Etiologic studies of birth defects often use family history information provided by parents of patients. The validity of this information has not been adequately assessed. Using data from the Atlanta Birth Defects Case-Control study, we evaluated sensitivity, specificity, and positive predictive value of mothers' responses regarding the presence of birth defects in their offspring. A total of 4929 mothers of infants with major structural defects ascertained by the Metropolitan Atlanta Congenital Defects Program and a total of 3,029 mothers of normal infants were asked whether their babies had had a birth defect or a health problem diagnosed during the first year of life. Interviewers and coders of maternal responses were blinded to the case-control status of infants. Sensitivity (the proportion of case mothers who gave responses that could be coded as denoting a major birth defect) was 61%. Specificity (the proportion of control mothers who gave responses that could not be coded as denoting a major birth defect) was 98%. The positive predictive value (the proportion of mothers who gave a major-birth-defect response who in fact had babies with major birth defects) was estimated as 47%. Sensitivity, specificity, and positive predictive value varied by maternal sociodemographic factors such as race and education, as well as by type of defect. These results suggest that family history data obtained through maternal interviews should be cautiously interpreted and, if not properly validated, may alter estimates of recurrence risks.