Background: The CAGE and the Brief MAST questionnaires are widely used to screen for alcohol problems. We tested the performance of these instruments in 2 population-based groups: a high-risk sample composed of relatives of alcoholic subjects and a community sample consisting of families not selected for alcoholism (ie, alcohol dependence disorder). Methods: A total of 3435 relatives of alcoholics and 795 control subjects were interviewed with the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) semistructured interview in a multicenter collaborative study on the genetics of alcoholism. The performance of "CAGE" and "Brief MAST" equivalent items in the SSAGA was characterized by their positive predictive value, sensitivity, specificity, and percentage of the sample who screened positive. Results: Both questionnaires performed well in the high-risk sample, where the base rate of alcoholism was 35%. However, in the community sample, where the 16% rate of alcoholism was comparable to that of the US population (14%), an acceptable positive predictive value could be achieved only through a substantial reduction in sensitivity. Results were similar when men were compared with women and when lifetime alcoholics were compared with current alcoholics. Conclusion: The "Brief MAST" and "CAGE" can be effective instruments to screen for significant alcohol problems in both community and high-risk patients; as expected, their positive predictive value increases with the base rate of alcoholism in the population being screened.