Courtesy Yonat Zwebnerg
Take a good look at this photo. His name could be Jacob, Dan, Josef, or Nathaniel—which do you think is right? You might think your chance of guessing correctly is only one in four, but a new study shows that you may know more than you think.
We learn new names all the time, but sometimes we have real trouble remembering the name of someone specific, or we constantly call them by another name. “It’s common to hear that someone’s name really suits them or that it doesn’t suit them,” says graduate student Yonat Zwebnerg of Hebrew University of Jerusalem. “It made us think. Could it be that most people do look like their names?”
So together with colleagues from HEC Paris and the Interdisciplinary Center and Columbia University, Yonat began a whole series of studies to see whether it’s possible to predict someone’s name simply by looking at clues in their face. They assembled thousands of head shots, then suggested four possible names for each one, including the person’s real name, and asked volunteers to guess which one was correct. The results, published in the Journal of Personality and Social Psychology, were astonishing.
Instead of the statistical likelihood of correct guesses (25 percent), the percentage of correct guesses was significantly higher at around 30 percent or more. (That may not sound like a lot, but in research terms, it’s a huge jump.) “What we found is that people repeatedly identified the true name of the person in the picture significantly above chance level,” Zwebner says.
The scientists were careful to eliminate any other clues that might influence the guesswork, such as clothing, but the results remained the same. It seems that humans are able to recognize certain characteristics of a face, which help them identify their name.
Over time, we seem able to “learn” facial characteristics that match individual names, a kind of face-name matching. However, this only works when we’re looking at names within our own culture, and doesn’t include middle names.
If humans can learn this skill, then perhaps computers could learn it too? The team took their research further. Using only two names, they showed multiple images to the computer, which then mapped the facial characteristics of each name.
Yonat explains what happened: ‘We gave the computer a new set of faces of A and B, and asked the computer to decide who is A and who is B. Surprisingly, we found that the computer also identified correctly at significantly above chance level,” she reports. “It tells us that it’s nothing about the name, or some human bias. It’s something in the face that looks like that name.”
The team is planning to explore the implications of their findings, such as whether our expectations about a name can influence whether or not we trust someone (these are the subtle habits that make people trust you). But meanwhile, if you find yourself wondering about the name of the commuter sitting opposite you, you could be more accurate than you think.
Oh, and the name in the photo? It’s Dan. Were you right?