The effect of continuous epidural analgesia with bupivacaine 0.125% and sufentanil 0.83 micrograms.ml-1 on the plasma cortisol concentration and postoperative pain relief was compared with that of intermittent intravenous peroperative and on-demand intramuscular postoperative analgesia with nicomorphine 0.2 mg.kg-1. The study was performed on two groups of ten patients for three consecutive days after thoracic surgery. In the epidural group, a better quality of analgesia was found as measured with the Inverse Linear Analgesia Scale (ILAS) (1 = severe pain, 10 = no pain) than in the group that received intramuscular analgesia. The mean ILAS scores in the epidural group were: day 1, 8.1 (SD, 1.26); day 2, 8.3 (SD, 0.87); day 3, 8.6 (SD, 0.71). The corresponding mean ILAS scores in the systemic group were 7.3 (SD, 1.38), 7.4 (SD, 1.16), and 7.6 (SD, 1.15). There was a significant difference between the ILAS scores for the two groups on all days (p less than 0.01). The mean plasma cortisol concentration was not significantly different between the groups during the period of investigation. In the epidural group, the highest measured mean cortisol concentration of 520.2 (SD, 266.4) nmol.L-1 was on day 2, whereas that in the systemic group was 738.3 (SD, 255.1) nmol.L-1 on day 1. The measured mean plasma cortisol concentration was found to exceed normal limits (150-700 nmol.L-1) only in the systemic group and only on day 1.