Humans are exceptionally good at recognizing faces they’ve seen before. It doesn’t take much study to accurately recall whether or not you’ve seen a particular face. However, this pattern breaks down when faces come from unfamiliar races. A white person who lives primarily among other whites will have more difficulty recognizing Asian faces, and vice versa.
But how engrained is this difference? How much experience with other-race faces do we need to have before we can recognize them as well as same-race faces? Is learning to recognize other races as difficult as recognizing any new category of objects — cars, say, or birds? When we do learn to recognize other-race faces, do we really know them as well as more familiar races?
While it has been known for some time that we can learn to recognize other race faces as well as our own, this last question hasn’t been studied as thoroughly. Maybe in more difficult tests of recognition, we wouldn’t do as well with different-race faces.
To explore this question, a team led by Elinor McKone developed a clever set of three experiments. In the first experiment, white Australian students were exposed to 32 different faces — some white, and some Asian — for three seconds each. After a brief break where they were distracted with multiplication problems, they were tested on a set of 64 pictures — 32 they had seen before, and 32 new pictures. Their job was to say which were old and which were new. As expected, the students were significantly more accurate with same-race faces compared to different-race faces. This showed that with brief exposure, different-race faces aren’t recognized as well as same-race faces.
Next, a new group of 23 students was trained extensively to recognize 6 pictures: three white faces, and three dogs. The pictures were only subtly different, as you can see here:
The students were required to be highly accurate — nearly 90 percent correct — for both faces and dogs — this took hundreds of trials. Even when the pictures were shown upside-down, respondents had to be able to accurately identify each face and dog.
Next the test was made more difficult. Viewers had to focus on the center of a computer screen, and images were flashed for 150 milliseconds, again either upright or inverted, at varying distances from the center of the screen. Since the pictures were flashed faster than eyes can react and refocus on a new spot, viewers had to rely solely on peripheral vision to identify the pictures. Here are the results:
So after extensive training, there’s a big difference in the results for faces and dogs: With upright faces, viewers were significantly more accurate than inverted faces, but with dogs the situation was reversed and viewers were actually slightly (but significantly) more accurate recalling inverted dogs. This shows that there’s something different about how we learn faces than other objects.
In the final experiment, 48 white students were trained to recognize four “friends,” selected from the initial set of 64 pictures in the first experiment. The friends matched the gender of the viewer, and each viewer’s friends were all the same race — either white or Asian.
The training again consumed hundreds of trials, and only when viewers were very accurate were they allowed to move on to the test phase (in fact, over half the participants were rejected because they never attained sufficient accuracy). They were tested in the same manner as experiment two, with images displayed at varying differences from the center of the screen. Here are the results:
This time, there was no significant difference in the results for white faces and Asian faces. In both cases, when the faces were displayed upside-down, accuracy was significantly lower than for upright faces. The results for Asian faces were similar to the results for same-race faces in experiments 1 and 2, and different from the results for Asian faces and dogs in those experiments.
In other words, memory for different-race faces can be trained to work in the same way it does for same-race faces, even in a difficult peripheral-vision test, in a relatively short period of time. It doesn’t take years of immersion in a foreign culture, just an hour or so studying pictures (albeit hundreds and hundreds of them!).
This suggests that humans have a general pattern for recognizing faces that is adaptable even to unfamiliar faces. McKone et al. argue that we recognize same-race faces holistically, instead of feature by feature. Initially when we see a different-race face, we attempt to remember it using individual features, much the same we remember a animal or other object. But after some training, we learn to recognize even different-race faces holistically, which can be more accurate, but which doesn’t work as well when faces are upside-down.
McKone, E., Brewer, J.L., MacPherson, S., Rhodes, G., Hayward, W.G. (2007). Familiar other-race faces show normal holistic processing and are robust to perceptual stress. Perception, 36(2), 224-248. DOI: 10.1068/p5499