One hundred and twenty patients presenting for admission were randomly allocated into two groups. Controls received standard hospital care and after-care. Projects were not admitted if this could be avoided; instead they and their relatives were provided with comprehensive community treatment and a 24-hour crisis service. Patients with a primary diagnosis of alcohol or drug dependence, organic brain disorder or mental retardation were excluded. During the 12 months study period, 96% of controls were admitted, 51% more than once. Of the projects, 60% were not admitted at all and only 8% were admitted more than once. Controls spent an average of 53.5 days in psychiatric hospitals; projects spent an average of 8.4 days. Community treatment did not increase the burden upon the community, was considered to be significantly more satisfactory and helpful by patients and their relatives, achieved a clinically superior outcome, and cost less than standard care and after-care.