The outcome of 807 gamete intra-Fallopian transfer (GIFT) cycles following the elective or non-elective transfer of two, three or four oocytes has been retrospectively studied. Electively replacing either three or two oocytes did not reduce the clinical or ongoing pregnancy rate when compared with replacing four oocytes. The incidence of high-order multiple gestation (triplet or more) was significantly reduced by replacing fewer oocytes, but the occurrence of twin pregnancy was not altered. Lower pregnancy rates were found when the number of oocytes available for replacement was limited and non-elective replacement was performed. It is suggested, therefore, that a higher number of oocytes available may allow selection of higher quality oocytes for transfer. We conclude that the overall expectation of pregnancy from the GIFT procedure is high (30-40%) and the number of oocytes replaced should be two in order to minimize the risk of high-order multiple pregnancies. It is not clear whether increasing the number of oocytes transferred will benefit subjects who failed to become pregnant previously with GIFT, but limited data suggest that transferring large numbers of oocytes to women > 40 years does not improve the expectation of pregnancy.