09 Apr Heritability and Malleability in Individuals and Groups
Heritability and Malleability
There has been some interesting discussion on Twitter of a very old topic, the relationship between heritability and malleability. The standard thing to say here is that there is no relationship between the two– the fact that an individual difference is heritable is not the same as saying it isn’t malleable. There are a couple of standard examples. The whole business is more complicated than is usually acknowledged.
Example 1: Phenylketonuria
The most familiar is phenylketonuria, an autosomal recessive disease that results in an inability to digest phenylalanine. If babies born with this condition continue to eat foods containing phenylalanine, they suffer permanent brain damage and eventually die. However, if they avoid such foods they develop normally. So PKU is a straightforwardly genetic disorder, the treatment for which is completely environmental. Would it be correct to say that PKU is a genetic cause of low IQ? In a world without dietary treatments, yes; in a world with them no. But that isn’t really a problem, no one is going to lose any sleep about whether the effects of PKU are or aren’t genetic, exactly because we understand the mechanism. We can tell the whole story: left unattended it causes brain damage, but in a world with the proper dietary amendments it does not.
Now consider two islands. Island One has a very high incidence of the PKU gene, and on Island Two the genes don’t occur at all. No one knows about the nutritional mechanism or the dietary treatment, so Island One is afflicted by the terrible consequences of untreated PKU. The average IQ on the island is very low, much lower than on Island Two. It seems perfectly reasonable to say that the IQ difference is caused by the genetic difference.
But then someone figures out the mechanism of PKU, establishes the proper dietary regime, and over a generation or two the mean IQ on Island One reaches the same level as Island Two. What happened to the genetic effect on the group difference? It went away, just like it did on the individual level, but as was the case above it doesn’t really matter whether describe the situation as genetic or environmental. We know the story. The gene caused low IQs on Island One under one set of conditions, but not under the other. Questions of whether people on Island One are genetically predisposed to have low IQs don’t matter, because we can tell a story about the mechanism. We understand what is going on and don’t have to mess around with statistics and percentages. It’s like the red-haired child story about IQ, in which red-hair genes only cause low IQ in a world in which red haired kids get abused. The difference is that PKU isn’t a thought experiment– that’s the way the world really works, and we understand it.
My main point here, which will continue throughout this series of posts, is that the group difference is a relatively unimportant part of this story. Once we understand the causal effect of the gene on individual people (which means, understanding the cause and the dietary domain in which the cause occurs) explaining the group difference just follows along. Groups, as I have been saying for a very long time, are just collections of individuals, and if there are things that cause differences in individuals, and some groups of people have more of those causes than others, those same things are going to cause differences in the groups. The important thing is the causal effect of the gene, and the extent of the domain in which the cause operates. How the gene happens to be distributed across groups is a side effect.
Example 2: Height
Height is another standard example of individual and group differences. Before the Second World War, people in Holland were a couple of inches shorter on average than people in the US. On an individual level height was highly heritable just as it is now. Think of how tempting it must have been to assume that Dutch people were shorter than Americans for genetic reasons. It’s the basic Harris-Murray argument: come on, we know height is super-heritable, Americans and Dutch differ in their gene frequencies, it’s gotta turn out that the difference is at least partly genetic!
But of course, today Dutch people are a couple of inches taller than Americans. What happened? Presumably, as Holland went from being a rural, relatively undeveloped country before the war to a developed modern country afterwards, its diet (and probably its medical and prenatal care, etc) were vastly improved, so all of a sudden those Dutch genes that used to be associated with (nb. not caused) short people now produced tall ones.
There are a couple of crucial differences between the genetics of height and the genetics of PKU. Height is polygenic, and more or less as a consequence we know very little about the biological mechanisms that determine height. Genes obviously have something to do with it, but there is no equivalent of a PKU gene, with a known causal chain that allows us to specify the conditions under which a particular gene will have a positive or negative effect on height. So although individual-level heritability tells us, “height is correlated with genes,” in human beings we have no way of knowing how causation works, and therefore no way of saying what the domain is within which that relationship will hold. We got fooled pre-war to post-war, because we couldn’t anticipate what was going to happen in a new environment. And we still don’t know: maybe in fifty years when we are all subsisting entirely on manufactured energy bars, Americans will once again be taller than the Dutch.
I want to repeat the important issue here. The key question is not really explaining the group difference, it is explaining the causal mechanisms linking genes and environment to a phenotypic outcome in individuals. We can do that for PKU, so talking about group differences in PKU and its consequences is unproblematic. We can’t do it for height, so there it will always be difficult to say whether a height difference between two groups is environmental or genetic. When the individual causal problem is solved the group problem is easy.
When is a Group Difference Genetic?
OK, here is a definition. A group difference is genetic when there is a causal mechanism (not a heritability or a polygenic risk score) linking a gene or genes to a phenotypic outcome across a known, wide domain of contexts, and the causal gene or genes is distributed unequally across the groups.
[I’m going to say it again below, but I want to put it here as well. If you don’t like my definition, propose your own! This discussion cannot be based on casual intuitions of what it means for a group difference to be genetic in origin.]
Here are some consequences of the definition:
- It makes clear why I think it is incorrect to say any group difference we currently know about in a polygenic complex behavior is genetic: we have no mechanisms, so we have no understanding of what might happen in a new environmental domain.
- As a corollary it provides a meaning to the standard hereditarian contention that a racial difference is “partly genetic.” This is mostly a rhetorical move, an attempt to put anti-hereditarians on the defensive by making us defend the null hypothesis that something is absolutely zero. What it means under my definition is that across a wide range of plausible contexts the gap might get smaller but it wouldn’t completely go away. That contention has exactly the same problems (How would we know?) as the 100% hereditarian view.
- It shows why I find the contention that the black-white IQ gap is genetic somewhere in between ridiculous and offensive, and in any case wrong: the assertion entails imagining that some portion of the gap would still exist if slavery and Jim Crow had never happened, and if there was no racism in the modern world.
- I have written before about a group difference in behavior that my definition suggests is genetic in origin: populations with the lactose persistence gene drink more milk than populations without it. Why is it genetic? There is a mechanism. Just like PKU, the effect of the gene is modifiable environmentally, but it doesn’t matter, because we get it. Some adults can digest lactose and others can’t, and those who can’t can take a lactase pill instead. You don’t need to compute any heritability statistics, and you don’t need to wonder what would happen if people with the gene were raised in Antarctica or on the moon. A mechanistic understanding of something means that you know how it will perform in environments that haven’t been observed yet.
- It’s interesting that the only good examples of genetic mechanisms, and thus of genetic group differences, involve single genes. It seems as though there ought to be examples of polygenic mechanisms, but there aren’t, at least not for human behavior. This is another way of making the challenge I always make to name a polygenic behavioral difference between groups that has turned out to be genetic: there isn’t a good one. I think you could tell a story about the modern course of behavioral science in these terms, eg, why do we never finally wind up with a meaningful biological theory of schizophrenia? Why does the endophenotype dream never come true? Another blog post.
- It answers a question I get asked a lot: what would it take to change my mind about the IQ gap? It would take a mechanism, a good causal story between a set of genes, a brain structure and IQ. Such a mechanism might or might not explain the particular group difference that interests you (this is the Lewontin argument about plants in good and bad soil, by the way) but it would explain SOME group difference. Put the people with the mechanism on Island One, and people without it on Island Two.
- It explains a lot about the arguments constantly put forward by the hereditarians. They always say:
- We have tried really hard to make the gap go away, yet it has persisted. This is a way of establishing that the genetic effect obtains across a wide range of contexts. But it also shows how hopeless it is: the one new context we really need to observe, the one where slavery didn’t happen and the effects of racism have worn away, is impossible to observe.
- Any day now we are going to have a biogenetic mechanism that explains everything! Here is Charles Murray in the Harris podcast: We are also within a matter of years—I don’t think that many years—before we will understand the functioning of intelligence down at the level of alleles and [SNPs]. We are already making a lot of progress on that. I propose a rule. I will stop declaring such advances to be impossible on philosophical grounds, and the hereditarians should refrain from basing their arguments on scientific developments that haven’t happened yet.
- If you don’t like by definition of what it means for a group difference to have a genetic basis, propose your own. In my opinion the worst thing about the hereditarian position is that it is based on a vague intuition about what genetic explanation means, which allows them to throw around statistics that look good but aren’t in any way decisive. There are only two other candidates out there that I am aware of. Emil Kirkegaard, who proposes what I call an analysis of covariance approach: If the genetic correlation within groups is sufficient to “explain” (in the weak statistical sense of explain) the difference between groups, then the difference is genetic. This is a very bad suggestion that will take another blog post to rebut. The other is the recent work on polygenic selection, which is very interesting biologically but a long long way from being an explanation of contemporary differences in behavior, as is emphasized by the people conducting it.