05 Jul Cochran on Zimmer, and Correcting an old Misimpression
Greg Cochran has a review of Carl Zimmer's new book in Quillette. Zimmer's book is long, and (as Cochran notes; I haven't finished it) has a great deal in it, most of which appears to be of little interest to Cochran. He wants to talk about race, and though he manages to talk around the topic, race and behavior. He gets the issue wrong in an instructive way.
Cochran starts out with a standard discussion of the reality of race, along the typical, "It's a social construction," "No it's biological" axis. None of this matters much for his actual argument, and in fact all concerned appear to agree: groups of people, defined geographically or phenotypically or ancestrally or however you want, differ to one degree or another in their gene frequencies. No one argues with that. Ancestral groups also differ in their proximal or distal evolutionary histories, and although that is an interesting topic for investigation, it is pretty much irrelevant to the conversation at hand, about why groups of people differ right now in their behavioral traits. Cochran gets it right:
The question is whether two groups are significantly, innately different in body or behavior, not how many genetic differences exist between them. We're not really concerned with the number of genetic differences, but with the consequences of those differences.
He then says something else that I agree with, which in fact I wrote about at length in an earlier blog post. If two groups differ on a single nucleotide (I used trisomy and Down's Syndrome in my example) that has large and invariant consequences for an outcome, it makes perfect sense to say that the two groups are (in Cochran's undefined terms) "significantly, innately" different. Once again, I agree. So, I think, does everyone else, but unfortunately that isn't what we are talking about. None of the individually heritable differences in behavior that are the real topic of interest here are determined by a single nucleotide. If behavior genetics had worked out this way it would be a very different field, and Cochran and his allies would have won the day a long time ago.
Human behavioral differences of the kind we are talking about are radically polygenic, in a way that has so far resisted any kind of mechanistic or deterministic explanation. There is an even deeper problem, which we can get to by way of Cochran's next topic, which is dog breeds. Comparing groups of humans to breeds of dogs is perhaps the laziest analogy in the history of human behavior genetics. It's what high school kids write me about when they first become interested in the subject. You could start with the fact that dogs have been systematically selected for physical and behavioral characteristics for many thousands of years; humans have not. That is part of what makes the argument so gross, but it isn't what is most important here. Yes, breeds are groups of dogs that differ in their gene frequencies. But what did Cochran say about that? Let's look at the quote again:
The question is whether two groups are significantly, innately different in body or behavior, not how many genetic differences exist between them. We're not really concerned with the number of genetic differences, but with the consequences of those differences.
So the question is, to what extent is the relationship between human behavioral differences and our genomes like the relationship between dog behavior differences and their genome? The answer is: they aren't alike at all. It's a funny argument to make, because it involves pointing out a difference between humans and other animals, in a scientific world that for many good reasons emphasizes the similarities. Humans, of course, are primates with genomes that evolved pretty much the same way as other genomes did. But the relationship between our behavioral differences and our genetic differences is very different. The behavior of dogs is "significantly, innately" (if not completely) fixed by their genomes. My two golden retrievers love to swim, like all golden retrievers, and they didn't need to be taught. Like many water dogs, they have webbed paws, an actual biological adaptation, a mechanism, for swimming. My old hound dog wouldn't jump in the water on a bet. That is part of being a dumb dog—they don't make choices about that kind of thing.
This is where the hereditarians appeal to the heritability of human behavior, which is why Cochran trots out Burt and Bouchard. (Zimmer, by the way, is perfectly well aware that human behavioral differences are heritable in the individual statistical sense.) If the heritability of intelligence is .5, doesn't that mean that humans are behaviorally constrained like dogs, just to a lesser extent? No it doesn't, and this is where the difference between big mechanistic single gene effects and complex polygenicity, and the difference between humans and dogs, become so important. The heritability of intelligence shows that within a population of people there is a correlation between how genotypically similar people are and how similar they are in measured intelligence. It tells you nothing about how you get from point A to point B, or the extent to which your polygenic genotype "significantly, innately" constrained your IQ to turn out the way it did. Our recent efforts to analyze the causal processes leading from DNA to human behavioural differences have only served to reinforce how complex and non-deterministic it all is. There is nothing, repeat nothing, like Cochran's single nucleotide underlying human behavioral differences in the normal range. And without even a hint of mechanism, there is simply nothing we can say about whether you would still be smarter than me if some large or small difference in our life circumstances had occurred, heritability be damned.
This isn't mystical special-pleading about the free will of human beings, it's just a fact of complex human development. You can work it out in a different way, playing by the usual heritability rules. Say the heritability of IQ is .5. What does the distribution of differences between identical twins look like? This is a concrete way of estimating the quantity we actually want to know: how much can IQ vary, conditional on a fixed genotype? There are complications, like measurement error which reduces heritability, and restriction of environmental variance and gene-environmental correlation, which inflate heritability—call it a wash. With a heritability of .5, the standard deviation of the non-genetic variance in IQ, the distribution of IQ within identical twin pairs is sqrt(.5*225)=10.58. I generated a normal distribution of twins with that SD and computed the absolute difference between them. Here is what you get.
Very large IQ differences are plausible for IQ with a heritability of .5. We don't really want to know the mean of this distribution (it's just a transformation of the variance), we want the plausible range. The very upper tail would be pairs in which one member got hit on the head by a brick or something, so let's take the 90th percentile as an estimate of things that might plausibly happen in the real world. The 90th percentile is 26 points. A 26 point difference for a heritable trait in people with exactly the same genome. Golden retriever swimming ability isn't going to come out like that. Sure, you could play around with the parameters, but it is almost impossible to get it down to anything resembling the 12-15 point difference that tends to be of interest, and this is between identical twins. Simple take away: human behavior is very malleable conditional on genotype.
Notice that this whole discussion has been about individuals. What about groups? I agree with Cochran: if someone found a well-understood genetic mechanism that had a deterministic effect on behavior within a close range, some IQ equivalent of webbed paws, and groups turned out to differ in that mechanism, the race-hereditarians would have what they want. But it hasn't turned out that way. What a well-intentioned hereditarian ought to be doing is searching for a mechanism of that kind, and some of them are; more power to them. I don't think they will be successful but I have no fundamental problem with the effort. That's science. What hereditarians shouldn't be doing is trying to twist basic data about polygenic heritability into false intuitions that genetically-based group differences in behavior are somehow inevitable. They aren't. As I have said many times, if different groups of people were genetically "tuned" (Sam Harris' term) to behave in certain ways, wouldn't you think we would have some examples by now? Maybe not for IQ, for which moral-panicking liberals like me are holding back the inevitable progress of science, but for some other polygenic behavioral trait, innocent and uncontroversial? "Oh yes, we know that the Japanese tendency to be introverted is based in their genome, but Latin American love of salsa dancing turns out to be environmental." But there is nothing, not one single thing. That is because: (a) There is no coherent theory on which to base such distinctions, other than the ever-hopeful appeal to imagined big-gene mechanistic causes, and (b) When an actual human adaptation, with an accompanying mechanism, is actually discovered, it is uncontroversial.
Using the word "moral" in the previous paragraph reminds me to address something I have wanted to write about for a while. Cochran will bring it up if I don't. Ten years ago, I was invited to the Cato blog to write about the genetics of IQ, in response to a target piece by James Flynn. The conversation, as these conversations will, drifted to group differences and I wrote a post about it, echoing many of the ideas I have expressed here. I was discussing the differences between offensive racial stereotypes and scientific hypotheses about groups, and I said:
Why don't we accept racial stereotypes as reasonable hypotheses, okay to consider until they have been scientifically proven false? They are offensive precisely because they violate our intuition about the balance between innateness and self-determination of the moral and cultural qualities of human beings. No reasonable person would be offended by the observation that African people have curlier hair than the Chinese, notwithstanding the possibility of some future environment in which it is no longer true. But we can recognize a contention that Chinese people are genetically predisposed to be better table tennis players than Africans as silly, and the contention that they are smarter than Africans as ugly, because it is a matter of ethical principle that individual and cultural accomplishment is not tied to the genes in the same way as the appearance of our hair.
My phrase about ethical principle in the last sentence has been widely ridiculed by the hereditarians, Cochran in particular. Although in reading the whole piece I think it is fairly obvious what I meant, I won't attribute it to taking me out of context, because I didn't express my point very clearly, making it too easy to suggest that I was arguing that scientists get to make ethical judgements about which findings they do and don't accept. For the record: I don't think that, so the hereditarians can officially stop saying that I do.
What did I mean? I meant– agreeing with Cochran– that our ethical evaluation of human differences depends not on the fact that they are correlated with genes (they are heritable), but instead on HOW they are related to genes. Characterizing Zimmer's view by way of an argument made by Richard Lewontin, Cochran says,
According to Lewontin, groups we call races only differ in superficial characteristics like "nose, lip and eye shapes, skin color, hair form and quantity" — and height, and disease susceptibility, the immune system, skeleton, fast-twitch muscles, cranial capacity, bone density, alcohol metabolism, fat metabolism, and so on and so on and so on.
Cochran's argument veers off after this but he wants us to think that the Lewontin view is foolish. If groups can differ in those superficial characteristics, why can't they differ in behaviors as well? This is the very core of the hereditarian argument. My answer is above: the problem is that the relationship between genotype and "nose, lip and eye shapes" on the one hand, and extraversion and IQ and criminality on the other, are fundamentally different.
And that difference has ethical content (implications?). Think for a minute: why do we hold someone responsible for their criminal behavior, but not for the texture of their hair? Hair texture is determined by our genes; we don't have any choice about it, and groups of people with a certain kind of hair just are the way they are. There is no ethical content to hair texture, no hope that things will be different someday. On the hereditarian view, IQ as a heritable trait is just the same, at maybe a slightly lower level of heritability. That's the way people are born, there is nothing you can do about it. And because everything is heritable, blunt hereditarianism leads to an obviously false view in which everything, including all of the innumerable behavioral differences among various groups of people– are just a reflection of the way people are born. Hair texture and complex behavior are both heritable, but the causation that underlies the heritability of hair texture is not like the heritability of behaviour, for which the human central nervous system, the greatest engine of anti-determinism ever designed by evolution, is interposed between genotype and phenotype.
So these distinctions are "ethical" not because scientists get to reject conclusions they don't like, but because complex self-determination is a crucial component of an ethical evaluation of behavior. It is why the universal heritability of human differences does not compel genetic determinism. My contention is that self-determination can and does co-exist with a considerable amount of behavioral heritability. Genes are correlated with the IQs of individuals, but they don't determine the IQs of individuals. Groups differ in their gene frequencies, they differ in their behavior, but the gene differences don't determine the behavioral differences.
Cochran is infuriating because he doesn't feel the need to argue his case on the merits. He assumes that the facts are obvious to everyone, but only he has the intellectual courage to face them:
The real problem with this book is that, to Zimmer and many other people, genetics itself is the enemy. The facts, not the discipline, particularly in how they apply to humans. We now know that everything is heritable, to varying degrees — and the more that life is determined or influenced by genetics, the less blank the slate, the less that can be accomplished by egalitarian social policies (or by aristocratic social policies, for that matter). The facts of genetics are caltrops on the road to a 'just' society. Zimmer is moderately fair-minded, usually mentioning both criticism of genetic claims and the response to that criticism — but he still gives the impression of wishing these claims had never been made and dislikes scientists who discovered unpleasant truths.
This is theoretical slop. Notice the "determined or influenced by". What's the difference exactly? Is height determined by genes or influenced by them? Is it surprising that the heights of many populations have increased by more than three inches in the last 100 years? Intelligence: determined or influenced, and how does the Flynn effect fit in? Divorce: determined or influenced? Does the heritability of marital status mean that we should give up on social policies relevant to marriage?
It is not Carl Zimmer who lets his policy preferences color his view of science. It's Cochran, and the fact that his view of the world is one that motivates him to put the word just in scare quotes only makes it worse,
No comments:
Post a Comment