All admissions to the Singer Mental Health Center, Rockford, Ill, with a diagnosis of alcoholism in a one-year period (N = 466) were randomly assigned to one of two inpatient programs. One program, "intensive incare," had a high staff-patient ratio with the operating assumption that intensive staff-patient interaction is significant in patient outcome. The other, "peer-oriented incare," was of low staff density with the assumption that patient-patient interaction is critical in rehabilitation. In addition, the patients resided in communities with outpatient services classified as either "network" (an organized set of community services) or "no-network" (no special funding or deliberate outreach effort). Thus, each patient could be treated in one of four systems of care. Data on treatment outcomes were collected via semistructured interviews at 3, 6, 12, and 18 months after admission. The "peer-oriented incare" system showed superiority to the "intensive incare" treatment approach in improved drinking behavior. There were no other significant differences among the four systems on the outcome criteria for alcoholic patients. Along with other recent studies, these findings have implications for policy planning, particularly with today's emphasis on cost effectiveness.