The literature was surveyed in an attempt to assess whether corticosteroids have a detectable teratogenic effect. Reporting bias was illustrated by the fact that the frequency of malformations in offspring of women treated with corticosteroids was much higher in reports of single cases than in reports of series of cases. In the latter group there were 468 exposed women, and the frequency of malformations was 3.5%, no different from the usual population frequency. There were, however, 2 cases of cleft palate, whereas 0.2 cases would be expected; it is impossible to tell whether this is a real increase. We conclude that the teratogenic potential of corticosteroids is so low as to be undetectable from the data available.