Eighty-three infants received i.v. morphine following surgery as a continuous infusion to a targeted morphine concentration of 20 ng ml(-1) (n = 56) or as intermittent bolus doses as needed (n = 27). Ventilation was compared in the two groups by continuous pulse oximetry, by venous blood gases on postoperative day 1 (POD 1) and by CO2 response curves. Infant pain scores were done to assess analgesia every 4 h. Both groups achieved pain scores consistent with analgesia but the bolus group showed a higher percentage of pain scores indicating distress (32 vs. 13%, P < 0.001). Room air saturations of < 90% were seen for 2.3% of POD1 in infusion-treated infants and for 2.5% of POD1 in bolus-treated infants. Mean venous PCO2S were normal in the two groups. Four infants showed ventilatory effects in the infusion group (4/ 56 = 7%); venous hypercarbia in two (2 days, 36 days), oximetry desaturation in one (240 days), both effects in one (6 days). Ventilatory effects were not statistically different between the intermittent bolus-treated and infusion-treated infants but may be clinically important. Monitoring with continuous oximetry is necessary. Morphine clearance increased with age. Infants with detectable morphine also had measurable morphine-6-glucuronide in both groups. Oral intake began at 16 h in both groups and other side effects were infrequent.