OBJECTIVES: To compare use, effect, and cost of personalised computer education with community psychiatric nurse education for patients with schizophrenia. DESIGN: Randomised trial of three interventions. Modelling of costs of alternatives. PARTICIPANTS: 112 patients with schizophrenia in contact with community services; 67 completed the intervention. INTERVENTIONS: Three interventions of five educational sessions: (a) computer intervention combining information from patient's medical record with general information about schizophrenia; (b) sessions with a community psychiatric nurse; (c) "combination" (first and last sessions with nurse and remainder with computer). MAIN OUTCOME MEASURES: Patients' attendance, opinions, change in knowledge, and psychological state; costs of interventions and patients' use of NHS community services; modelling of costs for these three, and alternative, interventions. RESULTS: Rates of completion of intervention did not differ significantly (71% for combination intervention, 61% for computer only, 46% for nurse only). Computer sessions were shorter than sessions with nurse (14 minutes v 60 minutes). More patients given nurse based education thought the information relevant. Of 20 patients in combination group, 13 preferred the sessions with the nurse and seven preferred the computer. There were no significant differences between groups in psychological outcomes. Because of the need to transport patients to the computer for their sessions, there was no difference between interventions in costs, but computer sessions combined with other patient contacts would be substantially cheaper. CONCLUSIONS: The computer based patient education offered no advantage over sessions with a community psychiatric nurse. Investigation of computer use combined with other health service contacts would be worth while.