This study examines sensitivity and specificity figures associated with screens used to predict harmful drinking and alcohol dependence among current drinkers. The study population comes from a probability sample of emergency room patients in Jackson, MS. Data are presented by gender and injury status (injured versus noninjured) for Blacks and for Whites. The Composite International Diagnostic Interview was used to assess ICD-10 criteria for harmful drinking and alcohol dependence, which were taken as standards. Predictors include screening instruments (CAGE, AUDIT, brief MAST, TWEAK, and History of Trauma Scale), breathalyzer reading, self-reported consumption before the injury or noninjury event, quantity and frequency of drinking, and an abbreviated alcohol dependence experiences measure used in general population surveys. Single items from these screening instruments were also tested as predictors. Overall, the TWEAK and the AUDIT performed best in terms of sensitivity and specificity, but variation across subgroups suggests that the search for a good screening instrument for general use must be continued.