return to the MAIN pageGo BACK PRINT this page
The Official Website of Andrew Vachss

What are the Odds on Predicting Violent Behavior?

by Joel A. Dvoskin, Ph.D., A.B.P.P.

Predicting violent behavior among persons with mental illness has always been an important part of the mental health professions, but especially so since the 1960s, when commitment laws began to specifically require such prediction in order to involuntarily hospitalize people in psychiatric crisis. Unfortunately, these predictions were made with a confidence that turned out to be largely unfounded.

In the 1970s, researchers such as Henry Steadman began to systematically test psychiatric predictions of dangerousness. By the late seventies, these studies and others had led to a general belief that mental health professionals were no more accurate at predicting dangerousness than the general public. This dramatic turnabout from clinical overconfidence to clinical nihilism when it came to predicting violent behavior had an unfortunate side-effect: a long and disappointing dearth of research interest in the topic. "We can predict anything" had sadly become "we can predict nothing."

Yet practitioners continued to make predictions. Indeed they were forced to by laws which continued to require dangerousness (or predictable violence) for a variety of legal purposes. At the same time, it was intuitively clear to many practitioners that we could indeed accurately predict a great many things, such as response to various treatment modalities or medications. As is often the case when science and intuition conflict, a second generation of prediction researchers paid closer attention to what the earlier researchers had really said. Our failures had been around the "long-term prediction" of violent behavior. What about short-term predictions?

During the second half of the eighties, researchers such as Binder and McNiel, Mulvey, Klassen, and others began to look toward violence predictions in a whole new light. "We can predict nothing" became a question: "What can we predict, when, and for whom?" They began to look at the interactions between certain types of people with mental illness and the situations in which they were likely to find themselves. In short, it is clear that mental illness by itself does not necessarily increase the probability of violent behavior. But for some people in some circumstances, it may. While most people with mental illness are not violent most of the time, it is also true that some people with mental illness will indeed behave violently some of the time. The challenge for practitioners and researchers alike is to try to identify the treatments and life situations which raise or lower the odds for individuals. Judgments about groups of people can only lead to stigma and discrimination, while judgments about individuals, if based on reason and information, can lead to better treatment outcomes and increased safety for the individuals and their communities.

Those of us who work in forensic mental health systems are repeatedly faced with the need to recommend continued hospitalization or release to the community for people who may have committed serious acts of violence in the past. According to actuarial data, these historical acts indeed raise the likelihood that the person will again behave violently. Yet our inability to make accurate long-term predictions has left us with a difficult dilemma. Suppose that each of ten people had a one-in-ten chance of committing a violent crime if released. We would be forced to choose between retaining all ten, thus unnecessarily taking liberty away from nine people, or letting them all go and thus guaranteeing an injury to an innocent member of the public. While such odds are of course impossible to ascertain, the example is nevertheless instructive. We are often torn between the wish to maximize the liberty of each citizen and the need to protect the public—truly a rock and a hard place. How have we managed?

The answer, for me at least, has been to reject the dilemma. I do not believe that the odds are stagnant. For example, there are mentally ill people who have only behaved violently when drunk. For these people, release without alcohol treatment raises their own risk of violence. Others have never behaved violently when taking their prescribed medication. Obviously, their attitude and knowledge about medication will have an effect on their odds of success. The absolute odds are impossible to determine, but the relative odds are not. Our challenge is to continue to develop responsive community supports and treatments so that each person who is returned to freedom after an involuntary hospitalization has the highest possible likelihood of a successful, non-violent life.

The other principle which governs our management of the risk of violence among our patients is to provide small increments of increased freedom and reduced control. This graduated path to freedom not only allows the patient to demonstrate trustworthiness and an ability to adapt to new situations, but allows a sometimes retributive public to adapt to them as well.

In a number of studies, including our own recent study in New York, persons found "not guilty by reason of insanity" have shown recidivism rates for violent crime which were dramatically lower than those of convicted felons. Generally these patients receive treatment and supervision in the community following release, while other parolees are often placed on ridiculously large caseloads where help is virtually impossible to find. Thus, a societal response of support and treatment seems to yield better results than one of punishment and scrutiny. Yet there remain a number of mentally ill citizens who, because of their criminal-justice status, may be receiving less support and treatment than they need to maximize their chances of success. An investment in these mentally ill probationers, parolees, detainees and inmates would seem to appeal not only to our generous hearts, but to our selfish minds as well.


  1. Trust your instincts. They are born of years of valid experience. If you feel afraid, you probably have a good reason.

  2. Be honest about your own feelings, but don't judge those of your mentally ill family member.

  3. Violence is often the product of anger and fear. Try not to make it worse.

  4. It is okay to ask your family member if you are in danger, and what they would like for you to do.

  5. If you become frightened during an argument, back off. It's not a good time to negotiate, anyway.

  6. Persons with mental illness are often more afraid of their violent feelings than we are. Pretending that everything is fine when it clearly isn't will only further confuse and frighten both of you.

  7. Don't feel compelled to either confirm or challenge delusions or hallucinations. If pressed try: "I believe these things are real to you." Be kind but honest. Lying is disrespectful and humiliating.

  8. What has worked for you in the past? Things that defused past situations are likely to work again.

  9. Know your local support systems. If your community has a mobile crisis team or a supportive police force, know how to contact them and what to expect.

And, for heaven's sake, don't be afraid to ask for help.

This copyrighted article is reprinted by special permission from the publisher, The JOURNAL of the California AMI. The entire issue from which the article "What are the Odds on Predicting Violent Behavior?" was taken is available by sending a check for $10 to The JOURNAL of California AMI, 1111 Howe Avenue Suite #475, Sacramento, CA 95825.

For more information about Dr. Joel A. Dvoskin, or to read more articles by this leading leading forensic psychologist, click here.


Search The Zero || Site Map || Technical Help || Linkage || Contact The Zero || Main Page

The Zero © 1996 - Andrew Vachss. All rights reserved.

How to Cite Articles and Other Material from The Zero
The URL for this page is: