Psychology, Department of


Document Type


Date of this Version



Published in American Psychologist, 51:9 (September 1996), pp. 928-930; doi: 10.1037/0003-066X.51.9.928 Copyright © 1996 American Psychological Association. Used by permission. “This article may not exactly replicate the final version published in the APA journal. It is not the copy of record.”


Recent developments in the law have made the assessment of risk of violence a required professional ability for every clinical psychologist. About 30 years ago, laws controlling involuntary civil commitment evolved to require more than merely a finding of mental illness. They also required evidence that patients, if not committed, would be dangerous to themselves or to others. During that era, states also developed laws that made it mandatory for clinicians to report evidence if their child clients, the children of their adult clients, and disabled or older adults were in danger of abuse by their caretakers.

Clinicians’ obligations to assess risk of violence were driven home perhaps most dramatically by the infamous “Tarasoff case” (Tarasoff v. Regents of the University of California, 1976). Together with subsequent cases across the states, Tarasoff required that clinicians take measures to protect third parties from their clients’ potential for violence. This implied that clinicians should take reasonable steps to assess and be aware of their clients’ potential for violence. Indeed, by 1978, Shah was able to describe no less than 15 legal and clinical contexts in which mental health professionals were expected to assess the risk of violence and would be potentially liable for failing to do so.

Spurred by these new laws, many researchers in the 1970s began to explore mental health professionals’ abilities to assess violence risk. What they found was in stark contrast to society’s apparent faith in clinicians’ assessment skills. Summarizing those early studies, Monahan (1981) concluded that when clinicians predicted that a person would be violent, available research indicated that they were accurate no more than one in three times.

Three things could have happened in the ensuing decade of the 1980s as a result of this discouraging news: Courts could have discontinued their reliance on clinicians’ judgments about patients’ potential for violence; clinicians could have heeded the news by avoiding roles requiring violence predictions; or researchers could have given up on the empirical questions of violence predictions. None of these things happened.