Non-response bias is a potentially serious problem in studies involving human subjects. A parental questionnaire/negative consent form was used in a dental epidemiological survey of 5-year-old children in Leeds. The aim of this study was to determine whether and to what extent there was any variation in caries experience and oral cleanliness between responding and non-responding groups. 6,494 children were examined and completed questionnaires were returned for 4,069 (63 per cent). The response was higher for white children (65 per cent) than for Asian (39 per cent) and Afro-Caribbean children (45 per cent). Lower response rates were found from parents living in inner city areas associated with social deprivation. Children of non-responders had poorer oral cleanliness and higher caries experience (mean dmft). This study demonstrated that members of a non-response group are likely to be substantially different from responders.