A number of instruments have been developed to screen for alcoholism. With the advent of DSM-III and lay administered psychiatric diagnostic instruments, a test of the performance of these screens relative to diagnostic instruments is critical. In this paper, we document the relative effectiveness in a general medical clinic of the Michigan Alcoholism Screening Test (MAST), the Veterans Alcoholism Screening Test (VAST), and the CAGE questions in comparison to the DSM-III-R criteria for alcohol dependence as measured in the Composite International Diagnostic Interview (CIDI). All of the screens performed at acceptable levels, but the MAST and VAST tended to have higher performance characteristics. At the recommended cut points, they had higher sensitivity for lifetime alcohol dependence (VAST 95.1%, MAST 90.2%, CAGE 78.0%) as well as higher specificity (VAST 80.3%, MAST 81.7%, CAGE 76.1%). For present alcohol dependence only, at the recommended cut points the MAST and CAGE had sensitivity of 100% but specificity of 62.0 and 61.0% respectively. The VAST had sensitivity of 83.3% and specificity of 89.0%. We conclude that all three perform well relative to DSM-III-R criteria.