The comparative validity of the Michigan Alcoholism Screening Test (MAST) and the Alcohol Dependence Scale (ADS) in screening for current DSM-III alcohol abuse/dependence disorders is evaluated. These scales were administered to 501 patients presenting for treatment of alcohol or drug problems. DSM-III alcohol disorders are diagnosed using the Diagnostic Interview Schedule. Receiver Operating Characteristic (ROC) analysis is used to determine optimum threshold scores for the MAST and ADS and to compare the screening ability of the two instruments. Optimum cut points for the MAST and the ADS are 12/13 and 8/9, respectively. The overall accuracy of classification for both instruments using these threshold scores is 88%. The areas under the ROC curves are .91 and .90 (SD = .02) and there are no significant differences between the MAST and the ADS in their ability to screen for alcohol abuse or dependence in this population. The MAST and the ADS correlate highly with each other (.79). The results reported in our study should be applicable to the revised DSM-III since a field trial found a high level of agreement on alcohol disorders between the diagnostic systems. Categorical versus dimensional approaches to the assessment of alcoholism are discussed.