OBJECTIVE To evaluate the sensitivity and specificity of several alcohol screening instruments (CAGE, brief MAST, AUDIT, TWEAK, History of Trauma Scale) and other measures (breath alcohol analyzer reading and reporting of drinking before the event), in an emergency department population, against International Classification of Disease, revision 10 (ICD-10) criteria for harmful drinking and for alcohol dependence from the Composite International Diagnostic Interview (CIDI) by gender, race, and injury status. METHODS A probability sample of patients was subjected to breath alcohol analysis and interviewed (N = 1,330.) METHODS University of Mississippi Medical Center. RESULTS Overall, the TWEAK and AUDIT methods were the most sensitive, identifying 84% and 81%, respectively, of patients who were positive for alcohol dependence. Sensitivity was not as high for females, whites, or the noninjured. Sensitivities for the breath alcohol analysis and self-reported drinking were 20% and 29%, respectively. CONCLUSIONS These data suggest that the effectiveness of screening instruments varies by gender, race, and injury status and that positive breath alcohol analysis readings and reporting of drinking before the event are not good indicators of alcohol dependence in this population.