Oral fluid is a non-invasive biological sample, which can be returned by post, making it suitable for large-scale epidemiological studies in children. We report our experience of oral fluid collection from 14 373 preschool-aged children in the UK Millennium Cohort Study. Samples were collected by mothers in the home setting following the guidance of trained interviewers, and posted to the laboratory. Samples were received from 11 698 children (81·4%). Children whose mothers were of Black Caribbean ethnicity and who lived in non-English-speaking households were less likely to provide a sample, and those with a maternal history of asthma more likely to provide a sample [adjusted risk ratio (95% CI) 0·85 (0·73–0·98), 0·87 (0·77–0·98) and 1·03 (1·00–1·05) respectively]. Collection of oral fluid samples is feasible and acceptable in large-scale child cohort studies. Formal interpreter support may be required to increase participation rates in surveys that collect biological samples from ethnic minorities.