The quality of occupational histories as provided by surrogate respondents was considered, specifically addressing how often surrogates have no knowledge of the index subject's work history or are unavailable for interviews, how often the information matches the information from the subjects themselves or from company records, and what impact does exposure misclassification resulting from the use of surrogate data have on the estimate of relative risk. A number of features were shown to potentially be associated with the ability of the surrogate to provide complete and valid occupational data. The surrogate nonresponse is reduced if a spouse of a close relative is the surrogate used, as the incomes increases, as less detail is requested, if the events in question are more recent, and if there is little chance that the subject ever worked in the industry or occupation in question. Studies have indicated that the information provided by the surrogate is more close to the actual fact if the information sought pertains to the last job as opposed to the entire work history. Misclassification of exposure occurs even when great care is taken. When a study includes a mixture of surrogate and self respondents, nondifferential misclassification can be explored by stratifying the analyses on respondent status. Using a structured questionnaire when dealing with surrogates and administering the questionnaire carefully helped prevent affirmation bias.