Patients undergoing thoracotomy experience severe postoperative pain and marked respiratory impairment. Analgesics(narcotics or loeal anesthetics) administered via epidural catheter in epidural space have been shown to provide postoperative analgesia and improve respiratory mechanics after thoracotomy. Several different methods have been utilized in an attempt to reduce pain and pulmonary mechanics after thoracotomy. These include epidural blocks using local anesthetics, epidural narcotics, ketamine, steroid, and clonidine. These methods have been shown to provide pain relief with relative preservation of lung volumes in the postoperative period, but have disadvantages. Especially epidural local anesthetics may cause hypotension and motor blockade of lower extremities, and epidural narcotics may cause pruritus, nausea and vomiting, urinary retension and respiratory depression. In an attempt to provide excellent analgesia and improve pulmonary mechanics after thoracotomy and to decrease the side effects associated with the intermittent bolus administration of epidural narcotics or local anesthetics, we performed a study of continuously administered epidural infusion of small concentration of fentanyl combined with low concentration of bupivacaine. Twenty eight patients undergoing thoracotomy were randomized into groups based upon a postoperative pain regimen as indicated: Group I: intermittent intramuscualr injection of nalbuphine 0.2 mg/kg(n=13), Group II: continuous epidural injection of mixtures of 0.2/ bupivacaine and fentanyl 3 ug/ml(n= 15). Two, 8, 24 and 48 hours postoperative, the following indices were measured: visual analogue pain scale, vital capacity, tidal volume, arterial blood gas analysis(pH, PaCo2, PaO2), side effects, and 24 hour urine 17-ketosteroids. The results were as follows: 1) Pain score was evaluated by visual analogue pain scale postoperatively and the pain scores significantly decresed in group II as compaired with those in group L 2) Vital capacity and tidal volume in group II were more improved than group I. 3) There was no difference in arterial blood gas analysis except for decreased PaO2 at 2 hour and 24 hour compared with preoperative value in group L 4) Major complications in group II were two cases of nausea and vomiting, one case of urinary retension, whereas only I patient in group I complained of nausea and vomiting. 5) No significant difference occurred in 24 hour urine l7-ketosteroid at 24 hour and 48 hour postoperatively in group I and group II, which were within normal limits.