Continuous interscalene brachial plexus block with a single dose of 0.5% bupivacaine 1.25 mg/kg, continued with an infusion of 0.25% bupivacaine 0.25 mg/kg/h, was performed on 24 patients to provide analgesia during shoulder surgery and in the postoperative period. The drugs for general anaesthesia included glycopyrrolate, thiopentone, vecuronium, enflurane and N2O/O2. All patients had signs of regional analgesia 30 min after the block without haemodynamic problems. The infusion of local anaesthetic was interrupted in six patients because of a failure in catheter function. Of the remaining 18 patients, nine needed no complementary analgesics and nine patients received, on average, 1.6 doses of oxycodone (0.15 mg/kg/dose) during a 24-h period. Displacement of the interscalene catheters could be prevented by a fixation suture to the skin. Two patients noted a metallic taste during the bupivacaine infusion. The most common complaints were numbness of the hand (n = 15) and hoarseness (n = 5). The mean (+/- s.e.mean) plasma concentrations of bupivacaine at 30, 60, 180 min and 24 h were 0.68 +/- 0.06, 0.62 +/- 0.05, 0.52 +/- 0.04 and 0.76 +/- 0.01 micrograms/ml, respectively. During the 24-h period, the alpha 1-acid glycoprotein (AAG) concentration (mean +/- s.e.mean) in plasma rose from 0.41 +/- 0.04 g/l to 0.54 +/- 0.04 g/l (P less than 0.001). The concentration of free bupivacaine was below detectable levels (less than 0.01 micrograms/ml) after the 24-h infusion. The rise in AAG probably increases binding of bupivacaine to plasma proteins, diminishing the risk of systemic toxicity.