Using the same questionnaire, personnel, and field procedures, we conducted interviews in 1985 with the next-of-kin of 108 deceased study subjects, and with 197 surviving subjects who had been interviewed earlier in a 1980 case-control study. Concordance in response between the two interviews for cigarette smoking, coffee drinking, and alcoholic beverage intake was high, generally 90% or above. Next-of-kin, particularly spouses, provided information as reliable as the self-informants for cigarette smoking and coffee consumption, but showed a somewhat higher discrepancy rate for alcohol drinking. Thus, surrogates, especially spouses, appear to be an accurate source of information in epidemiologic investigations.