While alcohol use disorders have been found to be overrepresented in some primary care settings, the majority of patients with these disorders go undetected by primary health care providers, and little is known of the performance of screening instruments for identifying problem drinking, particularly among females or ethnic minorities. The performance of several screening instruments—CAGE, Brief Michigan Alcoholism Screening Test (BMAST), Alcohol Use Disorders Identification Test (AUDIT), TWEAK, Rapid Alcohol Problems Screen (RAPS)—and holding five or more drinks (HOLD), against International Classification of Diseases, 10th revision (ICD-10) and Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria for alcohol dependence/harmful drinking/alcohol abuse, was compared by gender in a probability sample of black primary care clinic patients (n = 711). Sensitivity of all measures except BMAST and HOLD was significantly better for males than for females. The RAPS performed best among males (sensitivity, 91%; specificity, 98%), while sensitivity of all measures was unacceptably low among females (none was above 44%). Among current drinkers, 19% of males and 10% of females met diagnostic criteria for an alcohol use disorder, suggesting the importance of identifying those who may benefit from a brief intervention in primary care settings, and suggesting the need for further research in determining those instruments that may have the best potential for identifying problem drinking among female primary care patients.