62 Comments

It seems to me clear that the heritability estimates from twin studies are correct, and the estimates from GWAS molecular genetics much too low. Here's the argument:

1) Twin studies measure the effect of the whole genome at once. That means that a sample size of ~ 50 will give a fair result.

2) GWAS studies are attempting to identify each genetic variation that contributes to differences in intelligence, and thence measure the effect of each genetic variation, and add these up.

3) Intelligence is a hugely complex trait; necessarily, any recipe for intelligence thus has to include a huge amount of information. That can only be if it involves a huge number of genes. You can't encode large amounts of information in a small number of genes.

4) Hence, variations in intelligence likely involve thousands of genetic variations (let's say 3000) each having (on average) a ~ 1/3000th effect on the overall difference.

5) The sample size needed to detect small signals scales as the square of how small the signal is.

6) Hence, for GWAS studies to find all of the genetic contributions to intelligence you'd need a sample size of ~ 50 * (1/3000)^2. That is 500 million. That is, you'd need to sample 500 million people's genomes. Obviously that hasn't been done.

7) If your GWAS study samples 1000 genomes then you can only find genes that contribute more than sqrt(1/1000th) of the variation. That is, the genes making ~ 3% contributions (actually. it's worse than that, for decent accuracy you need to repeat ~50 times, so you need a much bigger sample)

8) And that's exactly what they find, a handful of genes that affect intelligence at the ~ 3% level.

9) And that would be fine, if intelligence were indeed a trait controlled by ~ 30 genes each having a ~3% effect. But it can't be. There's no way one can encode the necessary oodles of information in 30 genes or even 300. Necessarily, it's going to involve 1000s of genes.

10) And that's why GWAS estimates are way lower; their sample sizes are only big enough to find the tip of the iceberg -- the small number of genes that each have a substantial effect -- but completely miss the thousands of genes that each contribute a small amount.

Refutations welcome; am I missing something?

Expand full comment

I would add to your good points the following:

As far as I understand, GWAS estimates assume linear interaction between genes. Not only it has to catch thousands of genes affecting IQ, with which I strongly agree with you, there is another important problem. That is, in reality I think there is a complex linear and non-linear interaction of genes with each other with various local maximas of their cumulative effect on IQ.

From that perspective twin studies are much more reliable in that it catches not only thousands of genes but even their complex linear and non-linear interactions with each other affecting IQ.

Expand full comment

"1) Twin studies measure the effect of the whole genome at once. That means that a sample size of ~ 50 will give a fair result.

2) GWAS studies are attempting to identify each genetic variation that contributes to differences in intelligence, and thence measure the effect of each genetic variation, and add these up."

All your points are well taken, but for me, the first two are a synopsis, and the rest are details.

Expand full comment

The point is that GWAS looks for additive genetic elements. In reality, genes, regulatory elements, can be both positive and negative elements in the metabolic network, depending on its state. Intellect is thus not the additive result of the effect of slightly different genes or regulatory elements, but a combinative result. Just as in a game of poker, not only cards with figures win, since one can have a straight of cards of small value. The genetic determinism of intellect requires finding the rule first - as in poker. It would be worth using well-researched endogamous populations with very different average intellect: e.g. Ashkenazi, Icelandic, Scandinavian, European Gypsy. From these study groups, whole genome sequencing should be done, and then imposed on them by a rule-seeking self-learning artificial intelligence might find the philosopher's stone...

Expand full comment

The underlying logic of twin studies assumes the effects of different genes are linear.

If it's not, they will overestimate genetic effects.

Expand full comment

Explain to me why twin studies must assume that gene effects are linear. Don’t they just compare outcomes for different levels of genetic relatedness, without making any assumptions about how genes are operating?

Expand full comment

> Don’t they just compare outcomes for different levels of genetic relatedness

Yes, and they assume the genetic component of the effect will be linear in the genetic relatedness, which is how they seek to separate the effects of genes and shared environment.

Expand full comment

Again, explain to me why it makes any difference to twin studies whether genetic effects are linear or non-linear.

Expand full comment

Take the classic case of comparing identical and fraternal twins.

We subtract the average variance between identical twins from the average variance between fraternal twins. Assuming linearity, this cancels the effect of shared environment and half the effect of shared DNA, so by doubling the variance, we get the genetic contribution.

Of course, the above calculation assumes genes interact linearly. If they don't, when we do the subtraction above, we will have cancelled out less then half the genetic effect, thus the doubling step will result in an overestimate of the genetic contribution.

Expand full comment

The only assumption there is that the environmental differences for identical twins (separated at birth and raised apart) will be the same (on average) as for fraternal twins (separated at birth and raised apart). And that seems to me a pretty robust assumption. And given that, the subtraction is valid. I don’t see that that subtraction (once equal environmental differences is assumed) makes any assumption about whether genetic effects are linear (though I’m not fully sure what the latter even means).

Expand full comment

It's a question for a professional geneticist like Sasha, and I am certainly not one. I've read a few molecular genetics article and have never encountered this view in any of them. If you can find a geneticist who makes similar points, it would increase your credibility. You also shouldn't be blind to the reasons to suspect that twin studies might have their own problems, which I address in my article. https://open.substack.com/pub/eclecticinquiries/p/on-race-racism-iq-and-heritability?r=4952v2&utm_campaign=post&utm_medium=web

Expand full comment

I would indeed be interested in a geneticist’s response to it (I’m not one), though at root the argument is just basic statistics about the sample size needed to detect small effects. (Though I’m open to being told that I’m misunderstanding something.) I’ve read your article, and other criticisms of twin studies, though I don’t find these criticisms persuasive. For reasons as expounded by Noah Carl in the article, twin studies (along with things like adoption studies) seem about as clean and reliable as we could hope for. The ways suggested that they could be wrong seem to me far fetched, especially as they’d have to be systematically operating in multiple families.

Expand full comment

Here is a race realist video by your one-time mentor and now nemesis, Jared Taylor.

https://rumble.com/v5gm1ph-david-reich-reluctant-scientist.html

Expand full comment

why would I want to watch this?

Expand full comment

"why would I want to watch this?"

To get in touch with reality.

Expand full comment
Sep 26Liked by Aporia

Great article, your response to Mr. Gusev was quite thorough. Nevertheless this quote that you stated in relation to the “missing heritability” is the main point of contention where we “shouldn’t shift our priors nearly as much as Gusev claims.” I’m not sure the response you laid out is convincing enough to not shift our priors much more. In regard to this, I believe we should at the very least be much more cautious and have a much more probabilistic approach as it relates to the estimates of heritability. There still seems to be much we don’t understand in that regards it seems!

Expand full comment

This reads like you haven't really absorbed Sasha's post about twin studies. The papers he cites have a more complex causal model than the twin studies you seem to prefer, but they are themselves Twins Studies, and ones that seem to offer an explanation worth considering.

Even if the EEA holds, the high heritability estimates from twins studies are from models that don't incorporate assortative mating, and that effect will inflate apparent heritability given some very plausible assumptions. Some of the studies Gushev cites also incorporated data from mixed families to test the assumptions around shared environment, and those suggest that this more complex model fits the data better and that EEA doesn't hold.

The other thing I observe is that most of the arguments in this about genes mattering more than environment, pointing to eg adoption, are arguments that genes matter more than shared environment. That could be true, and we could still be overestimating heritability.

> The focus of Gusev’s first article is a handful of recent studies in molecular genetics

I would say that he focuses at least as much on the Cloninger papers, which are far from recent, the last being published in 1979

Expand full comment

First of all, thank you for engaging with Sasha Gusev's work, which has really challenged my understanding of heritability. I would suggest including assortative mating, gene-environment interaction, population stratification, homogamy, and cultural transmission in the discussion, as these factors currently strengthen Gusev's argument. As long as someone thoroughly addresses all his points and the evidence he draws upon, and can present compelling arguments against them (see his "Comments on:" posts), his position remains my working hypothesis.

Expand full comment

Thankyou for writing this. I think there has been a certain laziness about responding to Guzev's arguments from hereditarians, which, if nothing else, doesn't look so great.

Expand full comment

So when do we start seeing articles from Gusev types headlined like this: My Career Is A Lie, Genes Aren't Real"? Published in The Atlantic, of course.

Expand full comment

"The weight of evidence suggests genes matter far more than family environment."

Indeed, it does. Those who can not accept that some races and ethnic groups have higher cognitive abilities than others make the argument for nurture over nature.

Expand full comment

Indeed, current view is 80% Hereditability for IQ.

If environment was dominant, we would have seen dramatic increases in Black IQ over the last 50 years.

Expand full comment

I always think back to the grand-daddy of all such Great Society programs, “Head Start”. HS is still in effect today, and even has an “off shoot”, Early Head Start. Although extensively funded, expanded, and now approaching 3 generations of effort—primarily in the Black community—I’ve seen no studies indicating more than a transient effect that wears off in early grades after the program stops. This is the same result confirmed by our author in today’s posting.

Surely if there were an environmentally amenable factor in IQ we’d be hearing about such studies incessantly in the literature. If HS can’t do the job, what can?

Expand full comment

Even more alarmingly, on the environmental side, take a look at the heart and Risley study that explored the 30 million word gap and tween upper middle class, children and lower class children

https://www.leadersproject.org/2013/03/17/meaningful-differences-in-the-everyday-experience-of-young-american-children/

Expand full comment

You are correct. Charles Murray, discussed this in the bell curve as well as his more recent book facing reality. Turns out there’s small raise in IQ, which then sinks back to their pre-Headstart Norm.

Expand full comment

"Indeed, current view is 80% Hereditability for IQ."

I agree; that is the percentage I have agreed with for years.

Expand full comment

Given the substantial uncertainty about "missing heritability," it is perhaps most reasonable to place our prior for the "true" direct heritability of social and cognitive phenotypes somewhere between the molecular estimates and the twin estimates. I do think the picture that is materializing is that the genetic and gene-by-environment effect ratchet and this seems to inflate the twin and extended family heritabilities some. There are also some genetic effects we're missing.

Expand full comment

I've written a new article explaining why heritability estimates from both twin and adoption studies are probably hugely exaggerated. These studies can separate out the effect of shared environment, but are very ineffective at measuring other types of environmental effects that amplify small genetic differences. I focus on gene-environment correlations. If someone is born with some small genetic advantage in performing music, they tend to create and select environments that enhance that advantage. Twin/adoption studies lump this kind of environmental effect in with the effect of genetic differences. The extremely low heritability for IQ, less than 0.2, derived from GWAS merely confirms that twin/adoption studies always substantially exaggerated IQ heritability. The Flynn effect also supports my point. https://open.substack.com/pub/eclecticinquiries/p/twin-studies-exaggerate-iq-heritability?r=4952v2&utm_campaign=post&utm_medium=web

Expand full comment

You don't actually take on one pillar of his argument. Which is that the value you're trying to estimate is not identified (in the statistical sense) in the kind of studies you list.

Expand full comment

You can quibble about semantics all you want but to call it "pseudoscience" is ridiculous. You're describing different things and I don't think many hereditarians are unaware of that.

I mean, OBVIOUSLY we can imagine hypothetical environments where innate genetic abilities will not be amplified, but if we're discussing the implications in the real world, those quibbles are irrelevant.

Expand full comment
Sep 28·edited Sep 28

This is really fascinating stuff. A few observations and questions.

1. Gusev seems to focus mostly on educational attainment, which is plausibly less influenced by genetics than IQ, and is also less important for explaining social outcomes. For example, the Young (2018) study that Gusev relies on is about educational attainment, not IQ.

2. I’m struggling to reconcile the results from natural experiments on the impact of education on IQ with the randomized controlled trials showing massive fade-out. Furthermore, I’m not aware of any evidence that parenting explains a large amount of the variance in adult IQ scores.

If genetics are not explaining most of the variance, what is? This seems like a big mystery for the environmentalist camp. How does the Gusev camp address this issue? Why aren’t the RCTs generating sustainable IQ gains if education can durably increase IQ? If parenting only has small effects on IQ variance, then why would education or other environmental interventions have big effects? Something is up here.

3. The research that you (Noah) cites suggests that genetics explains more of the variance in cognitive ability than other factors. But I’m puzzled by the disagreement between these studies about how much genetics explains. For example, some studies on the Wilson effect say that heritability of adult IQ is .80 while the Emily Willoughby study finds that it’s only 0.40. What explains this discrepancy?

Expand full comment

"Gusev seems to focus mostly on educational attainment, which is plausibly less influenced by genetics than IQ, and is also less important for explaining social outcomes."

IQ is somewhat correlated with education. Those with higher IQs tend to seek higher education.

"Why aren’t the RCTs generating sustainable IQ gains if education can durably increase IQ?"

Education does not generally increase intelligence. Those of higher intelligence tend to seek higher education.

Expand full comment

Great article. The only thing I can think of that would have been worth adding is some of the data showing that IQ correlations with adoptive parents reduce to ~0 by adulthood. This has been discussed in some of Plomin’s articles.

Expand full comment

Those interested in further research in this area should look at the article entitled the Wilson Effect, by University of Minnesota psychology professor Thomas Bouchard

Expand full comment
deletedSep 29
Comment deleted
Expand full comment

I've read the study you link to, but it just completely misunderstands twin studies. It postulates that, for twin studies to work, the effects of genes vs environment have to be "separate", whereas in reality they are hugely intertwined. (e.g. "Such advances have exposed the conceptual framework of heritability studies—that of identifiably separate effects of genes vs. environments—as unsound.)

First, she's right, genetic and environmental influences are inded hugely intertwined. When she quotes: "Development ... interweaving of dynamic processes within a system that is inseparably both the organism and its environment”, she is entirely right.

But this does not invalidate twin studies because twin studies do not in any way depend on how complex the gene/environment interactions are! Really, they don't!

All they're doing is measuring inputs (different levels of genetic relatedness in fraternal and identical twins and unrelated children) and life outcomes. What happens in-between is indeed hugely complex, dynamic and entwined, but that does not invalidate the twin studies methodology. The author just completely misunderstands this.

If, say, one is comparing the taste of a cake for different levels of sugar in the ingredients, it does not matter how complex the chemical reactions in between are, it's still entirely valid to (1) alter levels of sugar, (2) taste the resulting cakes, and (3) make deductions about how the amount of sugar alters the taste of the cake. You don't have no know anything about or assume anything about the chemical reactions in the baking process for that to be a valid methodology. You do not have to assume that sugar acts "separately" from the other ingredients.

In the same way, twin studies allow us to discern the magnitude of the effects of genes on life outcomes, without having to assume anything at all about how complex gene/environment interactions are.

Expand full comment

Yes, but when you just compare the genotype with the phenotype, you miss all sorts of environmental covariations and interactions that are amplifying the effect of genes.

"Due to our vast ignorance regarding the development and expression of human intelligence, a significant portion of rGE is mostly beyond the reach of current (and ethically permissible) methods in genetics. As we mentioned above, when estimating the heritability of IQ, those gene-environment correlations that we don’t recognize or don’t know how to measure will be attributed to the genetic component."

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5754247/#R135

Expand full comment

No, you don't "miss" all the GxE interactions, you *include* them. The article you link to is largely correct, but all it is doing is clarifying what "heritability" estimates actually mean.

The strawman here is that some people think that genes act alone, independently of environment, and then you simply add the two up. No-one thinks that. It is *obviously* wrong.

Take two kids with the same genes for "liking maths". One is born into a middle-class family in today's West. His parents and teachers respond to his liking for maths by encouraging him, giving him maths puzzles, extra lessons, etc. The other is born into a tribe of illiterate farmers, where he is sent out to help his parents in the fields from age 5 and never goes to school. It is *obvious* that the complex interplay between genes and environment then has a *large* influence on that kid's maths ability at age 18. This is not some new idea that refutes twin studies, it's just obvious and something that everyone doing twin studies has always accepted.

For this reason, estimates of heritability from twin studies are only valid for the range of environments sampled in the study! If kids with the *same* range of genes were born into some very *different* range of environments (say, much more intensive one-to-one coaching for kids *poor* at maths), then the estimate of heritability would be different. Again, this is accepted and well known and textbooks on this stuff explain all of this.

All the article is doing is clarifying a correct understanding of what "heritability" means and amounts to (as oppose to some simplistic interpretations of the concept that some might have). And yes, a trait may indeed be more malleable than a *simplistic* interpretation of a heritability value might suggest. That is, if one imposed environmental interventions that are *outside* the range currently sampled in the studies, then the effect of environment will be higher and thus the heritability lower. (That's obvious if one thinks about it, and is well understood by those in the field.)

Really, all the article you link to is doing is refuting simplistic interpretations of heritability (Table 1 caption: "observations that cannot be accommodated by an “all gene” or “all environment” interpretation"), but nothing in it calls into question the basic validity of twin studies or the heritability estimates that come from them.

Expand full comment

The environmental difference between Africa and the USA is indeed captured by twin studies as shared environment. But I'm talking about rGE. Read the article again.

Expand full comment

I'm well aware that you're talking about rGE. But you (and the article) are not refuting twin studies or their estimates of heritability. All you are doing (and all the article is doing) is elucidating what such estimates mean.

Expand full comment

Say a kid has a slightly higher genetic musical ability than average and because of it practices playing the piano a lot and therefore amplifies the effect of that slight genetic difference and grows into an adult whose musical ability is vastly greater than average. This is active rGE. It's a gene changing the environment of its carrier in a way that amplifies the gene's effect. In a twin studies design, the adult musical ability gets interpreted as a genetic effect, but it's mostly an environmental effect in my view. Do you disagree?

Expand full comment