The purpose of this study was to compare the performance of a number of standard screening instruments for alcohol dependence and harmful drinking/abuse by ethnicity (black, Hispanic and white) and by ethnicity and gender in an emergency room setting. A probability sample of patients (N = 1,429) was breath analyzed and interviewed at the Santa Clara Valley Medical Center in San Jose, California. Sensitivity and specificity were analyzed among current drinkers (n = 857) for the CAGE, Brief MAST, AUDIT, TWEAK, RAPS and other items against combined ICD- 10 or DSM-IV criteria for alcohol dependence and separately for alcohol dependence or harmful drinking or abuse. Screening measures were not found to perform equally well by ethnicity or gender, with lower sensitivity found for women compared to men. Consistency in sensitivity of measures was found to vary considerably across ethnic and gender groups, with some measures (most notably the RAPS and the AUDIT) showing consistently high sensitivity across subgroups. None of the instruments performed nearly as well for identifying alcohol dependence or harmful drinking or abuse combined as for alcohol dependence alone. Analyses suggest that, while the RAPS may hold promise for identifying problem drinkers across ethnic and gender subgroups, it and other screening instruments currently in use require additional evaluation in a variety of settings to determine their usefulness for identifying those who could benefit from a brief intervention or referral for problem drinking.