»  Taki's Magazine

March 21st, 2013

  Genes and Smarts


A couple of months ago here on TakiMag I reviewed responses to geek website Edge.org's Annual Question. The 2013 question was: What should we be worried about? The leadoff answer, I noted, was from evolutionary psychologist Geoffrey Miller, who thought we should be worried about Chinese eugenics. I expressed some skepticism:

I have no doubt that the ChiComs will indeed steam ahead on this. Steaming ahead is one thing, though; attaining actual success is another. The Titanic was steaming ahead pretty confidently there for a while.

Leaving aside my skepticism, which is still intact, and leaving iceberg-avoidance prospects for another time, what exactly are the ChiComs up to?

There is no short answer to that question. This is just really complicated stuff. The accounts given in the better-quality U.S. news outlets aren't bad, but if you want to understand what's going on, you have to invest a little time in getting to grips with the fundamentals.

How much time is "a little"? Well, 1h33m32s has to be close to the minimum. That's the length of this excellent presentation given by Steve Hsu at Michigan State University last month, title: "On the Genetic Architecture of Intelligence and Other Quantitative Traits." You can even shave five minutes off that: the emcee (not identified, but looks like cognitive-science boffin Taosheng Liu) makes a lengthy meal of introducing Steve.

Having already told you that nothing less than watching the whole thing will do, I can't claim that the following notes are a substitute. I just want to point up some highlights and offer a helping hand to the math-challenged.

That word "quantitative" is a good start. A trait is quantitative if it can vary smoothly across a range, like height; as opposed to a yes-no trait like earwax texture (sticky or crumbly) or hairiness of elbows. Height is in fact a handy notion to help you remember what "quantitative" means here. And as Steve says at 11m40s, if the idea that some people are intrinsically smarter than others is intolerable to you — as it is to most well-socialized early-21st-century Americans — you can just substitute "height" for "intelligence" in the rest of the lecture, because intelligence is a quantitative trait, too.

The diagram at 12m22s (slide 3 here) is very informative. The collection of all your genes is your genome; the collection of all your traits is your phenome — how you appear to the world (from Greek phainein, to appear). When, one day, we are possessed of perfect knowledge, we shall understand how our genome shapes our brain structure, how our brain structure shapes our mentation, and how our mentation shapes our abilities, which are part of our phenome.

We are far away from that yet, but we can take a shortcut — what Steve calls "a cheap statistical hack" (12m37s) and "just a little bit of math" (1h24m15s) — from the genome to the phenome. Just assemble a large number of subjects with some interesting oddity in the phenome (e.g. they are very smart, or tall), and try to find corresponding oddities in their genomes. That's called a GWAS, a Genome-Wide Association Study.

That's what these researchers are trying to do, with intelligence as the focus of interest. Previous attempts have been dry wells. As Steve says at 39m56s: "At the genome-wide level of significance so far there are no hits." That does not mean that genes have nothing to do with smarts. We know from twin studies (30m40s and more pointedly — "It's almost all genetic!" — at 1h12m18s) that intelligence is highly heritable, so that genes must be implicated. What it does mean is that we haven't yet used big enough samples, as Steve explains very elegantly at 40m49s. That's what they're trying to do.

It's tough because there are a lot of genes involved. Earwax texture is determined by just one gene; eye color, by a dozen or so; height by hundreds. Each one of those hundreds contributes a teeny bit to shaping the trait, but the most powerful single one we know affects no more than three or four millimeters of height. Intelligence is probably like this, but more so: more genes involved, less effect per gene.

If you are either kind of creationist, or if your specialty is some pseudoscience like Economics — James Heckman gets unhorsed by Steve at 1h27m32s — you will of course scoff at all this. (For pity's sake don't send me creationist email, I beg you.) If you respect empirical inquiry, though, and don't mind a bit of math, this genes'n'smarts project is worth your attention.

And even if you're not interested in it, it's interested in you — and in your kids, your grandkids, and your civilization. Answering a question at 1h19m35s, Steve goes into the Idiocracy zone.

And now modern life has probably flipped the sign, so … everybody can reproduce now, and the smartest people seem to have the most trouble. I'm 46 and I have 7-year-old kids, and I only have two … Anybody, if they wanted to, could have five kids nowadays …

Thence (1h20m15s) to what I am going to christen Neo-Social Darwinism, a.k.a. "survival of the richest":

In economic history there's good data on wills and family records in China, medieval Germany and medieval England. You can see that in those days economic success was incredibly correlated with reproductive success … so there was very strong selective pressure even in recent history. But that's all gone now.

Ron Unz turned his plow into this field over at The American Conservative a few days ago. Ron's piece is titled: "How Social Darwinism Made Modern China: A thousand years of meritocracy shaped the Middle Kingdom." Well, maybe it did and maybe it didn't (I had things to say in a follow-up piece); just be sure you'll be seeing a lot more on these themes over the next few years. As a primer in the underlying science, an hour and a half with Steve Hsu will be time well spent.