Two trained, nonclinical raters used the psychometric scale to assess the dangerousness of 210 patients interviewed in a pretrial forensic clinic. Reliability, item, and factor analyses refined the instrument into a 15-item Dangerousness Behavior Rating Scheme. Profiles of patients during a 2-year followup were constructed from contacts with the correctional system and 5 psychiatric hospitals. Profiles also were rated on dangerousness to others by 9 independent judges. A Pearson product-moment correlation of .34 was yielded between aggregate factor scores derived from the instrument and followup dangerous behavior. Results also indicate that the accuracy of dangerousness predictions is increased and the rates of false positives and negatives are curtailed when assessments are offered only regarding patients falling at the extremes of the continual for perceived dangerousness. Given the limited ability of either clinicians or psychometric instruments to accurately predict dangerous behavior and the socially constructed nature of the phenomenon, predictions about future violence should be adopted into practice only under restricted conditions. Some of these conditions are delineated. Tabular data and 59 references are provided. (Author abstract modified)