101 critically ill patients admitted to five intensive-care units were allocated randomly to receive a continuous intravenous infusion of either propofol or midazolam for sedation for up to 24 h. In addition, morphine was given to provide analgesia. The mean duration of infusion was 20.2 h (range 3.0-24.5) in the propofol group and 21.3 h (4.0-47.0) in the midazolam group and infusion rates were 1.77 mg/kg/h (range 0.40-5.00) and 0.10 mg/kg/h (0.01-0.26), respectively. The infusion rates were adjusted as necessary, and the desired level of sedation was achieved easily in most patients in both groups. There were slight falls in arterial pressure, but there were no significant differences between the groups. Heart rate was lower in patients who received propofol. Some small changes occurred in biochemical and haematological variables in both groups, but they were not clinically significant. There was no indication that either drug substantially impaired adrenal steroidogenesis. When the infusion was discontinued, there was less variability in recovery of consciousness in patients who had received propofol. In a subgroup of patients, weaning from mechanical ventilation was achieved significantly faster after discontinuation of propofol than of midazolam. Propofol proved to be a satisfactory agent for sedation of these critically ill patients and compared favourably with midazolam.