To date questionnaire surveys have been the most commonly used instruments to measure hygiene behaviours related to water and sanitation. More recently, a number of studies have used structured observations to study practices related to diarrhoea. During a trial of a hygiene education intervention to reduce diarrhoea among young children in Bandundu, Zaire, both instruments were used to measure the disposal of child faeces and various hand-washing practices. Three hundred families were observed and follow-up interviews performed with 274 (91%) mothers. At the individual level, agreement between observed and reported behaviour was little better than might be expected by chance. There was evidence of over-reporting of hand-washing before food preparation (44% vs 33%; P = 0.03), hand-washing before eating (76% vs 60%; P < 0.001) and disposal of the child's faeces in a latrine (75% vs 40%; P < 0.001). On the other hand, hand-washing before feeding the child was reported less often than it was observed (7% vs 64%; P < 0.001). Our data are consistent with the hypothesis that, in general, mothers over-report 'desirable' behaviours. At the same time, our data indicate that open questions may lead to under-reporting of certain behaviours. The repeatability of observations at both the individual and population levels remains to be established.