The variability of ovarian response in cattle after superovulatory treatment with PMSG is mainly due to differences in the follicular population on the day of PMSG injection. We have tried to modify this follicular population either by destroying some of the follicles at the time of PMSG injection or by pretreatment at the beginning of the cycle. The electrocauterization of follicles on day 16 delayed luteolysis and did not reduce variability. The injection of 1 000 i.u. of PMSG on day 3 post-oestrus produced 35.7% of twin ovulations when 600 i.u. were injected on day 16. Another explanation of response variability is the duration of PMSG action which can be modified by controlling the PMSG-oestrus interval with prostaglandin analogs; lengthening that interval increased the ovulation rates from 1.75 to 4.60 and decreased the number of large follicles growing after ovulation from 5.13 to 0.60. Moreover, when PMSG prostaglandin treatment was associated with progestagen, PMSG efficiency increased. So after 500 i.u. of PMSG at the end of a progestagen-prostaglandin treatment we obtained a 34.1% twinning rate. Finally, repeated treatments (2 000 i.u. PMSG-0.5 mg PGF2α analog) every 6 weeks during one year produced the same variability in the response pattern: response was constant in 29.8% of the heifers, 27.7% showed a definite decrease after 3 to 5 treatments, and 36.2% a decrease followed by an increase.