The outcome of patients receiving rehabilitation on a stroke unit (SU) was compared with that of patients on general medical and health care of the elderly wards (conventional wards; CWs) in a randomised control trial. Of the 315 stroke patients admitted to hospital, 176 were randomly allocated to the SU and 139 to CWs. Outcome was assessed on the Barthel Index, Rivermead Activities of Daily Living Scale (ADL), Nottingham Extended ADL scale, General Health Questionnaire, Mood Ratings and Adjustment scales at 3, 6, and 12 months after stroke by an assessor who was ''blind'' to their group allocation. SU patients spent significantly longer in hospital. There was a significant difference in personal ADL at 3 and 6 but not 12 months after randomisation. On measures of extended ADL, SU patients were significantly more independent at 6 and 12 months but not at 3 months after randomisation. There were no significant differences between the groups in mood at 3 and 6 months, but SU patients showed better mood 12 months after randomisation. There was a significant difference in adjustment 6 months after randomisation. Patients who received rehabilitation on SUs were more independent in activities of daily living compared with those who received rehabilitation on CWs. By 6 months they were also psychologically more able to cope.