During the 8-year period from 1976 to 1984, 202 patients with a primary ovarian cancer underwent a second-look laparotomy at the Institut Gustave-Roussy (Villejuif). One hundred and nine patients had a macroscopic tumour, in 77 of which there was clinical evidence of disease before the laparotomy. Fifty-seven patients underwent an optimal resection of the tumour (largest residual tumour less than 2 cm) and 52 underwent non-optimal cytoreductive surgery or isolated biopsies. In 22 cases the optimal resection necessitated a bowel resection. Survival curves suggest: (1) that the removal of macroscopic residual disease does not improve life expectancy except in the cases of optimal resection without bowel resection. (2) When there is evidence of disease before the second-look operation the prognosis remains the same whatever the surgery performed.