Follow me on Twitter

Blog archive

We Participate In:

ABA Journal Blawg 100!

Subscribe to The Jury Room via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.


I can tell from your face that you are suicidal

Friday, December 14, 2012
posted by Douglas Keene

Now this is strange. We’ve written before on studies showing you can identify Mormons by faces alone. And Scientific American tells us you can also identify gay men by faces alone. So now we have research telling us you can also identify who is suicidal from looking at photos of their face. We’d caution you to not try this at home. And while you are not doing this at home, you may also want to take these findings with a grain of salt.

University of Toronto researchers decided to examine whether research participants would be able to identify the difference between those who successfully suicided and “living controls” based on yearbook photos alone. (How do you even think of ideas like this?) Photos of people who had committed suicide (40 photos in all, 12 women and 28 men) were culled from high school and university yearbooks. Each photo was then matched with the nearest photo in the yearbook of a student (who remained alive) and was of the same gender and race. (Because we know you want to know, the authors searched Facebook and other social networking sites to verify that the control photos were indeed of persons still living.)

The authors standardized the photos so that each was converted to grayscale and cropped close to the face/head (as in the A side of the graphic illustrating this post).

In Study 1A, 33 undergraduate student participants were asked to view all 80 photos (40 successful suicides and 40 living controls) and to quickly determine (“based on your gut instinct”) whether they believed the person in the photograph had committed suicide or was still alive. While it did not matter whether the research participant or the individual depicted in the photo was male or female, participants were able to successfully identify those who had suicided at a level significantly above chance (p = .01).

For Study 1B, the researchers wanted to be sure that hairstyle and photo backgrounds or face shape were not affecting decisions. So they cropped the photos further to show only “internal facial characteristics” (as in the photo marked B in the graphic illustrating this post). 30 participants examined the photos and again, were able to identify the individuals who had committed suicide at a rate better than chance guessing (p = .005).

For Study 1C, the researchers (who had by then exhausted their own yearbook sources and those of their friends) looked to the internet for photos of people who had committed suicide. Naturally, since everything is on the internet, they found a website. They gathered the “first 25 targets from the website between the ages of 14 and 19 years old and posing face-forward to the camera”. A research assistant (who did not know these were photos of people who had committed suicide) gathered matching photos of those who were gender and race equivalent to the photos of people from the suicide website. If the person who had killed themselves wore eyeglasses, so did their matched (and verified to be alive) photo-partner. Finally, all the photos were cropped to display only the internal facial characteristics as in Study 1B. This time 29 participants viewed all the photos and again (p = .02) they were able to identify those who had committed suicide at a level higher than chance.

For Study 2, 161 undergraduate students rated the photos from Study 1A for depression, hopelessness, satisfaction with life or impulsivity. Suicide victims were seen as evidencing more impulsivity (p < .05) but there were no differences seen on the other factors. When the researchers looked at the specific photos adjudged to have suicided, they found the participants saw those individuals as seeming more depressed. “Thus, inferences of depression and impulsivity contribute to individuals’ perceptions of suicidality. Distinct from depression, only inferences of impulsivity actually predict whether an individual commits suicide”.

For Study 3, the researchers asked 133 participants to rate each face in terms of how likely “they thought the person pictured might be to make an impulsive purchase, to engage in an impulsive sexual behavior (unprotected sex), or to be involved in an impulsive violent act (a bar fight)”. Successful suicides were judged more likely to be engaged in a violent altercation in the heat of the moment (p < .05) but not seen as more likely to engage in unprotected sex or an impulsive purchase. Researchers conclude that since suicide constitutes a violent act against the self, there is a cue in the facial appearance that indicates impulsive violence is possible to the observer.

For the final study, researchers wondered if the size of the effect seen in the previous studies (55% accuracy in identifying those who had suicided) would be higher in mental health professionals than in community laypersons. They asked 36 psychotherapists (master’s and doctoral level) and 39 community members to examine the photos used in studies 1B and 1C and identify those they thought had suicided and those that they believed to be alive. Again, both the community members (p = .009) and therapists (p < .001) but the therapists were not statistically more accurate than were the laypersons.

The researchers (accurately) point out that it is important to realize these sorts of judgments are inaccurate almost as often as they are accurate and that there is no evidence to support the use of facial appearance as a diagnostic tool in suicidality.

Rather, this work speaks more to the basic phenomena surrounding the expression and perception of suicidality at rates that are statistically significant but perhaps only modest in practical significance (e.e., a mean identification rate of 55%).”

We are grateful to these writers for pointing out a flaw in their research and attempting to circumvent a rush to judgment (like the one we’ve seen in the nerd defense) wherein readers of this research think you actually can “see” suicidality on someone’s face. The discussion of how to understand the statistical nature of studies like this (and statistical base rates in general) is beyond the scope of this post, but expect to see it here soon. It’s an important concept too often glossed over by the popular media as they “cover” research findings.

***We appreciate being included in the ABA Blawg 100 for the third year in a row! If you like our blawg, take a minute to vote for us here (under the Trial Practice category). Thanks! Doug and Rita***

Kleiman, S., & Rule, N. (2012). Detecting Suicidality From Facial Appearance Social Psychological and Personality Science DOI: 10.1177/1948550612466115

Image: Study stimulus

%d bloggers like this: