This research developed a model for forensic release decisions that in-
INTRODUCTIONPrediction of successful outcome in the insanity acquittee population is crucial in facilitating patient release decisions. During the 1970s, there was heightened interest in the prediction of violent behavior due to the trend away from civil commitment based on a need for treatment and toward civil commitment based upon a standard of dangerousness to self or others (Monahan, 1984). By the end of the decade, however, a number of studies had suggested that mental health professionals possessed poor predictive abilities with regard to future violent acts (e.g., Cocozza & Steadman, 1976; Pfohl, 1978).One major problem in prediction literature has been the overprediction of violence, that is, high false positive prediction rates (Wenk & Emrich, 1982;Wenk, Robinson, & Smith, 1972). One reason for this, suggested by Monahan (1978) is that violence has a low rate of occurrence and can easily be overpredicted. In addition, violence can be unreliable as an event, and there is little consensus on the definition of violence or reliability in verifying its occurrence.A second problem in prediction research is to define what constitutes a successful outcome. Recidivism is frequently used as the sole criteria of outcome. However, its use has been criticized as overlooking the value of programs whose goals may be other than to alleviate an individual's proclivity toward criminality (Maltz, 1984;Gottfredson & Gottfredson, 1980). Further, there is little agreement among researchers as to a definition of recidivism. Maltz (1984) has identified nine different definitions of recidivism but advocates using rearrest rates as the most accurate, though limited, definition. In this research, we have chosen rearrest as our definition of recidivism as well.A third major problem in prediction literature is the lack of agreement about which independent variables are significant in a prediction model. Some studies have used standardized psychiatric interviews, mental status findings, and other clinical parameters (Strauss & Carpenter, 1977). However, attempts at predicting adjustment on parole have relied largely on actuarial tables and have been somewhat more successful (Gottfredson, Wilkins, & Hoffman, 1978;Hoffman & DeGostin, 1974;Glaser, 1962). The Salient Factor Scale, used by the U.S. Parole Board (Gottfredson, Hoffman, Sigler, & Wilkins, 1975) consists of nine items: prior convictions, prior incarcerations, age at first conviction, offense involved auto theft, history of drug abuse, employment prior to arrest, high school graduate, prior parole revocation, and release plan to live with spouse and/or children. While nearly all of the items in the Salient Factor Scale are based on an individual's actions prior to incarceration, Monahan (1978) has suggested the importance of additional postinstitutional factors, such as with whom the subject would be working and living, family support, and the type of environment to which an individual is released.Though...