Of 2,996 welfare recipients applying for CETA benefits at the Milwaukee office of Jewish Vocational Service between 3/1/78-9/30/78, a 10% sample (N = 309) was screened for assessment of alcohol problems. After obtaining voluntary informed consent from participants (6% declined), trained interviewers individually administered a 16-item alcoholism At-Risk Questionnaire (ARQ) based on observations by NCA's Criteria Committee; a standard form of the 25-item Michigan Alcoholism Screening Test (MAST); and a 35-item interview structured around a selectively modified version of NCA's Criteria for the Diagnosis of Alcoholism (CRIT). Analyses of data suggested that our ARQ was of little value in discriminating between problem drinkers and other persons, although significantly correlated with MAST and CRIT scores. Using a conventional scoring of the MAST, 53.6% of the sample appeared to have significant alcohol problems, while our CRIT identified only 31.9% as problem drinkers. By combining the MAST + CRIT in a unique scoring system, a more conservative estimate of 36.57% problem drinkers, with an estimated error rate of 1.63% false negatives and 23.45% false positives, was determined. Further modification of MAST + CRIT scoring led to a revised estimate of 25.41% problem drinkers with estimated false-positive and false-negative rates of 7.55% and 6.5% respectively. Implications for research and plans for further modifications of screening procedures are discussed.