Attempts to develop effective treatment and preventive strategies for chronic diarrhoea have been confounded by inconsistencies in defining the disorder. In the present paper, we illustrate some of the methodological considerations encountered in using a distributional approach to define chronic diarrhoea. We employ data obtained from 3470 children aged less than six years residing in Dhaka, Bangladesh, between January 1985 and January 1986. The importance of defining the age of the target population is demonstrated by the inverse correlation between age and duration of episodes required to reach or exceed the 95th percentile, ranging from 27 days in children aged less than 12 months to 18 days in children aged over four years. Altering the percentile cut-off qualifying as 'chronic' from the 95th percentile to the 90th percentile changed the minimum qualifying duration from 24 to 16 days. We also demonstrate the impact of altering criteria for termination of an episode and of altering the duration of the monitoring period on the resultant length of episodes classified as 'chronic'.