What does an Emma look like? Is there a prototypical Emma out there against whom all other Emmas are compared?
Turning the question inwards, we can ask ourselves if we look like our names.
It seems like a silly question. After all, most parents name their children before they are born. Even when named shortly after birth, babies tend to look alike. How could a face mature to fit a name it was given before being seen?
Yet, there is evidence that many of us do look like our names, and that strangers can pick our name from a short list better than if they were resorting to chance.
This face-name matching effect, beyond being strange, exemplifies how, in science, the way in which you ask your research question can affect the kind of answer you get.
What a Yaakov looks like
This effect was described in a 2017 paper authored by a team of Israeli and French scientists, who tested both humans and a trained computer in a total of eight studies in which a face had to be matched to its correct name. What is particularly interesting about these studies is that they reveal the scientists’ thought process, specifically how these scientists addressed the possibility that they were wrong.
They showed Israeli students 20 photos of young Israeli adults, one at a time, and asked them to choose which of five given names was the correct one. Someone choosing at random would, on average, achieve a 20% success rate. Each photo has five possible names, and one in five is 20%. Yet these students did better than that. They were, all in all, successful 28% of the time, with an above-chance success rate for 17 of the 20 faces.
But the scientists thought, “Hold on. What if the filler names we chose—the wrongnames—are somehow different from the true names, and our participants are picking up on it?” To give an egregious example, what if the list of possible names was Youssef, Yves, Yan, Yorick, and Bob, and the correct name was Bob?
So they conducted a similar experiment in which those filler names were the real names of other faces used in the exercise. Again, the participants achieved a statistically significant success rate above chance.
But what if there was something about the Israeli culture that was biasing the results? The researchers turned to France and repeated the experiment with 116 French participants looking at 10 headshots from young French people whose names were neither too common nor too rare. Each headshot had four possible names to choose from, for a success rate by chance alone of 25%.
The participants reached an overall success rate of 41%, much greater than in the Israeli studies. In fact, for each of the ten faces, the participants were better than chance at figuring out that person’s name.
There is something in those faces that a human recognizes as looking like a particular name, and a computer can do the same. The researchers trained a computer on a database of over 36,000 female faces of 15 female names and over 58,000 male faces of 13 male names. When the trained computer was given new faces to match, it also did better than chance on every single face.
This is the point at which we have to talk about Bob.
There is this idea that we might match names with faces strictly by the sound that these names make. A classic exampleis the presentation of two images, one of an athletic man with a narrow nose, a full head of hair and a goatee, the other of a bald, larger man with a broader nose. Students are asked who is Tim and who is Bob. Inevitably, the leaner man with the hair is said to be Tim, while the man with the rounder face is Bob. “Bob” sounds round, therefore we may associate a rounder face with the name.
A more abstract version of this phenomenon is the bouba/kiki effect. Two shapes are shown. One is a sharp, pointy explosion. The other is a blob with round fingers. We are told to match the shape with its name, either bouba or kiki. Most people say the pointy one is kiki and the blobby one is bouba. We expect the shape to mimic the sound of its name.
But that’s not what’s happening in the study of matching names to faces. Because when French students were asked to match Israeli faces and names, and when Israeli students were asked to do the same for French faces and names, their success rates plummeted. The effect seems to be culture specific. You have to know what a “Yaakov” typically looks like in order to recognize it in a face. It’s not in the sound.
These eight studies taken together are quite impressive, but they all emanate from the same group of researchers. Has this effect been replicated? And if it is real, how can a face start to approximate some invisible cultural prototype?
A Dorian Gray effect
In 1891, Oscar Wilde published his novel The Picture of Dorian Gray, in which a man remains young and beautiful while his painted portrait exhibits the signs of the man’s real-life debauchery. In psychology, the Dorian Gray effect refers to internal factors, like someone’s personality, affecting their external appearance. The link between a person’s face and their name might be a sort of Dorian Gray effect, by which growing up with a name leads to alterations in our physical looks.
The most obvious, of course, is hairstyle. It is possible that, within a particular culture, a certain name tends to evoke wilder hair, and that a person with this name, subconsciously wanting to belong, lets their hair grow out in a wild style to match this stereotype. In fact, in the Israeli paper, one of the studies consisted in presenting 10 faces either in full or as stand-alone hairstyles or the reverse, with the inner features of the face devoid of hair. In all three conditions, participants did better than chance at matching these photos to their names. In the hairstyle-only condition, they found the correct name better than with chance alone for nine out of the ten photos of hairdos!
And if someone does not look like their name, they may inherit a nickname that provides a better match. The name Robert is particularly plastic. It’s easy to imagine that parents and friends who see the boy Robert grow up not looking like their idea of a Robert might choose instead to call him Bob, or Rob, or Bobby, or Robby, or even Bert.
As for the Robert cultural prototype—the über Robert—it may be hard to imagine how it even comes about. Who decides what a Robert should look like? But given that certain names get inherited in some families, with Robert Jr. being the son of Robert Sr. and perhaps raising a Robert the Third, and with genes being also passed down from one generation to another, it’s possible to imagine how, in certain communities, a Robert might carry certain genetic traits recognizable on one’s face.
It may sound like I’m grasping at straws, trying to explain a very unreal finding in the scientific literature. Especially since, two years before the Israeli-French paper was published, a similar study was made public and it came to a negative conclusion, so much so that it was published in the Journal of Articles in Support of the Null Hypothesis, which is a very geeky way of saying “Papers In Which We Didn’t Find the Cool and Sexy Thing We Were Hoping For.”
In that study, they found that American women, not men, were more accurate than they would be by chance alone to match faces and names, but in a bid to replicate these findings using British students looking at British faces, the effect vanished.
However, there is a big difference between this study and the Israeli-French one. In the former, participants were given a name and they were asked to find which of two faces had that name. In the latter, participants were given a face and they were asked to find which of four or five names belonged to that face. You may think I’m splitting hairs, but matching a name to a face and matching a face to a name could, in theory, trigger different processes in the brain. There are many examples in the research literature of real effects that were missed or of non-existent effects that were detected because an inappropriate method was used. The devil in science is always in the details.
Because apart from this American-British mostly negative study (but not completely negative: let’s not forget the American women), there are other indications that the face-name matching effect might be real.
A decade before the publication of the Israeli-French paper, psychology students were asked to play with a software used by law enforcement to recreate the face of a suspect; only, they had to come up with a composite face that represented a typical Andy, and Bill, and Justin, for a total of 15 male names common among college-age students. The best ones were morphed into the ultimate exemplar, and students who had not participated in the exercise could match these faces to their names better than chance. Even though these faces were all fictional, the face that had been assembled to fit some cultural idea of a Justin was also easier to memorize when students were told its name was Justin than when its name was said to be Bill.
Another study shows humans performing above random noise when shown real faces and ten possible names. By chance alone, they should get it right 10% of the time, but they reached a nearly 14% success rate.
Finally, there are the slightly more fun replication attempts that followed in the wake of the Israeli-French paper, although no firm conclusions can be drawn here. The French magazine Slate ran a survey where readers were asked to match faces to names, although I was unable to access its results. Meanwhile a professor at the University of California at Berkeley had 106 students try to match the faces of male faculty members in her field with their names. She sent me her results but pointed out that this very loose exercise was partly done “for entertainment purposes” and that there were many limitations to it. Given that every photo was accompanied by the same list of 16 names, but it wasn’t clear from the instructions that each name corresponded to one photo, calculating the odds of getting a correct match by chance becomes harder to compute. Also, unlike in the Israeli-French study, the people in the photos were not told to adopt a neutral expression. Many are smiling; others are not; and their clothes are visible. An interesting attempt, but not one that allows us to come to a conclusion either way.
There is a big limitation to the research surrounding this effect. Most of these attempts at detecting it were done using participants in their 20s looking at photos of other people in their 20s. Do older adults look like their names? Can middle-aged people perform better than chance at this matching exercise? Future research might tell us.
But if this effect is indeed real—if, to a small but detectable degree, many of us resemble some cultural idea of what we should look like as we reach adulthood—it would be an extraordinary example of the Dorian Gray phenomenon. To quote one of the authors of the Israeli-French paper, our face, like a stone, is metaphorically sanded down by the sea over years. In small ways, we adjust our appearance to match some intangible, abstract archetype that is shared within our culture.
Our names, it seems, are printed faintly onto our faces. I look forward to more research confirming this effect and digging deeper into exactly how we come to look a little bit like the name our parents gave us.
- Studies report that when people are asked to find someone’s first name from a short list after looking at their face, they tend to do better than if they were purely guessing, and trained computers can also do the same
- It is thought that we might subconsciously alter our appearance as we grow up, for example our hairstyle, to match some cultural idea of what someone with our name looks like
- The studies done so far are limited in that they involve people in their 20s trying to match names and faces of other people in their 20s