"They all look alike": Understanding the "other race effect" | Hacker News Reader

"They all look alike": Understanding the "other race effect"

arstechnica.com

28 pointsevo_920 comments

I remember looking at a high-school yearbook from the beehive hairdo era and remarking to the owner, Geez, you all really wanted to look just alike. I went to high school in a day and place where the girls grew their hair long and left it down, and in looking at that yearbook, all I could see was the beehives.

In fact, at a second look, of course the faces were quite distinguishable. No doubt they were more readily distinguishable then and there because varying hairdos did not distract one from the faces.

There seems to be many factors besides race that could be the actual cause of this, such as that people of one race are most likely to be around people of the same race, and people's family is from the same race. Having the hypothesis that race is the cause of this implies this is a genetic phenomenon, whereas it seems more likely an environmental one to me.

There is no implication at all that this is a genetic phenomenon. The research simply states that we are better at recognizing people within our race than outside of our race. From the point of the experiment both genetics and "what we're used to" are equally valid ways of explaining this effect. And neither the article nor the research attempted to put forth either.

I think it would be interesting to repeat the study with people of the same ethnicity but different developmental backgrounds. IE. East Asians that in east Asia and Asian-Americans. Presumably, there is a large developmental component. Anecdotally, my cousins in China have a significantly harder time distinguishing between white people than I do.

Scratch my earlier comment. As found in the article:

" The scientific literature has provided clear evidence that the ORE [other race effect] and the popular belief that OR faces all look alike are not accounted for by the paucity of anthropometric variations in OR [other race] faces, but by a genuine lack of expertise. Although this theoretical explanation has been supported by numerous behavioral (for a review, see ref. 2), computational (e.g., refs. 3–5) and neuroimaging (6–15) studies on the ORE... "

so yes, it is because we are used to seeing faces from our own race. And no, the researchers are not implying this is a genetic phenomenon.

>such as that people of one race are most likely to be around people of the same race

Not to mention that that likelyhood can be (much) more so than any individual would prefer:

http://web.mit.edu/rajsingh/www/lab/alife/schelling.html http://ccl.northwestern.edu/netlogo/models/Segregation

There's a fair amount of research that suggests that the FFA (fusiform face area, where we supposedly process faces) is really just specialized for processing objects that we're very familiar with. For example, bird-watchers' FFAs light up more when they're differentiating between birds, and I think they have increased repetition suppression too.

Anyway, the theory behind that research is that we process faces differently because we're all very familiar with them, so it would make a lot of sense if this other-race effect is purely because Western Caucasians see Western Caucasian faces more and East Asians see East Asian faces more.

I'd like to see the study done with people who are living in a country where their race is a tiny minority.

I'd like to see the study done with people who are living in a country where their race is a tiny minority.

You mean like people of East Asian descent in Finland? ;)

I joke, but I'm also making a point. That's valuable if that's what they did, but we don't actually know if that's what they did; we don't know where their subjects grew up, not just their nationality but more importantly their immediate environment. We have no idea. The researchers make no attempt to detail those factors.

Why? Wouldn't somebody with a PhD doing a study on racial anything want to a) know those things and b) tell us about them? a) Yes. b) If and only if it supports their already-formed conclusion.

That's a really good point. This may be telling us something more general about what the cerebral cortex is doing, too.

For example, when people learn a new language that involves phonemes very different from those of their native tongue, they often go through a phase where not only do they have difficulty producing the foreign sounds but have difficulty distinguishing sounds that would seem very different to a native speaker. For example, R and L are perceived as very similar to a native japanese speaker and the nasal vowels of french and portuguese are notoriously difficult for english speakers register.

It could be that Wernicke's area (were language is supposedly processed) is really just specialized for processing complex sounds that we are very familiar with. So you'd expect a bird-watcher specialization effect from ... opera goers? safe crackers? sonar operators?

When I (caucasian) moved first to Japan for six months I had a difficult time telling Japanese people apart. By the time I had been there for 4 months this effect was significantly less. In addition I remember being startled to realize that people looked "less japanese" to me.

I know exactly what you mean. I have spend enough years interacting with people from Turkey that I no longer recognize any kind of Turkish accent.

Somewhat related: you can quiz yourself on whether you can tell apart East Asians' ethnicity by looking at them.

http://alllooksame.com/exam_room.php

I scored no better than random for face recognition, though some of the other quizzes had fragments of text that gave the answer away.

Does the facial recognition signal N170 generalise/apply to other species too? Dog faces? Cat faces?

The article took a 2nd reading to grok, but the experiment shows that N170 recognizes the same face, but only if that face is the same ethnicity. Meaning it is very specific. Presumably, if it can't recognize the same face of another ethnicity, it surely can't recognize the same of another species.

Simple explanation: we don't know how to distinguish between faces "out of the box", we have to be trained to do it. Most of our training happens within our own race, so we become very good at distinguishing faces.

When presented with completely different type of faces, we have to work hard to tell two faces apart, we can't just distinguish them on first sight.

It's like knowing to tell the alphabet from your own language vs some other language you're encountering for the first time.

We have specialized circuitry to process faces. People are congenitally deficient in this regard suffer from prosopagnosia (in the same way that some people with normal hearing are tone deaf).

That said, you're not completely off track. It is very likely that said circuitry becomes better at disinguishing the kind of faces it has been initially trained with.

A nice complement for this study would be to test mixed race and adopted people.

I don't have any studies on hand, but I specifically remember a lecture I attended where it was noted that infants do not have specialized circuitry to process faces. Their affinity for their caretakers, at least at a very early age, is actually not a result of recognizing people, but of the caretakers subconsciously learning how to better provoke an affectionate response.

You also saw this effect with the emotional robot Kismet at MIT. Its reactions were totally hardwired, but it managed to fool people into believing that it "liked" them.

This seems to me a simple instance of the general fact that, for telling things apart, you only observe as deeply as you need to, and it takes a while to get used to observing deeper.

For example, suppose I learn the phone numbers of my neighbors. Then I meet a friend who lives across town, and his phone number starts with 322-, while all my neighbors' numbers start with 494-, so when I see a caller ID beginning with 322-, then I immediately know it's him. But then I make several more across-town friends, and holy crap now there are four numbers beginning with 322-, so now I learn to pay attention to all seven digits.

Then I meet someone at a summer camp, and his phone number is in a whole freakin' new area code, so if I see "(408)" then I know it's him. But then if I meet more people in 408, I'll have to learn to look at the rest of the number when it begins with 408.

Or, think of backpacks. I have the only black backpack in my class; that makes it easy for me to tell mine apart. But now someone else gets a black backpack, and I say to myself, hmm, well, mine's got thick straps. Then maybe someone gets the exact same model of backpack, and I'm like, dammit, fine, I'll write my name on mine.

Cars. Mine is a big white car. That worked fine for a few years. Then I saw other big white cars. Mine is a Toyota minivan. Worked fine for some more years. Then I saw another white Toyota minivan, but it had different hubcaps. Fine. Now I've seen a few that are the same model and color of car with same hubcaps; gotta look at the license plate to be sure.

With people, there are many ways to distinguish people. Male/female. Long hair, frizzy hair, buzzcut. Really tall, really short. Color of skin. Color of hair. Facial features common to certain races. Fat. Always wears a particular hat or jacket.

At any time, people's methods of telling people apart are likely not much more sophisticated than they need to be, and if they're thrust into a situation in which those methods don't work well, then obviously they'll have trouble telling people apart. And, given time, they will learn and use new methods and get better at it.

So: There's nothing special about race, there's nothing special about faces, and there's nothing special about finding it hard to distinguish things in ways that you're not used to.

This is absolutely what is going on.

My brother had an interesting experience with this. He grew up with Caucasian faces. Then moved to Taiwan. After several years there he came back to Canada. What he quickly realized is that he had learned how to tell Chinese people apart, but had forgotten how to tell Caucasians apart. Given that he actually remembered a lot of Caucasians, a large fraction of the population looked like (but wasn't) someone he knew.

He found this very confusing.

Great summary, and what you're describing is a good practical/social example of information theory[1], where the entropy[2] is increasing with the unusualness of the phone number/backpack/whathaveyou, causing the efficiency of your ability to recognize the widget in question to decrease.

[1] http://en.wikipedia.org/wiki/Information_theory

[2] http://en.wikipedia.org/wiki/Entropy_(information_theory)

Reply on news.ycombinator.com