Say you have twins, identical in every way. Will their difference in names cause them to look different over time?
Yes, argues Marketing Professor Jacob Goldenberg and his colleagues. Perhaps more surprisingly, the twins’ facial appearances will also conform to a cultural stereotype around that name: Jacobs, for example, all tend to look like a Jacob, while Rachels all have something very Rachelish about them.
Goldenberg and his research colleagues are now collecting the data to prove it, following up on a headline-making study published a year ago called, “We Look Like Our Names,” co-authored with Yonat Zwebner of the University of Pennsylvania, Nir Rosenfeld of Harvard, Anne-Laure Sellier of HEC Paris, and Ruth Mayo of the Hebrew University of Jerusalem.
“The fact that you call someone Pinocchio doesn’t mean that their nose will be longer,” says Goldenberg. But he adds that other, more malleable facial features and expressions, such as the wrinkle in a smile or the squint of an eye, may conform to the stereotype of Pinocchio. “You can control a lot of changes in your face — your smile, your haircut, your glasses, your makeup, the way you shave, the way you comb your hair. And that’s a lot.”
This is peak season for naming babies, as late summer tends to see the highest number of births in the US. Goldenberg’s ongoing research into the manifestation of name stereotypes in facial appearance suggests that parents have an especially crucial decision to make in these weeks. What you name your children may influence how they look.
Faces reveal a wealth of information.
Studies have repeatedly demonstrated something called the “Dorian Gray effect,” named after the Oscar Wilde novel in which the deeds of the protagonist affected his portrait. Faces reveal a wealth of information, including political affiliations and willingness to cooperate. We judge political candidates by their appearance, electing faces that are perceived to be more competent-looking.
But whereas earlier research focused on the effect of facial appearance on social perception, Goldenberg and his coauthors focused on the effect of social perception on facial appearance. Receiving a name is a form of “social tagging,” as every name comes with associated expectations regarding characteristics, behaviors, and even a stereotypical “look” within a culture.
“We argue that once a baby receives a name, this social label (name) leads to certain social expectations, inferences, and interactions,” Goldenberg and his co-authors wrote in their 2017 paper, published in the Journal of Personality and Social Psychology. “The social attitude that a name triggers and its influence on one’s identity may affect the individual’s self-perception and development of traits, which eventually may be realized in his or her facial appearance.”
Across eight studies with more than 94,000 faces in two countries, Goldenberg and his colleagues found that people and computers could match a name to a face at a surprisingly high rate, but less so cross-culturally or when nicknames were involved. The results are not trivial, as having a good face-name fit might influence perceptions of attractiveness, trustworthiness, and even compatibility of job candidates.
In their first study, 120 Israeli students (half men and half women) were presented with the images of 20 faces and told to choose the “true given name” from among five options. The faces, all between the ages of 20 and 30, were drawn from a social media website. Chance would have seen the correct name chosen one in five times (20 percent), but in fact the participants accurately matched the target’s true name 28.21 percent of the time. Another cohort of 67 Israeli students were likewise shown 25 faces in succession and told to choose the correct name from four options; they were correct 30 percent of the time, above the guess level of 25 percent.
A second study essentially replicated the first experiment with French students and names. When presented with one face and four name options, the 116 participants accurately chose the target’s given name 40 percent of the time — well above the chance level of 25 percent. A third study attempted to further control for age and ethnicity variables, again finding that information conveyed in the image of the face enabled face-name matching.
If it’s true that faces conform to a cultural stereotype around a given name, then someone from outside that culture should not be able to match names to faces. To test this, the researchers assigned the French students to identify the Israeli faces and names, while the Israeli students tried to match French faces to names. As predicted, the students could not match faces to names above chance level for foreigners.
Computers can also match faces to names, the researchers found. To “train” a computer to recognize the prototypical face for the name of Benjamin, for example, the researchers gathered photos of 3,014 people named Benjamin from a popular business-oriented social platform in France, enabling the computer to then create an “average” of that face. This was done for 15 different female names and 13 male names using a total of 94,000 images.
While a human may possibly use personal, social, and historical contextual information to match faces to names, a computer is bias-free and only acknowledges pixel values; to the computer, a name is just a binary number without social meaning. Yet the computer still chose correctly 60 percent of the time when presented with two options. For some faces, the computer identified the correct name up to 95 percent of the time, underscoring that some faces look especially like a certain name. The computer also tripped up more often with certain names, such as Emilie and Jeanne, suggesting that some names share a similar facial profile.
Hairstyle plays a part in name recognition, the researchers found. In another study, 192 Israeli students were shown only a hairdo, and still they matched it to the correct name 33 percent of the time (when 25 percent would have been random). Hairstyles are relatively within one’s control, which implies that a person actually chooses to look more or less like their name.
Nicknames also affect appearance. While study participants correctly matched people without nicknames 52 percent of the time (when 25 percent would be random), face-name matching dropped to 19 percent for people who go exclusively by unique and uncommon nicknames. The results suggest that someone like the jazz trumpeter Louis Armstrong, who went by the nickname “Satchmo,” would have looked less like a Louis both to himself and to others because he didn’t identify with his legal name.
Goldenberg cautions that the research doesn’t imply that we’d all look alike if we shared the same name. The face-name matching effect breaks down when a name is either too rare (so it lacks a cultural stereotype) or too popular (so there’s too much diversity in the name). For example, the most popular male name in the US from the 1960s through the 1990s was Michael, according to the Social Security Administration, resulting in the name “Michael” possibly losing its stereotypical meaning.
Based on all that research, Goldenberg suspects that twins’ appearances will over time also align with whatever they’re called —implying that for twins to remain especially twin-like, they’d need to share the same names and not just the same genes. He plans on testing the hypothesis in a new study with twins in the US.
“It’s amazing how much information is encapsulated in the face, and even more amazing how we can decode it,” says Goldenberg. “I still don’t understand exactly how it works. We don’t control every part of our faces and how we look. Yet we still convey all this information.”
Read the Research
Read the original piece on Columbia Business School’s Ideas and Insights blog.