Cohort studies of putative human teratogens can identify the full spectrum of phenotypic effects, including both major malformations and minor anomalies. Cohort studies which include the much more common minor anomalies make it possible to use a relatively small number of exposed and unexposed infants to identify an increase in the frequency of malformations. We evaluated this use of minor anomalies in a cohort study of newborn infants who had been exposed in utero to three putative teratogens: insulin-dependent diabetes mellitus in the mother and the use of the anticonvulsant phenytoin and exogenous sex hormones by the mother. In addition, the reproducibility of identifying minor anomalies was tested by comparing the results of examinations by two independent observers of 444 unexposed infants. The frequency of minor anomalies was increased among infants of diabetic mothers. However, the reproducibility of identifying minor anomalies was poor. We conclude that the examination of teratogen-exposed infants for minor anomalies cannot be used in epidemiologic studies of putative teratogens unless special efforts are made to maximize consistency in the identification of these features.