We have prospectively studied the effects of three types of patient controlled anesthesia (PCA) in 30 patients who underwent elective upper abdominal surgery. Patients were allocated into three groups of 10. The first group received postoperative anesthesia by means of intravenous conventional PCA at fixed demand doses (PCADF). A second group was treated with demand doses that were varied in each bolus according to the intensity of pain (PCADV). A third group received continuous infusion with additional demand doses in variable bolus (PCAC + DV). The analgesic agent used in this study was morphine chloride. The total accumulated doses of morphine in PCADV group were lower than those of the other two groups. These differences were statistically significant when they were compared with PCADF group (p less than 0.01 at 2 and 6 hours; p less than 0.05 at 12 and 24 hours) and with PCAC + DV group (p less than 0.01 at 2, 6, 12, and 24 hours). Differences in morphine consumption during the different time intervals were markedly significant between 0 and 2 hours after the onset of PCA. PCADV required lower doses (p less than 0.01). After the first 2 hours, patients of PCADV group continued using lower morphine doses, but these differences were not significant when compared with PCADF group, although were present with respect to patients of PCAC + DV group (p less than 0.01). The incidence of secondary effects was not significantly different among the three groups. Based on these results we propose a modification in present PCA pumps.