The validity of self-report in alcoholism treatment research is controversial. Our recently completed Veterans Administration Cooperative Study evaluating the efficacy of disulfiram treatment for alcoholism provided an opportunity to assess the validity of self-report. To assess treatment response, patients and household contacts were interviewed at seven scheduled points during the 1 year of follow-up. Blood specimens also were obtained from the patients at these times and were analyzed for ethanol. Eighty-eight percent of the patient and/or collateral interviews were obtained at 6 months and 90% at 1 year. The mean number of blood and urine specimens collected per patient was 4.3 and 14.4, respectively. Outcome criteria included continuous abstinence during the year and total number of drinking days. Continuous abstinence: If we had had only the patients' self reports, we would have significantly underestimated the percentage of men who drank. By self-report 58.7% (355/605) relapsed whereas the combination of self report, collaterals' reports, and laboratory tests indicated that 72.4% (438/605) drank (p less than 0.001). Using Bayes' theorem, the conditional probability that a patient is continuously abstinent for 1 year when he so claims is 65%. Total drinking days: Of the 213 patient-collateral pairs each of whom provided all seven scheduled interviews, 46.9% (100/213) agreed on the total number of drinking days during the year.(ABSTRACT TRUNCATED AT 250 WORDS)