Eighty patients scheduled to undergo knee arthroscopy were studied in random and double blind fashion. Spinal anaesthesia with hyperbaric 0.5% bupivacaine was selected for 40 overnight-in-patients. At the end of arthroscopy, 1 mg morphine or saline was injected intraarticularly. Local anaesthesia with 1% lidocaine plus adrenaline, was selected for another 40 out-patients. At the end of the arthroscopy either 1 mg morphine or saline was injected intraarticularly. As a rescue medication the spinal anaesthesia patients received oxycodone 0.14 mg kg-1 i.m. or ketoprofen 100 mg p.o. and the local anaesthesia patients received ketoprofen 100 mg p.o. The need for additional postoperative analgesic was almost similar in both spinal anaesthesia groups. The patients having local anaesthesia and given intraarticular morphine needed fewer doses of ketoprofen (22 doses) postoperatively than the control group (39 doses) (P < 0.05). Duration of analgesia was slightly longer after morphine than in the control group (ns). There was no difference between the morphine patients and the control patients in the two studies regarding the incidence of side effects. We conclude that postoperative analgesia in patients undergoing knee arthroscopy under local anaesthesia, but not under bupivacaine spinal anaesthesia, can be improved with a single intraarticular injection of 1 mg morphine.