Heavy maternal alcohol intake is a major perinatal risk but, unfortunately, is difficult for obstetricians and gynecologists to detect. To develop a brief questionnaire appropriate for office detection of "risk-drinking," that is, alcohol intake potentially sufficient to damage the fetus, defined here as greater than or equal to 1 ounce of absolute alcohol per day, we obtained a quantitative drinking history at the first prenatal visit from 971 consecutive gravid women who admitted ever having drunk alcohol. In addition, we administered the 25-question Michigan Alcoholism Screening Test and the four CAGE questions (C = cut down, A = annoyed, G = guilt, E = eye opener), a screening test previously unstudied in pregnancy, and sought evidence of tolerance to the inebriating effect of alcohol, a question which does not appear to trigger psychologic denial. The patient was considered tolerant if it took greater than 2 drinks to make her feel "high." Among 42 (4.3%) risk-drinkers and 929 women who did not report drinking at risk levels, four questions were found to contribute to reliably differentiating risk-drinkers from non-risk-drinkers (R2 = 14.6%, p less than 0.0001). The probability of risk-drinking increased from 1.5% for those responding negatively to 62.7% for those responding positively to all four questions (T = tolerance, A = annoyed, C = cut down, E = eye-opener; odds ratio = 109X). A simple scoring scheme (2 points for T and 1 each for A, C, or E, with a total score of greater than or equal to 2 considered positive) correctly identified 69% of the risk-drinkers (sensitivity) with a positive predictive value of 23%). The T-ACE questions take about 1 minute to ask and represent the first validated sensitive screen for risk-drinking appropriate for routine use in obstetric-gynecologic practice. If validated in further samples, broad application might contribute to better risk identification, secondary prevention efforts, and improved pregnancy outcomes for offspring at risk from heavy prenatal alcohol exposure.