NCJ Number
86033
Journal
Social Forces Volume: 61 Issue: 2 Dated: (December 1982) Pages: 475-483
Date Published
1982
Length
9 pages
Annotation
Although statistical predictions of violent behavior are superior to clinical prediction, statistical predictions based on socio-demographic, criminal, and mental hospitalization history variables are of little practical value in assessing future violent behavior.
Abstract
Equations to predict future violent behavior during hospitalization and in the community after release were developed for a study group of incompetent felony defendants based on retrospective analysis. While these analyses were underway, two comparison groups were designated. The study group was used to generate the prediction equation, while the two comparison groups were used to validate the accuracy of the questions. The study group included all males (n=257) indicted for a felony and found incompetent to stand trial in New York State from September 1, 1971, and August 31, 1972. The first comparison group included all incompetent males (n=282) charged with felonies during the same period but not indicted. The second comparison group was comprised of a stratified sample of male involuntary civil hospital admissions to six State mental hospitals in New York City during fiscal 1969 (n=250). The study produced three principal conclusions: (1) the variables considered do not strongly discriminate assaultive and nonassaultive cases for any of the three groups, regardless of the type of classification function used; (2) the use of either indicted-specific or comparison group-specific classification functions as a base for predicting future behavior leads to large error rates or misclassifications; and (3) community assaultiveness appears to be even more elusive to predict than hospital assaultiveness. It would appear that the reformulation of the problem for research in terms of a comprehensive assessment of individual characteristics in relation to social situational factors may be more promising for significant advances both for sociological understanding and more informed social policies. Tabular data are provided.