Can you tell criminals from non-criminals by sight?

Inspired by this GQ thread, I had a look at this article, which discusses a recent paper (pdf) presenting evidence that people can judge whether or not a person is a ‘criminal’ (the crimes looked at by the study were arson, assault, drug dealing, and rape) by simply looking at pictures of them, better than chance.

Now, this post is not about the validity of the science, presented in either the article or the study. I merely want us to conduct a little informal test of our own. Thus, I ask you to have a look at the pictures, write down the numbers of those that make your spidey-sense tingle, and check them against the list of numbers indicating criminals I will provide in a spoiler box below (which can also be found in the appendix of the paper). There are 16 criminals, and 16 non-criminals in the picture. Please, be honest and don’t peek!

Then, just indicate the number you got right in the poll options.


3, 4, 5, 8, 10, 11, 16, 20, 21, 23, 24, 27, 28, 29, 31, 32

Oh, and perhaps it would be interesting if you could add, in a spoiler box of course, which ones you got right, and which non-criminals you misidentified… My answers were:

got right: 3, 4, 8, 11, 16, 20, 21, 23, 27, 28, 29, 31; misidentified: 1, 6, 7, 15

Oh look, all the faces are male.

Was this a disguised essay on sexism?


got right: 4, 8, 10, 11, 16, 20, 21, 23, 27, 29, 32 ; false positives: 1, 6, 9, 14, 19. There were some, like 21, that seemed extremely likely. I’d stay away from someone who looked like that if I met them

Not only all male, but all young male. Do no old people commit crimes these days?

12/16 here.

I’m guessing young white males were chosen speficially not to skew the results as, if other demographics were included, that would affect perceptions of criminality and make people more/less likely to choose them based on their demographic group.


Correctly identified: 3, 4, 8, 10, 11, 16, 21, 23, 28, 29, 31. False positives: 1, 6, 7, 18, 19. I did waffle a long time on 20, 24, and 27, but I psyched myself out. I just stuck 6 and 7 in at the end to get to 16. And if I saw 28 and 29 coming up a dark street toward me, I’d definitely head the other way.

What characteristics did you tend to use?

As for me, some of the ones I used were:
A square jaw, looking straight into the camera, no effort to look amicable in the way that number 30 does. Generally, when they seemed to have a lot of testosterone and want to project a “I want to look tough” look, that was a reliable combination.

I got 8 of them and had 4 of the wrong ones in my list (I missed that part about choosing 16 people). I think what made this a bit difficult is that they all appeared to be mugshots and the person seemed to pick people that all looked somewhat similar so it (to me) seemed to be more about demeanor then anything else. For example, number 4 looks like he just got picked up by the cops and he’s waiting for his dad to come pick him up for the 3rd time (I picked him).

I positively identified seven, which is how I voted. I also kept a list of maybes and 5 out of seven of them were criminals – should have trusted my instincts more. Six falsely identified though.

I pcked them out according to whether they had a prison look, like a hardness about the eyes and also for “lairyness” a sort of false friendliness.



Correctly identified: 3, 4, 8, 11, 16, 21, 23, 27, 28, 29, 31. False positives: 1, 6, 19, 22, 25.


As it stood, every single criminal, every single one, were in some state of frowning. If they weren’t actively frowning, they had a rigid look to their mouths where it felt as if it were going downward.
To be sure, many of non-criminals were also frowning. But there was a considerable number that were either smiling or had a softness to their straight lined mouth that indicated passiveness, not tenseness.
I assume, but can’t be positive, that for the criminals they used the mugshots as the photo of choice. Obviously with the non-criminals no mugshot would be available. I think it would have been interesting to see a set of 32 where none of the pictures were mugshots. Can you pick out the criminal when he’s just smiling with his friends?

11/16. I realized a few rows in that I had started trying to pick out the “mugshot look,” which I suspected might skew things; if I had it to do over again and really specifically looked for that, might do better. I also recall seeing a photo piece in the NYT a while ago that focused on “candid” shots of criminals in the London(?) underworld with their friends, families, etc. I don’t recall them looking all that unusual in the face, though obviously personal style was a bit different.

Correct: 3, 4, 8, 10, 16, 20, 23, 27, 29, 31. Misidentified: 1, 2, 6, 9, 12 18.

They all looked like criminals to me.

But I used to work for a bank.

I got 12.

3,4,8,11,16,20,21,24,27,28,29,31. I incorrectly picked 1,12,14,18

Yeah, I think if they weren’t mugshots, it would be a lot harder. I’ve known people who committed some of those crimes, and when they haven’t just been apprehended, they generally look like normal people. I think what we’re seeing is a tinge of either, “Oh shit, I’m in trouble now,” or “Fuck you, coppers,” or a combination of the two - attempting to hide their fear with a hard, cavalier attitude. The people who I thought weren’t criminals generally looked serious, but not especially scared, sad, or angry.

There was one exception: #3 looked too comfortable, like he was thinking, “Yeah, I did it, and I’m totally going to get away with it - you’ve got no proof.” I pegged him, too. But it was a similar expression on #19 that made me choose him incorrectly.

The study attempted to control for the mugshot issue – participants that thought some of the images looked like mugshots did not score any better than participants who didn’t pick up on that, so apparently this didn’t introduce any helpful additional information.

I got what I expected: 8/16, which is what I’d get if I had picked people at random.

Right - I don’t think the participant knowing that they’re mug shots makes a difference. But I do think the situation (being arrested) may have added some element to the criminals’ expressions that would not have been present if they were in an emotionally neutral situation. I suspect that, whether or not I know a specific photo is a mugshot, I’m still able to pick up on the subtle cues of anger and fear that the criminals would have been feeling and showing while having a mugshot taken, that the non-criminals would not.

In other words, they look guilty because they feel guilty - because they just got caught. If you took photos of them at some other time, outside of the context of their crime, I’d guess they would be harder to identify.

I got 9 right. The Court of Terraplane’s Facial Judgment Ability would wrongfully imprison only a mere 44% of defendants. Sounds great!

Correctly Identified: 3, 4, 5, 10, 11, 20, 27, 28, 31
Misidentified: 1, 7, 14, 15, 17, 19, 22