The files of the Oxford Record Linkage Study were used to identify 223 infants delivered to 168 epileptic women as the result of 218 pregnancies. There were six stillbirths, two of which were grossly malformed. It was shown that the population of epileptic mothers differed significantly from the total reproducing population in respect of social class. Each pregnancy resulting in a livebirth was therefore matched exactly for social class, civil status, maternal age, parity, hospital, and year of delivery with three control deliveries resulting in livebirths. The defects noted at birth were abstracted from the Record Linkage files, and any subsequent hospital admissions or deaths of the children were also abstracted.There were highly significant excesses of congenital abnormalities among the infants born to epileptic mothers (13.8% of livebirths had some degree of defect of congenital origin compared with 5.6% of controls, P <0.0005). It was shown that neither the frequency with which the mother had fits nor the length of time she had had the epilepsy seemed to bear any relation to the frequency of defects in the offspring-with the exception of the two mothers who developed epilepsy in the first