Although responses to 24-hour recall and knowledge-attitude-practice questionnaires are commonly used in water-sanitation studies as surrogates for direct observation of behaviour, the validity of this approach is questionable. We therefore compared questionnaire data with those obtained by direct observation of practices related to water storage, handwashing, and defecation among 247 families in urban Dhaka, Bangladesh. Analysis of the results indicates that accord between the replies to the questionnaires and the data collected by direct observation was poor and that the responses to the two questionnaires were often contradictory. Significant disagreements between the results of questionnaires and observations arose usually because desirable practices were over-reported by the respondents. The results of the study suggest that in urban Bangladesh 24-hour recall and knowledge-attitude-practice questionnaires should not be used as proxies for direct observation of hygiene practices.