In a retrospective study of 813 oocyte retrieval-embryo transfer cycles in women with normal follicle stimulating hormone and luteinizing hormone concentrations, we sought to investigate the relationship between the amount of human menopausal gonadotrophin (HMG) used for ovarian stimulation and treatment outcome. Patients were divided into three groups: group A patients (495 cycles) required < 40 ampoules of HMG and had a predicted probability for pregnancy of 25% per embryo transfer; group B patients (165 cycles) required 41-77 ampoules per cycle, with a predicted probability rate for pregnancy of 5-25% per embryo transfer; and group C patients (153 cycles) required > 77 ampoules of HMG and the predicted probability for pregnancy was < 5% per embryo transfer. Groups C and A differed significantly (P < 0.005). The mean oestradiol concentration on the day of HCG administration in group C was 6412 pmol/l, and the mean number of eggs retrieved was seven. The highest success rates were found when up to 2.5 ampoules of HMG were required for each egg or 4.4 ampoules for each embryo. The lowest rates were obtained when > 4.8 ampoules of HMG were necessary for each oocyte or > 9.6 ampoules for each embryo (P < 0.005). We identified a group of infertile patients who required excessive amounts of HMG to achieve a fair degree of steroidogenesis, number of eggs and number of embryos but who had very low pregnancy rates. Although all other relevant parameters were normal, this may highlight the beginning of ovarian-gamete insufficiency before the basic hormonal status is affected. In cases of repeated failure, oocyte donation should be considered.