Campus & Community

Test improves prediction of self-injurious behavior

3 min read

Researchers have found a way to better predict self-injurious behavior by using a test that assesses subjects’ implicit attitudes toward self-injury rather than relying on self-revealing talk. The test addresses a major challenge in identifying people who engage in self-injurious behavior, because they are often uncommunicative in order to avoid unwanted treatment and are also often unable to articulate their feelings.

Conducted by two Harvard professors, the study was led by Matthew Nock, an assistant professor of psychology, with Mahzarin Banaji, the Cabot Professor of Social Ethics in the Department of Psychology and Pforzheimer Professor at the Radcliffe Institute for Advanced Study. The paper is published in the May issue of the American Journal of Psychiatry.

“This research represents a significant advance in risk assessment for self-injury. This test measures the associations people hold about self-injury, using their own behavioral responses rather than their verbal report — and therefore, it has the potential to significantly improve our ability to detect and predict self-injurious thoughts in a way not previously possible,” says Nock.

The study included 89 adolescents between the ages of 12 and 19, of which 53 had a history of nonsuicidal self-injurious behavior, and 36 were noninjurers that comprised a comparison group. The group was given a version of the Implicit Association Test (IAT), adapted to measure mental associations specifically to cutting of the body; prior studies have shown that cutting is the most common form of self-injury.

The IAT was computer administered with participants using keyboard controls to quickly associate self with cutting or not.

In the identity test, participants grouped together words and phrases such as “cutting” or “no cutting” with words such as “me” or “they.” Those who engaged in self-injurious behavior were more likely to quickly and consistently group the words that are relevant to cutting with words related to the self such as “I,” “me,” “mine,” “myself.”

In a second attitude test, participants were asked to group together words such as “cutting” or “not cutting” with words that captured the categories “good” and “bad.” While both groups showed an association between “cutting” and “bad,” the association was significantly weaker for self-injurers.

In addition to the IAT, participants were also assessed using more conventional measures of self-injury. The assessment also obtained measures of demographic factors, differences in IQ, and presence of psychiatric disorders. When considered in conjunction with the more traditional indicators of self-injurious behavior, the IAT significantly improved the prediction of who was a self-injurer.

“Among the most important applications of the IAT is its ability to reveal the mind’s state when it is psychologically compromised — as it is in a full range of psychological disorders,” says Banaji. “Clinical psychologists have been at the forefront of applying the IAT to improve the ability to predict and treat mental illness. Such applications show the importance of basic research in understanding mental illness.”

The IAT was developed in 1998 by Banaji, along with Anthony Greenwald of the University of Washington. Since then, more than 4.5 million tests have been completed, through the Web site at

Nock and Banaji’s work was funded by the National Institute of Mental Health, the Milton Fund, and the Talley Fund of Harvard University.