This report describes the conceptual and empirical basis for the development of a screening instrument to identify persons with potentially harmful alcohol consumption. As part of a larger project sponsored by the World Health Organization (WHO), alcoholic (N = 65) and nonalcoholic (N = 187) research volunteers completed a battery of assessments that included laboratory tests, a physical examination, a diagnostic interview, personality measures, and two standard self-report screening questionnaires. The data were analyzed to evaluate the validity of diagnostic measures that could subsequently be used to develop a briefer screening test. The results of a construct validity analysis indicated that the new diagnostic measures correlated well with generally accepted alcoholism screening tests (the MAST and MacAndrew scales) and with measures of hypothetical vulnerability (e.g., sociopathy and childhood problems). Analysis of discriminant validity indicated that alcohol-specific self-report measures differentiated well between male risk groups, but were less effective in identifying high risk females. In general, alcohol-specific measures differentiated best, followed by clinical and laboratory tests and vulnerability assessments. It is concluded that no single procedure is universally suitable for the early identification of harmful drinkers. The design of a screening test will depend on the purpose of screening, the groups to be identified, the resources available and the level of cooperation to be expected from the population screened.