There are a growing number of studies attempting to diagnose the degree and extent of mental illness among the homeless. Increasingly, these studies are relying on structured diagnostic interviews such as the structured clinical interview for DSM III R diagnosis (SCIDR). This study examines the sensitivity of the SCID in diagnosing major mental illness among the homeless. Comparing SCID interviews with hospital chart diagnosis of twenty-three homeless individuals, the study found that whereas the positive predictive value of the SCID is high in that it accurately predicts a positive diagnosis, its negative predictive power is quite low. A negative SCID diagnosis does not accurately reflect a negative history of mental illness. The study indicates that single point interviews cannot be relied upon to accurately diagnose past history and, therefore, future need for treatment.