Over the past two decades, there have been numerous attempts to develop clinical instruments and techniques for the prediction of dangerous behaviour (e.g., Carroll & Fuller, 1971; Hellman & Blackman, 1966; Holcomb & Adams, 1985; Lothstein & Jones, 1978; Megargee, Cook, & Mendelsohn, 1967; Menzies, Webster, & Sepejak, 1985; Syverson & Romney, 1985). The search for valid standardized scales, test batteries, diagnostic categories, and other clinical predictors continues despite meagre returns and growing doubt that the ambition of refining a predictive technology is at all worthwhile (Meehl, 1986; Megargee, 1970; Monahan, 1984). Unfortunately, much of the existing clinical research literature, dedicated as it is to the development of a technology of prediction, underestimates the complexities of actual clinical procedures. Clinicians, already handicapped by imprecise tools, find themselves further encumbered by the lack of substantive discussion about how, given the limits of their abilities, they can responsibly conduct assessments of dangerousness. The search for accurate predictors of dangerousness has its roots in the historical interdependence of psychiatry and the law. According to Foucault (1978), by the latter part of the 19th century, it was apparent that the penal system was unable to achieve the ideal of reforming the criminal. The emphasis of the criminal justice system began to shift from a focus on legal responsibility and rehabilitation, to the issue of protection for society. With this shift in focus, the demands on mental health professionals in the judicial system also began to change. Whereas previously, clinicians were called upon primarily for explanations of criminal behaviour and recommendations for treatment, now they were being asked to offer predictions as well. By the beginning of the 20th century the concept of the “dangerous being” (Petrunik, 1983) had become an established concept in criminal law and psychiatry. It attained this status not on scientific grounds but because it served legal imperatives. Clinicians in the justice system had assumed a role shaped more by judicial priorities than by a realistic appraisal of what mental health professionals could reasonably offer the legal process. As a result, the function of clinicians assessing dangerousness had grown increasingly narrow and in many quarters had come to be regarded as having a single purpose-accurate prediction. As Halleck (1987) has aptly remarked, clinicians