Most of the research on crime and mental disorder has been limited to studies of the prediction of “dangerousness.” This is not surprising because mental health professionals are regularly required by law to assess the risk that a patient presents of behaving violently or of committing other illegal acts. Inaccurate predictions can lead to horrific human tragedies or unwarranted confinement for the patient. Accurate predictions, on the other hand, can lead to placing patients in community treatment programs that effectively prevent crime and violence. To date, the accuracy of these predictions has not been good (Borum, 1996). In order to increase the accuracy of these predictions, it is essential that they be based on results of empirical research. However, on what type of research should these procedures for clinical prediction be based? Presently, the literature on the prediction of “dangerousness” among the mentally disordered consists of studies of variables, easily ratable in a clinical situation, that were entered into a linear regression model and found to account for a “significant proportion of the variance.” These types of studies are inadequate, both conceptually and methodologically, and consequently their results may be misleading. In order to improve the accuracy of these clinical predictions of dangerousness, their scientific basis must be made sound. This requires abandoning variable-oriented investigations that presume the same predictors apply in the same way to all patients (Brown, Harris, &
Lemyre, 1991) and developing prediction procedures based on a more general understanding of the development of individuals who as adults suffer from major mental disorders and commit crimes.