The Interior Situation of Suicide

The Boston Globe Sunday Magazine included an article by Peter Bebergal, titled “On the Edge.” (The teaser reads as follows: “Can a test reveal if a person has a subconscious desire to kill himself? Peter Bebergal, who lost a brother to suicide, goes inside Mass. General, where Harvard researchers are trying to find out.”) Here are a few excerpts.

* * *

Four years after my brother’s death, Harvard researchers at MGH are experimenting with a test they think could help clinicians determine just that. It focuses on a patient’s subconscious thoughts, and if it can be perfected, these researchers say it could give hospitals more of a legal basis for admitting suicidal patients.

* * *

This missing piece in the suicidal puzzle is what prompted the innovative research study now in its final phase at MGH. The study, led by Dr. Matthew Nock, an associate professor in the psychology department at Harvard University, is called the Suicide Implicit Association Test. It’s a variation of the Implicit Association Test, or IAT, which was invented by Anthony Greenwald at the University of Washington and “co-developed” by [Situationist Contributor] Dr. Mahzarin Banaji, now a psychology professor at Harvard who works a few floors above Nock on campus. The premise is that test takers, by associating positive and negative words with certain images (or words) – for example, connecting the word “wonderful” with a grouping that contains the word “good” and a picture of a EuropeanAmerican – reveal their unconscious, or implicit, thoughts. The critical factor in the test is not the associations themselves, but the relative speed at which those connections are made. (If you’re curious, take a sample IAT test online at implicit.harvard.edu/implicit/.)

The IAT itself is not new – it was created in 1998 – and has been used to evaluate unconscious bias against African-Americans, Arabs, fat people, and Judaism. But critics question whether the test is actually practical, and up until now no one has tried to apply it to suicide prevention. As part of his training, Nock worked extensively with adolescent self-injurers – self-injury, such as cutting and burning, is an important coping method for those who engage in it, though they are often unlikely to acknowledge it. Nock thought that the IAT could serve as a behavioral measure of who is a self-injurer and whether such a person was in danger of continuing the behavior, even after treatment. In their first major study, Nock and Banaji asserted that the IAT could be adapted to show who was inclined to be self-injurious and who was not. And more important, they said, the test could reveal who was in danger of future self-injury.

The next step, Nock realized, was to use the test to determine, from a person’s implicit thoughts, whether someone who had prior suicidal behavior was likely to continue to be suicidal. It would give doctors a third component, along with self-reporting and clinician reporting, and result in a more complete picture of a patient. Nock doesn’t assume that a test like the IAT would be 100 percent accurate, but he believes it would have predictive ability. “It is not a lie detector,” he says. “But in an ideal situation, a clinician who is struggling with a decision to admit a potentially suicidal patient to the hospital, or with an equally difficult decision to discharge a patient from the hospital following a potentially lethal suicide attempt, the IAT could provide additional information about whether the clinician should admit or keep that patient in the hospital.”

Over two years, researchers at MGH asked patients who had attempted suicide if they would be willing to participate in the test. About two-thirds of them agreed (some 200 patients) – even though some had tried killing themselves just hours before – and after answering a battery of questions about their thoughts, sat with a laptop and took the IAT.

During one test, a person was shown two sets of words on a screen, one in the upper left corner, one in the upper right. A single word then appeared in the center, and the test taker was asked to indicate with a keystroke the corner containing the word that connected to the center word. The corner sets were drawn from two groups of words (one group was “escape” and “stay,” and another was “me” and “not me”). In one version, the sets were “escape/not me” and “stay/me,” and the series of words that appeared in the center included, among others, “quit,” “persist,” “myself,” and “them.” The correct answers called for “quit” to be associated with the side that had “escape,” for “myself” to be matched with the side that had “me,” and so forth. In theory, a delay in answering on “quit,” even if the person got it right, could reveal that he was associating the idea of “quit” with the idea of himself. The word sets varied depending on the test, and bias could emerge in a positive or negative way. For example, if the sets were “escape/me” and “stay/not me” and a person hesitated in correctly matching “myself” to the side with “me,” it could reveal that he was associating himself with the idea of “stay.”

For about the next five months, Nock and his research team at Harvard will analyze all the data collected from MGH. If they think their findings show promise, they will follow up and run their experiment again to see if it yields similar results. If it does, they may seek to implement the test at an area hospital. For now, following up with patients will be pivotal in assessing the test’s effectiveness. Tragically, though, the only way researchers will know for sure whether the test can predict behavior is if a key number of patients attempt suicide again.

* * *

Nock says it’s still too early to tell how well the test will predict someone’s likelihood of engaging in suicidal behavior. But he says the hope is that the IAT will be able to record subtle distinctions between those who are at risk and those who aren’t by measuring how “positively or negatively people value the option of suicide as a potential response to their intolerable distress.

* * *

We recommend the entire article, which you can link to here. For a collection of Situationist posts about implicit associations, click here.