarrhoea and dysentery at health centres and hospitals in Kaengkhoi district, Saraburi province, Thailand, a community-based cluster survey of treatment-seeking behaviours was conducted during 21-23 June 2002. Interviews were conducted at 224 households among a study population of 78,744. The respondents reported where they sought care for diarrhoea and dysentery in children aged less than five years and adults aged over 15 years. Health centres or hospitals were the first treatment choice for 78% of children with dysentery (95% confidence interval [CI] 63-94%), 64% of children with diarrhoea (95% CI 54-74%), 61% of adults with dysentery (95% CI 40-82%), and 35% of adults with diarrhoea (95% CI 17-54%). A high degree of heterogeneity in responses resulted in a relatively large design effect (D=3.9) and poor intra-cluster correlation (rho=0.3). The community survey suggests that passive surveillance estimates of disease incidence will need to be interpreted with caution, since this method will miss nearly a quarter of dysentery cases in children and nearly two-thirds of diarrhoea cases in adults.