Predicting violent behavior: not guesswork, but far from certain

<i>This post has been corrected. See note at the bottom for details.</i>

After a person has committed horrific acts of violence, the signs often seem plain to see: He was always a loner, say fellow students; he may have been bullied, former teachers speculate; he was hit hard by a disintegrating family situation, neighbors suggest.

But predicting whether an individual is on a path toward violent behavior is anything but simple, say experts. Even for highly trained mental health professionals, assessing an individual’s risk of becoming violent is an exercise in complexity, in probability -- and in humility.


Some advances have been made since 1983, when the Supreme Court said the mental health profession was “not always wrong” in making such calls -- “only most of the time.” Forensic psychiatrists and psychologists have begun to detect patterns in the histories, behavior and even genetic makeup of violent offenders. But if they rounded everyone up who fit the emerging profiles, the dragnet would draw in many more harmless individuals than it would murderers.

As law enforcement officials in Newtown, Conn., pore over the life of 20-year-old Adam Lanza for clues to what set off his murderous rampage, forensic psychologists and psychiatrists say they are getting better at near-term forecasting -- if the troubled individual steps forward to get help.


“Skilled and practiced mental health professionals have gotten a lot better at predicting short- term dangerousness,” said Dr. Steven E. Pitt, a forensic psychiatrist who consults with the Phoenix Police Department and directed the Columbine Psychiatric Autopsy Project after the 1999 school shootings. “But who’s going to commit violence in some unspecified future? You might as well consult a Ouija board.”

For psychiatric patients who show up in emergency rooms and doctors’ offices, a recent study found that a simple checklist of five recognized risk factors for imminent violence could bring the judgment of newly minted psychiatrists nearly up to the level of far more experienced ones -- who can identify a potentially violent patient roughly 70% of the time.

Patients more likely to become violent in a healthcare setting were most often those with active symptoms of mental illness and impulsiveness, who lacked insight into their conditions or were disconnected from reality, displayed negative attitudes such as lack of empathy or remorselessness, and who responded poorly to treatments such as anti-anxiety medications.

In assessing an individual’s propensity for violence beyond that, a few signposts have come to be recognized as reliable warning signs, even if they do not always point to trouble. Most important is a history of past violence, and not always violence that would rise to the attention of others. Animal cruelty or torture, or even an escalating history of fistfights, can send up a red flag.


University of Michigan psychiatrist Alan Teo, whose study of short-term risk assessment was published in September, noted that patients with “command auditory hallucinations” -- unseen things or people ordering them to do something -- are manifesting a “potent risk factor” for violence. At the same time, Teo warned that people exhibiting paranoia are at no higher risk than anyone with evident mental illness to carry out violent acts against others.

“The vast vast majority of people with mental illness are never, ever going to be violent, and even schizophrenics with paranoia are not going to be,” Teo said. “Paranoia in and of itself does not raise my rating of violence risk. The public should not assume a person with paranoia is actually a danger.”

“A lot of people every day are walking around with many risk factors for engaging in violence who do not,” Pitt said. “Other psychosocial variables have to come together” to move a troubled person to mayhem, he added.

In recent years, research has turned from a sole reliance on matters of nurture to investigate the genetic underpinnings of violence. Studying violent offenders, a pair of Duke University researchers has found that while a personal history of childhood trauma and abuse is common among those who harm others, that experience interacts powerfully with a genetic variation that influences the neurotransmitter serotonin. Their research suggests that for roughly 3% of the population, nature and nurture combine to increase an individual’s risk of engaging in violence almost fivefold.


Even if abused children were to be screened for the genetic variation, however, “There would be a fair number of false positives,” said Dr. William Bernet, a forensic psychiatrist and emeritus professor at Vanderbilt University School of Medicine. “It does not inevitably make a person violent.”

Finally, research into suicide -- a violent impulse turned inward -- speaks clearly about the predictive value of access to guns. Where troubled people can get hold of them, said the University of Michigan’s Teo, their prospects of hurting someone -- or themselves -- go up dramatically.

“It’s not only common sense,” Teo said. “We have good evidence that removing access to lethal weapons reduces tragic outcomes.”

At best, researchers say, predicting who will be violent and when will be a matter of forecasting -- building models of probability and risk that look more like those that predict a hurricane’s track than issuing a yes-or-no judgment.

“There’s an art as much as there is a science to this,” Pitt said. “If it were that easy, that straightforward, we’d be doing a lot better at keeping these things from happening." 

[For the record, Dec. 19, 2012, 1:16 p.m.: An earlier version of this post misspelled the first name of forensic psychiatrist Steven E. Pitt as Stephen.]