Deconstructing “race science”

After I posted a quote from Ta Nehesi-Coates about The Bell Curve, a poster asked (paraphrased): OK, so that study is flawed, but how do we know that the various races are intellectually equal?

My answer is basically that “race” doesn’t correspond to any systematic genetic difference that might mean anything when compared to measured intelligence, assuming we could even define and measure intelligence. But I feel like that answer needs more unpacking and detail around it, so here goes.

Let me first say that I’m going to use the term “race science” to refer to all the purportedly scientific studies showing that white people are smarter than black people, Asian people are smarter than white people, and so on. The most popular recent example of “race science” is The Bell Curve, which still gets trotted out by people who ought to know better, but the theories go back centuries and I’m going to aim to explain why they’re all garbage.

So: The idea behind “race science” is that if we look at people of different races, we can measure differences in average intelligence for the different races — that some races tend to be smarter than others. Furthermore, “race science” holds that these apparent intelligence differences, measured using IQ tests or some similar tool, are because of racial genetics.

First of all, the scientific consensus is that this view is bullshit. As a specially convened task force of the American Psychological Association politely put it in response to the controversy over The Bell Curve:

It is sometimes suggested that the Black/White differential in psychometric intelligence is partly due to genetic differences (Jensen, 1972). There is not much direct evidence on this point, but what little there is fails to support the genetic hypothesis.

They then give a couple of examples of controlled studies prior to 1996 which showed no genetic link between skin color and intelligence.

That term “controlled studies” is key. It’s very hard to come up with a good controlled study on whether skin color and intelligence are linked — that is, a study where all the other possible factors are controlled for, and the populations whose IQ is measured only differ by race. And as the APA document notes, when people have managed to come up with such a study, the results haven’t supported “race science”.

But hey, let’s flog this dead horse some more, as people seem determined to try to revive it.

It’s common for people to further criticize “race science” based on the fact that IQ tests don’t really measure intelligence. There are many reasons why identifying IQ test scores with actual intelligence is problematic. I’d certainly recommend reading about those, and the Wikipedia article on The Bell Curve has a good summary. But I think that focusing on the weaknesses of IQ as a purported measure of intelligence is missing the point, as there are much bigger reasons why “race science” is bullshit.

Let’s start with the biggest single reason why “race science” is bullshit:

“Race” is bullshit.

Curiously, Penn and Teller never recorded that episode, but PBS has a nice list of 10 things everyone should know about race. The second one is that race has no genetic basis. That is, you can’t take the genome of a person and use that to assign them to a particular race. Race simply isn’t defined that way. Rather, the thing we call “race” is an arbitrary social construction. Not only does it not reflect the underlying genetics, it also varies depending where you are in the world. Your race is basically what the people around you decide it is by looking at your appearance and behavior, influenced somewhat by what you decide you want to identify as.

A person of Mexican descent who lives in England might be perceived as white. In Texas, they definitely won’t. When I moved to Texas I had to get used to the fact that a whole class of people I considered white, suddenly weren’t considered white by everyone else.

Ta Nehesi-Coates has talked to Professor Neil Risch about his paper Characterizing the admixed African ancestry of African Americans, which looked at the genetics of black African-Americans:

From cluster analysis, we found that all the African Americans are admixed in their African components of ancestry, with the majority contributions being from West and West-Central Africa, and only modest variation in these African-ancestry proportions among individuals. Furthermore, by principal components analysis, we found little evidence of genetic structure within the African component of ancestry in African Americans.

Quoting the background info from the paper:

For example, the largest study of African Americans to date, based on autosomal short tandem repeat (STR) markers, found an average of 14% European ancestry with a standard deviation of approximately 10%, and a range of near 0 to 65%

So just looking at African-Americans, i.e. those who live in America and have black skin and believe that their recent ancestors are from Africa, you still see a range of anywhere from 0% to 65% European genetic markers. So you can’t map race onto genetic markers in any straightforward way. Stand me next to a white American and a black American, and I might be genetically closer to the black American, even though my skin color is closer to the white American.

This isn’t some new revelation. UNESCO issued a position statement back in 1950 pointing out that race was not a biological fact, but an artificial social construct.

But hey, for the sake of argument, let’s pretend that race is just a matter of skin color, as coded for by genes. Let’s continue to pretend that IQ test scores measured intelligence. Is it true that skin color has some correlation with IQ test score?

Well, no. Again, as the APA document points out, when people have managed to find unusual populations that have differences in skin color alone, they’ve not found IQ test differences. And if you pause to think about it, why on earth would you expect them to?

Black skin is exactly the same as white skin, except that because of variation in a selection of genes, the amount of particular types of melanin in the skin is different. Eye color and hair color work the same way.

There are about 100 genes which code for some part of the melanin cycle. There’s one key mutation of one key gene which accounts for the largest part of the skin color difference between Africans and Europeans, but even that only accounts for around a third of the color variation. We’ve known for decades that there are actually many different ways skin can end up “black”:

The Negritos of the islands of Luzon and Mindanao in the Philippines, for instance, superficially resemble other dark-skinned groups in Africa and Australia. Yet their overall genetic affinities turn out to be far stronger to the lighter-skinned Asian peoples who surround them.

So if you measure IQ of “black” and “white” populations, you’re not even measuring populations that have anything genetically in common. The “black” group will include people who are genetically more “white” than many people in the “white” group. So “race science” is literally garbage data in, garbage answer out.

Creative Commons License Robert Taylor via Compfight

But let’s ignore even that problem, and imagine we are studying genetically similar people of European descent, and genetically similar people of African descent — so we’re looking at just that one key gene mutation mentioned earlier. Let’s imagine our subjects all grew up in the same cultural environment, so we can ignore all the problems with the cultural biases of the IQ test. What about then — might we expect to measure a meaningful IQ difference?

Well, skin color is a matter of small variations in genes. However, the melanin in the brain which we know to be important is a different kind to the melanin in the skin, eyes and hair. While there is eumelanin (the skin color type) in the brain, we don’t have any evidence that it serves any purpose. (As an aside, this means that the theory that black people are innately superior because of their increased melanin levels — a kind of reverse “race science” — doesn’t hold up either.)

So asking whether black skin leads to a genetic predisposition to lower intelligence is really just like asking whether blonde women are naturally more stupid, or whether gingers are naturally weak and overweight. It’s like calling for scientific research on whether green eyes mean you’re better at math.

But in fact, the genetics of human hair color have been teased out quite precisely:

Further study showed that the region of human DNA that contained the single nucleotide change associated with blondness specifically affected the expression of KITLG only in hair follicles. […]

(Emphasis mine.)

“It’s clear that this hair color change is occurring through a regulatory mechanism that operates only in the hair. This isn’t something that also affects other traits, like intelligence or personality. The change that causes blond hair is, literally, only skin deep.”

In other words, melanin in the brain and melanin in the hair aren’t even affected by the same gene mutations. The mutation that changes hair melanin levels has literally zero effect on brain melanin. So there’s simply no plausible mechanism by which the blonde hair mutation would affect intelligence. There’s no reason to believe that skin color should be any different.

If you want further reading, PBS has links to articles. I’d also suggest reading about Theodore W. Allen’s book “The Invention of the White Race”, which sets out the history of how race — and in particular, the “white race” — was an invention of 17th Century American plantation owners, who used it to shore up slavery by creating a new division between poor white laborers and black slave laborers.

White Squirrel Looking Alert
Creative Commons License Adrian Wallett via Compfight

In summary: Race is an invented social construct that is used to distract the disenfranchised from class issues. It doesn’t correspond to genetics or to skin color. Hence studies showing that some races are genetically predisposed to higher intelligence are garbage. You can’t even get as far as discussing nature vs nurture, because you haven’t grouped by anything that corresponds to a feature of nature.

At this point you might be asking: OK, so if the differences in The Bell Curve aren’t attributable to race, or even to skin color, what causes them? The answer seems to be that we don’t yet know. There are a lot of possibilities. Ultimately, I suspect that the data are just too noisy to let us analyze them and get a clean answer. But science can tell us a few things that aren’t plausible explanations…

Hopefully soon we can move on to more serious questions, like whether gingers have souls, or whether all humans having souls is just liberal dogma.

7 thoughts on “Deconstructing “race science”

  1. I sense that what a lot of racist views actually condense down to is: “It is self-evident that skin colour is determined in large part genetically. We can identify a lightness threshold for skin colour such that there is a significant difference in average innate mental traits, in particular intelligence, between those with lighter skin and those with darker.”

    We can argue until we’re blue in the face that skin lightness is determined by a very small set of genetic markers that “obviously” doesn’t affect much else. But that doesn’t demonstrate a lack of correlation between those genetic markers and any others. People can point to sickle-cell anaemia, for example, where incidence DOES correlate with skin lightness. Despite all the light-skinned people of African descent and all the dark-skinned Aboriginal Australians.

    The “ten things everyone should know about race” list includes, at number four, “Most traits are inherited independently from one another. The genes influencing skin color have nothing to do with the genes influencing hair form, eye shape, blood type, musical talent, athletic ability or forms of intelligence. Knowing someone’s skin color doesn’t necessarily tell you anything else about him or her. ” which is true, of course, but nicely illustrates how such reasoning misses the point. “Doesn’t necessarily tell you anything” is not the same thing as “necessarily doesn’t tell you anything”.

    I think a big part of the problem is that the people trying to debunk race science are starting from the premise that prejudice is bad and people don’t want to do it and if we could just debunk their arguments they’d relax and stop doing it. Whereas in fact a lot of people WANT to be prejudiced. We need arguments that their prejudices are actually wrong, rather than merely harmful and not supported by evidence.

    I don’t personally experience this much in relation to race, but I certainly do in relation to gender and sexuality. Evangelical Christians have often built their world-view around one specific Biblicist interpretation of Scripture. They want to continue having that view. “Because the Bible says so” is a mantra that makes them feel holy. They won’t give that up without strong contrary evidence. A lovely guy I know, an Evangelical Anglican minister, earnestly told me that gay relationships weren’t as successful as heterosexual ones. I pointed out that such attitudes were a self-fulfilling prophecy. He agreed… but it didn’t change his mind. His reasons are post-hoc and debunking them don’t help. Any effective challenge to his values has to go to the core. (Or sidestep such reasoning entirely with arguments about God’s loving grace.)

    I’ve also experienced the converse problem in relation to transgender issues. People accustomed to debunking prejudice in such flawed terms often come to believe subconsciously that any thinking involving a correlation of one phenomenon with another is inherently flawed. For example, I know trans people who get very cross that the people who compile breast cancer statistics ignore them. But the reality is that for such purposes they are a statistically significant minority. “Identifies as female” correlates with “has female-typical breast tissue” plenty strongly enough.

    So perhaps the acid test is to find modes of argumentation against “black people are innately more likely to be criminals” that avoid also arguing against “women are innately more likely to get breast cancer”.

    1. Null hypothesis, Clive. I’ve set out why “skin color gene differences don’t code for intelligence differences” is the null hypothesis, and why it would be extraordinary if they did — and particularly extraordinary if they reproduced the intuitions of early 20th Century eugenicists. And you apparently missed that there are controlled experiments testing the hypothesis that skin color is correlated with IQ differences, and they’ve found that it isn’t. Here’s some more reading with study citations, and here’s a NYT article talking about direct vs indirect evidence.

      The reason why books like The Bell Curve seem to find correlations is that they are population studies, and not controlled direct studies. And as I attempted to explain, the populations they study aren’t genetically distinct, and are stuffed with too many additional random variables (income, parental influence, nutrition, social surroundings, etc etc) to let you tease out the skin pigment influence (if it exists) from all the others.

      1. I think you’re treating the notion of the null hypothesis in magical and invalid ways.

        The importance of the null hypothesis is that one refines the notion of what one is seeking to prove by contrasting it with a default. And “default” merely means “thing that happens if the alternative doesn’t”. There must BE a default, or your hypothesis is untestable. And it matters what the default is, because it clarifies exactly what you’re saying.

        For example, if I say “women who drink more milk are more likely to get breast cancer” it matters a great deal whether I then say “compared with all other women worldwide” or “compared with other women of similar socio-economic status and nutrition”. (A lot of malnourished women don’t live long enough to get breast cancer.) It is, of course, true that the hypothesis under consideration should be precise about such things, but insisting on also seeing the null hypothesis is a valuable tool in teasing out the subtleties.

        Beyond that, as Wikipedia notes, there is a certain amount of controversy amongst statisticians in how to handle null hypotheses. But one thing is clear: “we devised an experiment to demonstrate with statistical significance that Ha holds instead of H0 and it failed” does NOT necessarily mean H0 holds, it just means Ha does not.

        Sure, repeated failed attempts to prove things similar to Ha tend to strengthen the claim that something like H0 holds. But conversely, an awareness that the methodological problems which have thwarted attempts to prove Ha would also thwart attempts to prove H0 tend to demonstrate that we simply can’t tell.

        Of the various studies I’ve seen referred to in the various articles you’ve linked to (and I confess I only skimmed, because there were a lot) the most interesting was that mentioned in the New York Times, comparing the intelligence of early-adolescent offspring of American servicemen and German women, where some of the servicemen were black and some white.

        The trouble is this: assume that study had shown a 10-point difference in IQ. People would immediately say “that shows nothing, because it’s only comparing the offspring of black rank-and-file military with that of white rank-and-file military and only relatively less intelligent whites enter the military”. And they’d be right to raise that doubt about its validity. But it’s scientifically bankrupt to raise a methodological doubt only if the experiment’s outcome is one we don’t like; we need to recognise that evidence of equality of IQ has to be treated with the same skepticism as evidence of inequality.

        1. Correction: I said “…does NOT necessarily mean H0 holds, it just means Ha does not.” and meant “…does NOT necessarily mean H0 holds, it just means Ha hasn’t been proven.”

          (I can’t edit replies here, can I?)

        2. Sure, such an experiment would not prove anything about black populations as a whole. But it would possibly show an IQ link to skin color, which controlled experiments so far apparently have not, which is the key point.

          1. It’s A key point, but it’s not the key point I have been making.

            I’ll try to be more succinct:

            Many people have tried to show a link between race and IQ. Their efforts have been bogus, partly because of prejudice and partly because there are grave methodological difficulties.

            Conversely, few people have tried to show that there is no link. Partly because they’re content to knock holes in the attempts to show there is one, partly they have less of an axe to grind, partly they can see many of the methodological difficulties would be a problem for them, too.

            Based on the evidence I’ve seen, I’d say it was unscientific to claim a link between race and intelligence (or any of various other similar factors). But I’d also say that, though more scientific, it’s also pretty suspect to claim there’s no link. “We suspect there’s no link, but really we have no way of telling.” may be the more honest appraisal.

            But more than this, I’m making the point that this is all post-hoc argument. Some people want equality of opportunity, for everyone to reach their best potential. And some people don’t. So far as I can see, racist white people don’t WANT black people to achieve parity with whites and that skews the entire agenda. In turn, it affects what modes of argumentation can be effective with them.

            I wonder: who are you trying to convince, and how do you expect to succeed?

          2. I’d also say that, though more scientific, it’s also pretty suspect to claim there’s no link.

            Well, I guess that’s the problem here. You seem to be saying that even though we have a bunch of science showing no link, and no rigorous controlled science showing a link, you want to carry on believing there might be a link anyway. Whereas I’m taking a scientific perspective, and saying no, we go back to the null hypothesis and take that as fact until someone rigorously proves a link.

            But more than this, I’m making the point that this is all post-hoc argument.

            Since your position is that it’s post-hoc on both sides of the argument, so what?

            I wonder: who are you trying to convince, and how do you expect to succeed?

            I was trying to answer your initial question, “so how do we KNOW people of one race are, on average as intelligent as those of another”. So I’ve been attempting to address that question of whether there might plausibly be some genetic link between skin color and intelligence. I’ve tried to explain why it’s incredibly unlikely, given what we know about genetics, and that the science doesn’t seem to support it, meaning we have to go back to the null hypothesis that skin color is not linked to intelligence. However, you seem to have adopted the hand-wavey “you can’t prove a negative” view, which makes me wonder if the whole invisible sky wizard thing hasn’t had a rather deleterious effect on your rationality.

Comments are closed.