In 32 patients subjected to total hip replacement, postoperative pain relief was achieved by random treatment with either 5 mg of morphine in 10 ml of saline (n = 15) or 6-8 ml of 0.5% bupivacaine with epinephrine (n = 17), both drugs administered by the lumbar epidural route. In an additional group of 10 patients, post-traumatic thoracic or post-operative abdominal pain was relieved first by 4-6 ml of 0.5% bupivacaine with epinephrine and subsequently by 5 mg of morphine in 10 ml of saline, both drugs being administered by the thoracic epidural route. The duration of analgesia was significantly longer, on average, with morphine (28 h) than with bupivacaine (4.3 h) when the drugs were given by the lumbar route. Thoracic administration of morphine also resulted in a significantly longer duration of pain relief (on average 9.8 h) than that of bupivacaine (3.8 h). Morphine gave satisfactory pain relief in all cases. It was not associated with motor block, loss of sensitivity to temperature, touch, or pin-prick, or any signs of sympathetic block, as was the case with epidural bupivacaine. Plasma concentrations of morphine were not detectable 8 h after injection, though the patients still had pain relief. One case of delayed severe respiratory depression occurred 6 h after morphine injection via the thoracic route. Epidural morphine analgesia should therefore be reserved for patients in whom continual surveillance is possible, at least until more is known about the pharmacokinetics of narcotics in the epidural and subarachnoid space.