Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

William F. Waters


Federal court decisions and state and federal legislation have made the legal concept of dangerousness the predominant criterion in involuntary mental health hospital commitments and in determining the least restrictive custody level in prison inmate custody assignments. However, no adequate consensual definition with empirical methods of measurement exists. Megargee's (1976) definition of dangerousness as the probability of violent behavior was used as the basis for an empirical approach to measuring dangerousness in this study. Recent studies in the literature (Monohan, 1984; Klassen & O'Conner, 1988) suggested that discriminant analysis might be an appropriate methodology for model development. Review of the literature suggested a variety of potential predictors to use in a discriminant model. Criterion groups of 50 violent psychiatric hospital patients and 50 violent prison inmates were identified, with violence defined as assault or overt threats of assault on another person within six months of admission. Comparison groups of 50 nonviolent patients and 50 nonviolent inmates were drawn from subjects hospitalized or imprisoned during the same years as the criterion subjects. Four validation samples of 25 subjects each were also drawn. Stepwise discriminant analysis of the combined hospital and prison derivation samples yielded a discriminant function containing eight variables which was 75% accurate in classifying the original sample and 73% accurate in classifying the holdout validation sample. The analysis of the hospital sample resulted in a discriminant function containing five variables which was 85% accurate in classifying the original sample and 74% accurate on the validation sample. The analysis of the prison sample resulted in a discriminant model with six variables which was 72% accurate in classifying the original sample and 78% accurate on the validation sample. It was concluded that the derived population-specific (i.e. hospital or prison) models did constitute legally defensible, empirically valid measures of dangerousness for the populations studied. It was also concluded that stepwise discriminant analysis is an appropriate methodology for deriving risk assessment models of dangerousness. Current plans for replication were discussed.