Time to respond to Eric Turkheimer’s post.
In my review of Carl Zimmer’s book, I said that the existing fixation index values between continental races are compatible with very significant physical and behavioral differences between those races, because wolves and coyotes have exactly the same value of Fst as Yorubas and Mormons. Lewontin said that two groups with that Fst couldn’t be very different, yet coyotes and wolves are very different – so Lewontin was wrong. Turkheimer seems to agree with me on this, although he probably doesn’t think Lewontin is full of shit. He should.
Turkheimer says that human behavioral differences are highly polygenic: true enough. And he says that makes them very significantly different from single gene effects – ” big mechanistic single gene effects and complex polygenicity” . He talks about how complex and non-deterministic that process is.
That’s bullshit: complexity doesn’t imply non-determinism. The closest thing to non-determinism you’re going to find outside of quantum mechanics (and to a sharp many-worlder that too is completely deterministic) is a physical system where tiny changes in initial conditions cause big changes later. Butterfly effect; the Kirkwood gaps in the asteroid belt: chaos theory. You know the story.
Is brain development in humans like that? Do little changes in initial conditions generally, or often, eventuate in big differences later? No: of course not. Big changes usually leave you dead or seriously fucked up: selection hates that. Development is robust because selection has made it so: generally small changes make no difference. There has, over a long time, been selection for reliable development of phenotypes that work, have high fitness. Obviously not always: shit happens, mutations happen.
One important example of that reliability of development : identical twins, even those raised apart, are very similar. Turkheimer says that 10% of the time, identical twins have significantly different IQs: well most of the time matters more than 10% of the time, and most of the time identical twins are very similar. Identical twins raised apart are more similar than dizygotic twins raised together. Some of them are insanely similar.
I’m going to mention some facts from agricultural genetics – not because I think that people are sheep ( although they are), but in order to make some points.
Lots of important traits in domesticated animals are highly polygenic. Milk yield in dairy cows is highly polygenic: no single allele explains a lot of the variance in milk production. Does this mean that milk production is not predictable, or that it varies in odd ways compared to traits that are monogenic or oligogenic? No, that doesn’t happen: the consequences of polygenicity that Turkheimer expects don’t actually occur. If they did, much of agriculture would come to a screeching halt. Speed in horses is highly polygenic: does that mean that development of running ability is a nondeterministic process? Is it logically impossible to know that certain breeds or horses ( say Thoroughbreds) are faster than other breeds (say Shetland ponies)? It may be logically impossible for Turkheimer, but it’s easy for everyone else. Along the same line, Kevin Mitchell, for equally mysterious reasons, has said that it’s really hard to select on highly polygenic trauts – even though we have a mathematical theory of selection that says otherwise, even though we do it every day in agricultural genetics.
Both are making use of nonexistent general principles to further a particular conclusion.
Let me tell you a real difference in a highly polygenic trait compared to a monogenic or oligogenic trait: the highly polygenic trait is generally more predictable, not less. Although quite a few alleles affect skin color, there are a few with large effects, SLC24a5 for example. Siblings will look a lot different if one has SLC24a5 and the other does not. This happens often in India. Underlying principle: central limit theorem. You’re throwing more dice.
Turkheimer also says that we don’t really know anything about polygenic differences unless we understand the mechanisms. And usually we don’t know the mechanisms: even when we know that a given allele boosts a horse’s speed, or a Guernsey’s milk production, or makes dachshunds have short legs, we usually don’t know exactly how it works. Often we haven’t the faintest. Which is why we couldn’t select for fast horses or cows that produce lots of milk- except that we could and did, hundreds of years ago, thousands of years ago. You don’t need to know how a plus allele or minus allele for trait X works to be able to [reasonably] accurately predict the consequences. Investigating mechanisms is going to be difficult in highly polygenic traits: those alleles favoring high trait value could work through a number of different mechanisms. We don’t know the mechanisms involved in the behavior of Turkheimer’s dogs – but in that case it doesn’t bother him. In humans, not knowing mechanisms bothers him: He wants humans to be special. And hey, they are, but not in the way he would like.
Actually I don’t quite believe this of either Mitchell or Turkheimer. I think they’re trying to arguing away stuff they don’t like.
That said, although knowing the details of mechanisms is not usually essential, I can think of cases where it’s worthwhile. There is a haplotype, now common in dairy cattle, that significant boosts milk production in heterozygotes while being lethal in embryo in homozygotes. Milk production went up while calf production went down.
What about the idea that some unknown small thing might cause someone with an identical genotype to be quite different? Not impossible – but we know it doesn’t happen very often, since identical twins, even though raised apart, are on average very similar. We know that genetically more similar people have more similar IQs. A robust development process, favored by selection, does this.
A note: we know that different races have significant average differences in behavior at birth.
Turkheimer says that I think he doesn’t have the intellectual courage required t0 face unpleasant facts. I don’t know the man: I don’t know whether he has intellectual courage. A priori, it’s unlikely, because few people and fewer professors do. But I do think he’s unsound on general questions in quantitative genetics, which is what I think of Kevin Mitchell – even though Mitchell just has a perfectly sound post on transgenerational inheritance. Once is not enough.