As part of the evaluation of an alcoholism orientation program conducted in 12 federal agencies, 378 female employees were asked to provide personal information about problems associated with their use of alcohol. The optimized form of the unrelated question randomized response technique (RRT) was used to provide for a comparison of estimates of frequency of problem drinking obtained with guaranteed confidentiality of response versus estimates obtained using a conventional anonymous questionnaire. The estimated proportion of alcoholics or possible alcoholics among participants in the orientation program was 34.3% (+/- 4.9) by the RRT and 21.9% (+/- 3.6) by direct response. Significant underreporting of alcoholism was found among older respondents (greater than or equal to 36 years), lower GS level respondents (less than grade 6), those with low seniority (less than 8 years), and those never married. Further, significant underreporting was found among those who claimed they were told they had to attend the program versus others, and among those who claimed they were not curious about the topic of alcoholism versus others. The RRT is recommended for use in experimental situations where answers to sensitive questions are needed to evaluate program success. The necessity for caution in interpretation of apparent differences in respondent characteristics with respect to problems with alcohol use is discussed. Finally, the findings indicate the importance of providing avenues by which confidential treatment for alcoholism can be obtained.