A group of 105 women (54 of whom were, and had only ever been, taking valproate for at least a year, and 51 who had only ever taken either lamotrigine or carbamazepine, for at least a year) were compared with a group of 50 women who did not have epilepsy: any oral contraceptive taken at the time of testing was recorded and blood levels of follicle stimulating hormone (FSH), luteinising hormone (LH), testosterone and prolactin were estimated from days 2 to 6 of the menstrual cycle (day 1 being the first day of bleeding) and an MRI scan made of their pelvis. Women with epilepsy in general were significantly more likely to exhibit evidence on MRI scanning, of polycystic ovaries (PCO): women taking valproate but not an oral contraceptive were significantly more likely to have clinical biochemical evidence of the polycystic ovarian syndrome (PCOS) with raised LH and/or testosterone levels between days 2 and 6 of their menstrual cycle than women who did not have epilepsy: this was not so for women taking lamotrigine or carbamazepine. Since the polycystic ovary syndrome has potentially serious consequences it is suggested that, where possible, valproate is avoided in women of child bearing potential.