A study released this month by Czech researchers proves that you can tell the intelligence of a man just by looking at him, but women are a closed book. This probably happens because of all the man hormones telling the ladies what good baby-makers they would be. Or something. At least that is what all the articles written about the study seem to suggest. The truth is a bit more complicated.
The researchers for the study, which is available for free on PubMed, took the photos of 40 men and 40 women, all biology students at Charles University in Prague, and gave them an IQ test*. They then had 160 Humanities students rate the photos on attractiveness and perceived intelligence. The researchers then ran a regression on it (basically a best-fit line with controls) and determined that there was a correlation between perceived intelligence and actual intelligence in the men but not so for the women. They then speculated on why this may be so.
The first thing I noted when reading the study was that the subjects that were photographed and rated were not particularly diverse. Their IQs were quite high compared with the general population, probably because they were all biology students. On the typical IQ scale, the median person has an IQ of100 and we would expect half of the population to fall below 100 and half above. The sample of people that were photographed for the study, however, had an average IQ of 125. Only 7 of the 80 photographed students (about 4 men and 3 women) had IQs under 100. Ethnicity was not mentioned anywhere in the study but considering this was done at a university in Prague, they were likely to be mostly white. So, already this study only applies to 20-something, smart, (probably) white Czech college students for both the raters and the students being rated.
The researchers found that the college students rated photographs of their peers as more intelligent the more attractive they were. However, there was no actual correlation between a subjects’ attractiveness and their IQ. In other words, student raters were misled by the attractiveness or lack thereof of their photographed peers and wrongly assumed the best-looking people were also the smartest.
The researchers also did some sort of fancy schmancy grid-based regression thing** on the photos of the subjects to determine which facial features are associated with intelligence. They determined that “in both sexes, a narrower face with a thinner chin and a larger prolonged nose” were perceived as being associated with intelligence. However, the researchers also noted that these facial features were not associated with actual intelligence. In fact, there were not any visible facial indicators that the researchers could find that were associated with IQ.
But wait, the title of the paper is “Perceived Intelligence is Associated with Measured Intelligence in Men but Not Women.” Except, I just told you that the students’ perceived intelligence ratings were strongly associated with attractiveness and certain facial features, but neither of those things were actually associated with IQ. So…what exactly is going on here?
Well, when the researchers looked at the association between perceived intelligence and actual intelligence, they controlled for attractiveness. They took out any bias that could be in the data regarding attractiveness and then saw that, for men at least, there was a small correlation between perceived intelligence and actual intelligence. So, if you are a robot who has no concept of “attractiveness” then perhaps you can tell whether a man is intelligent just by looking at a photo of him. If you’re human though, too bad because you’re probably going to be too distracted by his sexy “narrower face with a thinner chin and a larger prolonged nose” to be able to make any accurate judgment.
Ok, so maybe we each have a special super power inside of us that can accurately perceive the intelligence of a man if only we could stop looking at his sexy, thin chin for a second, but lady faces are still an enigma, right?
Uh, well… it’s a bit more complicated than that.
The researchers provided two handy charts showing the perceived intelligence versus the actual IQ for the men and women in the study along with the quadratic model they fit to the data. The chart for the men is on the left and the chart for the women is on the right.
Let’s start first with the chart for men. The first part of the quadratic model that they fit does seem to match the data. Just eye-balling it, there does seem to be a positive relationship between actual IQ and perceived IQ up to an IQ of around 130. At that point, because they fit a quadratic model the relationship actually becomes negative, meaning that for men with really high IQ’s the raters actually thought they were less intelligent. Looking at it though, it doesn’t seem like it necessarily is actually negative up at the higher IQ levels. There seem to be less than 10 data points of high IQ men to go off of, so it’s hard to really tell what happens to perceived intelligence for the men that were above an IQ of around 145 or so. It’s possible that rather than a negative relationship, it’s just that intelligence is easier to perceive at the lower levels of IQ but tops out at around an IQ of 120. Because the researchers used a quadratic model, the far edge is always going to have a negative slope, but that’s fine as long as you don’t read too much into it.
Except that the researchers took the negative correlation between perceived and actual intelligence at the high IQ level as fact and then speculated in their conclusion about why this might be saying it could “reflect an adaptation to an upper intelligence limit as men with an extremely high IQ could find such practical tasks such as care and protection of women and offspring, less rewarding.”
Yes, you read that correctly. The researchers used the fact that there was a slight negative correlation among the 10-ish male subjects that had high IQs to determine that very smart men are unable to care for women and children and therefore ladies don’t want to make babies with them. Remember that they used a quadratic model so it was going to have a negative slope at the high-IQ end of the spectrum because that is the shape of a quadratic equation. They’re basically wildly speculating on the nature of mankind based on a correlation that may or may not even exist.
Moving over to the chart for the women, it definitely looks like there is far less correlation between actual and perceived intelligence than there is in the chart for men. But, comparing the women’s chart to the men’s chart shows that there are actually some other interesting differences between the groups. You also need to be a bit careful when comparing them because both the x and y axes are on different scales.
The men seemed to have the strongest correlations down at the lower IQ end. For men under an IQ of about 110, the raters were correctly able to perceive that they were of relatively low intelligence compared to their peers. There were about 9 men with an IQ under 110. For the women, there were only about 6 that had an IQ under 110 and they are clustered closer to a 110 IQ. In other words, it’s possible that at the lower-IQ end, the same pattern holds for women as men in which it is easier to perceive a low-IQ, but we can’t really tell from this study because there weren’t enough women at that lower end to be able to actually see any correlations.
At the higher IQ end, the groups were even more different. There were about 11 men with an IQ above 140 and about half of them had an IQ above 150. For the women, there were only 7 with an IQ above 140 and none with an IQ above 150. We therefore can’t really compare perceived intelligence of high-IQ men to high-IQ women because there are just too few high-IQ women in the sample.
Additionally, the perceived intelligence ratings were widely different between men and women. In general, men seemed to be all rated at near an average intelligence. Only one man was rated on the relative scale as having a perceived intelligence above 0.5 and three were rated as being below -0.5. As for the women, seven of them were rated as being perceived as exceptionally intelligent (above 0.5) and 5 were rated as being perceived as having low intelligence (below -0.5). In other words, the students tended to rate the women as being of unusually high or low perceived intelligence, while the men tended to all be rated at around average.
There aren’t very many data points to go off of here, so you can fit a lovely model to it and it may even be statistically significant, but that merely suggests that perhaps there is a relationship. It certainly doesn’t prove anything. And likewise, even if there doesn’t seem to be a statistically significant relationship like we saw for the women, it doesn’t necessarily mean one isn’t there but possibly that we don’t yet have enough data to be able to see it.
The study itself is not a bad one and the researchers’ methods and results fit the work they did, but it’s such a preliminary, small study that its results shouldn’t be taken as proven fact. It’s merely suggestive that this could be an area that needs more study.
The only truly unscientific piece in the study was the discussion at the end where the researchers conjectured possibilities for why the subjects in their study were correctly able to perceive the intelligence of men but not women. They said that perhaps “cues of higher intelligence are sexually dimorphic and are thus apparent only in men’s faces” or that possibly “women are pervasively judged according to their attractiveness. The strong halo effect of attractiveness may thus prevent an accurate assessment of intelligence in women.” Except…did they not already control for attractiveness? They also mentioned that it was possible that “women prefer dominant men as extra-pair sexual partners while at the same time they seek men who are more willing to invest in their offspring as long-term or social partners.” They even go on to suggest that future research should run similar tests while keeping track of whether the women were menstruating.
Basically, the entire last page of their study is pure wild speculation with no actual evidence to back any of it up, and certainly no evidence from this particular study.
Media reports about this study claim it shows that men’s intelligence is written on their face but women’s faces obscure their intelligence. In actuality, this study showed that when we try to judge a person’s intelligence just by their looks, we’re really bad at it because we’re too blinded by their attractiveness or lack thereof to get a good assessment. If we were robots that could ignore attractiveness, we may be able to get an ok assessment of the intelligence of men provided they are white, 20-something smart college students and we are also a 20-something, white college robo-student, but even then it’s pretty unclear and in fact there are no objectively measurable facial features associated with intelligence.
In other words, you can’t judge a person’s intellect by just looking at them, not even if they are a man. You may all now wildly speculate on the evo-psych explanation for that.
* I am not going to go into the reasons here why IQ is not necessarily a very good measure of intelligence, mostly because that is not my area of expertise, but Will provided a link to this article that does a good job explaining it.
** I am not trained in nor have I ever done anything involving grid-based statistical analytics so their description of how they did their facial analysis was a bit over my head (pun intended).
Nice post, Jamie. I found this particularly interesting/problematic (from the article):
I did not see anything about sexual orientation in the article. Aside from the baseless nature of their speculation that you mentioned, they also seem to have made the assumption that all of the people involved were heterosexual. Whooooops heterosexism!
Oh fuck. I was totally going to mention their hetero assumptions of sexual orientation but then when it came to actually writing the post, I completely forgot that point. That’s my own hetero-privilege right there.
Those plots are laughable…
IMHO, a more reasonable interpretation is that both charts show no correlation at all. I don’t know whether you thought that was too obvious to require mentioning, but I wanted to put it on the record.
Another reasonable interpretation is that the researchers have no clue about statistics (or science), and they just blindly dropped the numbers into some curve-fitting formula. The phrase “cargo cult statistical analysis” comes to mind.
Show these plots to a physics or other scientific conference and ppl will laugh you off the stage…
Although I dot think it’s an excuse for these researchers or this study (which was pretty terrible), I do want to point out that I don’t think you can compare social science research to that of physics or other natural sciences. Social science research is always going to be messy and less clear because it cannot be properly controlled the way natural science research can. Social science in general has to work with less precision because real world data is less precise.
That doesn’t excuse this particular study, but it is something to keep in mind.
Yah, during my last edit I actually cut out a section explaining what it means when a model is statistically significant and why just because it’s statistically significant doesn’t mean it’s correct (that section was a little too nitty, gritty in the weeds and ruined the the flow and understandability of the writing. I felt that including it actually made the piece harder to understand so I cut it). The men’s chart is clearly clustered. It is not randomly distributed along the Y axis the way the women’s chart is. If any possible slope can be detected at any point in the chart and the model is closish to that slope at that point, it will be statistically significant, even if the model is completely wrong. I think this is fine to do I this type of preliminary research as long as the researchers understand the limits of how these types of regressions work and understand that the results can’t be anything more than suggestive.
The researchers attempted to fit both a linear model and a quadratic model and the linear model fit closest. They didn’t post their actual equation or even their resulting coefficients in the paper. When reading the paper, I did so under the assumption that they actually understand what a regression means and how it works and what it measures, but mostly because I can’t see anyone running a regression who doesn’t understand it. I mean, it’s easy to do. In most programs you just click a button or run one line of code, but It’s meaningless if you don’t understand the math behind it. Honestly though, some of the statements the researchers made in the paper made me suspect they might have no idea what they are doing. I don’t know how common it is for researchers who are not well trained in statistics to do studies like this, but it’s certainly worrying.
Oops! This was meant. As a reply to Amm1.
Can we really rate “attractiveness” from a photo? Is beauty skin deep? Have you seen your partner’s passport photo recently? Perhaps the results (if any) tell us more about the observers than the observed!
This study reminds me of a recent game show on TV where the panelists were asked to pick the criminal from a set of photos. Results were, as you might expect, pretty random, interesting and hilarious.