There has been a lot of arguing about intelligence lately. I think it was prompted by a remark made in passing about Jonathan Haidt in a (generally positive) review of his new book in the New Yorker: “He has been beset by a troubling fixation on the heritability of I.Q.—a contention widely dismissed as scientific racism“
I am of two minds about this. On the one hand, it’s a cheap shot, and a bad example of thinking about intelligence (and heritability, which I’ll do next) in a shallow and dismissive way. On the other hand, Jonathan (a former colleague of mine) is himself a superficial thinker about both intelligence and heritability, as immortalized in the classic clip in which he pairs intelligence and heritability “denial” as sins of the left, against denial of evolution and climate change as sins of the right. But this post isn’t about the problems of compulsive both-sides heterodoxy, which I leave to others. Let’s talk about intelligence.
Suppose you are running an online copyediting company. You hire people to look at documents and correct errors. The most common errors are spelling mistakes, you want employees who are excellent spellers, so you put together a 25 word spelling test and give it to job applicants, assigning everyone a score from 0-25. Voila, you are a psychometrician, measuring “spelling.”
It turns out that using sets of items scored correct/incorrect to measure ability is one of the better developed statistical theories in the social sciences. In the early part of the Twentieth Century, factor analysis was developed as a way of estimating a latent variable called “spelling ability” from the matrix of positive correlations among the spelling of individual words. The key insight here is that someone who is more likely to spell one word correctly is also more likely to spell another word correctly— that is what it means to say that some people are better spellers than others. If people just misspelled words at random, there would be no point in administering spelling tests.
Hopefully this is all completely uncontroversial, but we are right on the edge of a very tricky conceptual precipice. Let’s say that instead of calling our little measure a “spelling test” we call it a “measure of spelling ability.” That seems harmless enough, but it leads someplace very tricky. Talking about “spelling ability” instead of just “spelling” suggests there are two things out there: a person’s actual ability to spell some percentage of 25 words, and a latent, unobserved “ability” that explains why some people spell better than others. So under this model, people have some latent (or worse, innate) ability to spell, maybe coded in their genes or wired in their brains, and then an observed, actual spelling score in the real world. We might imagine people with potentially amazing spelling brains who, out of poverty or laziness, never got around to working on their spelling and wound up poor spellers in the real world.
I don’t doubt that such a model is true to some extent. There probably do exist people who could be much better spellers than they actually are. But here is the thing: spelling tests give us no insight at all into this state of affairs. All spelling tests do is tell us how good people are at spelling. They don’t tell us anything about some magical latent trait called “spelling ability”. So even if there is such a thing as spelling ability in people’s brains that explains spelling differences, it doesn’t matter because we have no way of finding out about it. All we know is spelling, period. The belief in an inner ability that explains our observed ability to get correct answers to questions has a name: it is essentialism.
I think I will save the extension of this argument to IQ for tomorrow, but for now this is the way out of the superficial argument about human abilities. Look around: do you doubt that some people are better spellers than others? Of course, who could deny that? It is a perfectly reasonable scientific question to wonder why some people spell better than others, or to investigate what other traits— like performance as a proofreader— might be correlated with scores on a spelling test.
The problem with Haidt and other heterodox thinkers about ability is that they give into the essentialist temptation, which quickly leads them down the road to Bell Curve style essentialist beliefs about human ability. It is what happened to Sam Harris when he interviewed Murray. It is what happened to Jon Haidt when he made those false equivalencies between “IQ denialism” and evolution denial. It is what trips up people from the left when they insist that any reference to human ability is troublesome. Does anyone really “deny” that some people are better spellers than others? Or, for tomorrow, that some people are better at getting correct answers to questions generally? That much is obvious. The hard question is why, and there we have to resist easy essentialist answers.
I have described one of the driving forces behind essentialism— common factor theory, which leads people to confuse latent statistical entities with real biological constructs. The other driving force is “heritability”, the other thing Haidt thinks liberals are denying. If spelling ability is “heritable”, doesn’t that mean there is an inner quality passed down from parents to kids, explaining why there are all those poor spellers out there? Doesn’t the left need to stop denying that some bad spelling is coded into people’s genes? No, but I’ll get to that in the next couple of days.