Although the results are preliminary, most of the homeless mentally ill patients who agreed to participate in this study and were randomly assigned to the assertive community treatment team successfully moved off the street. However, this change did not imply stability; most patients had multiple moves from one type of shelter to another. It will be important to assess if the number of moves diminishes as more patients obtain shelter that can become permanent.While loss of shelter sometimes seemed motivated by patient preference, the majority of housing losses were a result of problems amenable to clinical treatment such as psychiatnic relapse and addiction. This finding suggests that homelessness can be reduced by appropriate clinical interventions if housing is available.That patients referred to the team from hospitals rather than from the street had more favorable outcomes is noteworthy. It is possible that hospitalization stabilized these patients and facilitated their engagement in treatment. It is also possible that patients referred from hospitals were more inclined to tolerate the structure and constraints of living in a shelter.With a few exceptions, clinical and demographic factors did not appear to have a sizable effect on housing outcomes. The effects of such factors should be examined over time. Overall, these data suggest that even with a relatively resource-rich clinical program, helping homeless persons with severe mental illnesses live in adequate shelter is difficult and requires clinical flexibility as well as multiple housing options.