In a prospective study, patients undergoing cholecystectomy were randomly allocated to receive (a) intermittent intramuscular morphine (n = 25), (b) continuous intravenous morphine infusion (n = 25) or (c) epidural bupivacaine (n = 25) for postoperative pain relief. Morphine by intravenous infusion provided comparable pain relief to intermittent intramuscular morphine; there was no significant difference in the incidence of postoperative pulmonary complications. Patients receiving epidural bupivacaine for 12 h had better analgesia than patients receiving morphine (P less than 0.001). Arterial oxygen tensions were also significantly higher in the epidural group for the first three postoperative days (P less than 0.05). Epidural analgesia was associated with a significant reduction in the incidence of pulmonary complications (P less than 0.01) and chest infection (P less than 0.05).