A review was presented of potential sources of inaccuracy in data collecting for community based epidemiologic studies and aspects of questionnaire validation. A preliminary study of the accuracy of work histories obtained from subjects' relatives in a case referent lung cancer study was reported. Accuracy of questionnaire derived information was directly dependent on questionnaire structure, language used and interviewer attitude. Inaccuracies were observed to arise from memory failure or recall bias of subjects. Possible sources of bias were identified by a method of internal verification which included parameters of completeness, level of detail achieved and incoherent or impossible information. External validation was considered, but finding a valid external source was often problematic. Information obtained from surrogates was often inaccurate, and surrogate type influenced information obtained. A study of 281 male cases of lung cancer compared subjects' responses with those of the closest living relative after the deaths of 249 of the subjects. Final surrogate responding population included 197 relatives, primarily wives (73%). Subjects were interviewed, and surrogates answered mail questionnaires. For major occupational categories, validity revealed high specificity while sensitivity was sometimes quite low, corresponding to occupations which were less common and thus not as well known in the region. The kappa statistic generally demonstrated moderate or high agreement. For information for the period 15 to 24 years previous to disease onset (latency), sensitivity and specificity were improved. Methods of exposure assessment were reviewed. Generation of a-priori or a- posteriori job exposure matrices were compared, and benefits of using lists of workplace chemicals were noted. The authors conclude that quality of exposure assignments is affected by exposure assessment information quality, and questionnaire validation is inseparable from conception, formulation and administration methods.