Twenty-eight patients with severe pain due to cancer, who could no longer obtain acceptable pain relief from optimised doses of oral opioids, were entered into a study which compared pain relief, satisfaction with pain therapy and estimates of neuropsychological functioning during treatment with spinally administered (i.e., epidural and intrathecal) morphine as either repeated bolus doses or as a continuous infusion. These measures of efficacy and side effects were repeated every 2 weeks until either the patient died (82% of patients), withdrew from the study or were no longer able to complete the tests due to deterioration of their condition. The mean (range) duration of treatment was 169 (6-537) days for those patients receiving continuous infusion and 140 (28-378) days for those patients receiving repeated bolus doses. There was no significant difference in visual analogue pain scores, pain relief scores and satisfaction scores between the bolus and infusion groups. Furthermore, low pain scores and high pain relief scores indicated that both treatment modalities provided effective pain control. Similarly, there was no significant difference between the two groups in the various tests used to assess depression or neuropsychological function (i.e., memory, vigilance, attention and processing). There was a significantly greater degree of dose escalation in patients receiving continuous infusion compared to patients receiving repeated bolus doses. For 6 patients in the infusion group the catheter was sited in the intrathecal space, as the dose requirements by the epidural route exceeded the delivery capacity of the pump. For 4 patients in the bolus group the catheter was similarly sited, due to pain on injection and leakage/blockage.(ABSTRACT TRUNCATED AT 250 WORDS)