Because there is no consensus definition of asthma for epidemiology, we have examined the reliability of a questionnaire and the effect of its mode of administration on classification of asthma in children. A symptom history questionnaire was parent self-administered and then readministered within three months by a nurse. The questions of diagnosed asthma, cumulative wheeze, and recent wheeze (wheeze in the previous 12 months) were more repeatable than questions of night cough, but 7 percent of children changed diagnosed asthma category, 13 percent changed cumulative wheeze category, and 9 percent changed recent wheeze category at second questionnaire. Because the numbers who changed from symptom positive to negative roughly equalled the changes from negative to positive, prevalence estimates were not affected. Methods of measuring asthma with greater precision are urgently needed. Because of reporting bias, epidemiologic information collected by current questionnaires should be treated with some caution.