Interviewer-administered National Heart and Lung Institute (NHLI) and British Medical Research Council (BMRC) respiratory questionnaires were compared with each other and with a self-administered questionnaire of our own design in 2,350 adults enrolled in a longitudinal community study. There was a basic 10% disagreement between responses to any 2 questionnaires for all questions that inquired about a perception of a complaint or a disease, but much less disagreement for more factual questions, such as those concerning smoking. Little effect was noted from minor variations in wording, order of questions, method of administration, inclusion of other questions, or time between questionnaires (as long as this interval was less than 1 month). For questions with similar wording, the BMRC and NHLI questionnaires yielded very similar results in terms of over-all prevalence of responses, relationships to answers on an independent questionnaire, and inter-relationships of positive responses. The new self-completion questionnaire used in this study detected more abnormalities and better delineated cough and phlegm "syndromes" than either the NHLI or the BMRC questionnaire, and we believe that self-administration was a very satisfactory technique in the type of population surveyed in this study.