A study has been made on epilepsy and delivery using the Swedish Medical Birth Register, 1973-1981. A total of 635 women with 712 infants were identified with the diagnosis of epilepsy in the register-they represented 1/3 to 1/4 of the expected number. An analysis of the delivery outcome did not indicate a registration bias favouring poor delivery outcome. Hospital records were retrieved for 644 patients (again there is no indication for a selection bias favouring poor delivery outcome) and disease and drugs used were studied. Fifty-one women did not have epilepsy in early pregnancy but had their first attack during pregnancy, at delivery or in the puerperium. A total of 93 women with epilepsy had not used anticonvulsants in early pregnancy, 266 used drugs in monotherapy (10 different types of drug), 213 had 2 drugs, 65 had 3, 8 and 4, and 2 had 5 drugs. Differences in age, gravity, parity, marital status and smoking habits between the subgroups were studied. There was some statistically non-significant increase in perinatal death rate and malformation rate after polytherapy compared to monotherapy, but this may, at least partly, be a selection bias and not a drug effect.