We studied the effects of the timing of tourniquet release in 88 patients randomly allocated for release after wound closure and bandaging (group A), or before the quadriceps layer had been closed allowing control of bleeding before suture (group B). The groups were similar in mean age, weight, gender, preoperative knee score, radiographic grading, and prosthesis implanted. Patients in group B had less postoperative pain, achieved earlier straight-leg raising, and had fewer wound complications. Five patients in group A had to return to theatre, three for manipulation under anaesthesia, one for secondary closure of wound dehiscence, and one for drainage of a haematoma. The last patient later developed a deep infection, which was treated by a two-stage revision. There were no significant differences between the two groups in operating time, or the decrease in haemoglobin concentration at 48 hours postoperatively. Some of the adverse effects of the use of a tourniquet for knee surgery can be significantly reduced by early tourniquet release, with haemostasis before the quadriceps mechanism and the wound are closed.