OBJECTIVE: To assess the performance characteristics of the Brazilian version of the Tolerance, Annoyed, Cut down and Eye-opener (T-ACE) questionnaire to screen alcohol consumption during pregnancy. METHODS: Observational, cross-sectional study in a sequential sample of 450 women in the third trimester of pregnancy, attended in a maternity ward in a city of Southeastern Brazil, in 2001. The following instruments were used: a questionnaire to gather sociodemographic data, the T-ACE, a questionnaire to verify history of alcohol consumption throughout gestation, and a clinical interview to identify the harmful use of and dependence on alcohol, according to ICD-10 diagnostic criteria. Concordance tests among different interviewers as well as test-/re-test reliability tests were performed. RESULTS: A total of 100 women (22.1%) were identifi ed as positive by the T-ACE. The kappa indexes for concordance and reliability were 0.95, with 97% of concordant responses. When compared to the ICD-10 criteria and to the pattern of consumption, the T-ACE, with a cut-off point of two or higher, presented sensitivity and specifi city coeffi cients of 100% and 85%, and of 97.9% and 86.6%, respectively. CONCLUSIONS: The Brazilian version of the T-ACE seemed to appropriately meet the performance criteria that qualify it as a basic instrument for the screening of alcohol consumption during pregnancy. Its use in the routine and practice of obstetric services is recommended in view of the tendency for increased alcohol consumption among women, the diffi culties to identify alcohol abuse by pregnant women, and the risk of developmental problems in children.