The reliability of medical information collected from proxy respondents for deceased occupational health study subjects was investigated. A cohort of 586 workers from two chemical manufacturing factories (who had worked at least 1 day in either factory after 1955) was used in this study. A list of occupational exposure related health effects was collected from each living member of the study group. For the 147 deceased study subjects, occupational exposure health effect data were collected from a proxy respondent. Selection of proxy respondent was made on the basis of death certificate information, and information collected from newspaper obituaries and funeral directors. Proxy respondents were contacted via mail and telephone, and those who agreed to participate in the study were administered a questionnaire regarding the decedent's demographic characteristics, personal habits, and occupational and medical history. The information collected from the proxy respondents was compared to decedent's medical record data. Of the 147 first proxy respondents contacted, 96 participated in the study and 73 were able to identify potential second proxy respondents. Of these potential second proxy respondents, 59 participated in the study. Questionnaire nonresponses were lowest among spouses, and highest among friends and nonimmediate relatives. Agreement of response between paired respondents was high regarding such conditions as ulcer, cancer, diabetes, and lung disease. From this study the authors conclude that the use of proxy respondents as sources of medical information on deceased study subjects can be problematic, and that an effort should be made to select a proxy respondent best suited to provide the medical information needed.