Intelligence genes are hard to find.

Researchers led by Robert Plomin of the Institute of Psychiatry in London obtained intelligence scores for 7,000 seven-year-olds based on verbal and non-verbal reasoning tests.

They also took DNA samples from the children in the hope of identifying genetic differences between the high and low scorers.

The huge trawl identified 37 variants in six genes that appear to be play some role in differences in intelligence.

But the individual effects of these genes was barely detectable. Together they account for just one percent of the variation in intelligence between individuals.

We're a long way off from genetically engineered intelligence upgrades, it seems.

patrissimo has recently written a series of posts about intelligence and genetics, starting with a comment that race is genetic and real, there are genetic intelligence differences (and that he's not posting any more about this), that twin studies show that IQ is largely genetic, and finally a poll on the topic.

As usual, it's hard to get past the emotions in these conversations. One thing Patri is heavily relying on, though, are those twins studies. I think their informativeness is way overstated. Patri writes:

Consider the correlation between the IQ of a) identical twins raised together vs. b) fraternal twins raised together. We've kept environment almost constant - same family, same wealth, same education, same *prenatal environment* - the only difference I see is how parents/people respond to identicals vs. fraternals[1]. All we've varied is the genetic similarity, (a) shares 100% of genes, (b) only 50%.

The correlation in case a) is 0.88, and b) is 0.53. So we lose 0.35 of correlation between two people with almost identical environments, culture, wealth and education, based solely on whether they share 100% or 50% of genes. 50% genes, 0.35 correlation, hence, roughly speaking, IQ is 70% heritable (70% of variation is caused by the genes).

If environment played such a large role, the difference between (a) and (b) would be much smaller because in both cases the environment is the same. Variance within the family would be low compared to variance within a country, so MZ/DZ twins raised together would have about the same IQ correlation. We would not get such a dramatic drop by only varying whether the 2 siblings pull one or two sets of genes from the random Mendelian lottery. But that ain't the case. You can add data to this w/ various other kinship relations and level of shared environment, and you add more possible environmental confounders (in theory), but in practice the results get you about the same heritability as the simple (a) vs. (b) which cleanly short-cuts such problems.

This seems very persuasive. Separated identical twins show highly correlated IQs, and we've all seen those tv documentaries about how alike separated twins really are. Given that they have different environments, there can be no question that the sameness they show is genetic, to the tune of 70% of the correlation.

Did you catch the assumption there? This argument relies on the idea that the environments are uncorrelated. The more similar their disparate environments are, the weaker this evidence becomes for the dominant role of genetics in intelligence.

Why might their environments be similar? Because to at least some extent in our culture, your environment is not just thrust upon you with no consideration about you as an individual -- you have a lot of power, knowingly or unknowingly, to influence the environment in which you grow.

This is more clear in sports, so let's talk about baseball. Suppose you are a good natural athlete -- a bit faster than the other kids, a bit better at hitting a baseball. You stand out, just a bit, from your peers. Plus, you like being good at things, so you are excited to play and practice. You seek out good coaches, play for better teams, and work hard at getting better. By the time you graduate from high school, you are much, much better at baseball than the same kids you started out with, way back when you were a little better.

Here, the environment took a small genetic advantage, and became sort of a force-multiplier: it magnified your advantages, put them in the best light, and you now are far ahead of your peers. Who, of course, concentrated on things *they* were good at instead of baseball.

Now consider a set of identical twins, separated at birth. If both of them played little league baseball, they both would have had extra opportunities to learn and thrive at something they were good at. All that is required is a similar opportunity, one which is common.

The analogy to intelligence should be fairly clear. If you are slightly smarter than your peers, you will tend to do better at academic tasks and like them more. Your small genetic advantage will get multiplied into a large difference as you seek out tougher classes, better teachers, other learning opportunities, and so on. The very things you seek out are similar in nature to the things your identical twin with identical advantages will seek out, and suddenly your environments are also very similar.

If I'm right, then genetic differences start a positive feedback loop that exaggerates those differences, as children concentrate on nurturing the skills or talents at which they are good, and neglecting those at which they are not good. So they are at or near the top of their game on the things they concentrate at, and likely substantially far below their potential on the things for which they already had lower potential. And separated twins would end up looking very similar.

Now, if there are separate twin studies that look at children in radically different environments, then the story might be different. If a child raised by Madonna in England has similar IQ scores to a child raised in poverty in Malawi, then we can be sure that genetics dominates environmental factors (neglecting prenatal environments, maternal nutrition, breastfeeding, etc. And assuming Madonna is a reasonable parent.). I don't know of any studies that look at radically different environments though -- except generational differences.

IQs have gone up 15 points in the US since the 1930s among populations that have been here for that long (whites and blacks). It's hard to see how that could be genetic; however the environment in which we live is very different. Children learn abstract reasoning, math, and science at an early age now; education then had no emphasis on abstract reasoning, and people were bad at it. They had other things to worry about. Not to mention vastly greater childhood illnesses relative to modern times due to modern vaccinations and better nutrition. Indeed, the differences in test scores are because of large gains in abstract reasoning-type problems, and very small or negligible gains in things like vocabulary or arithmetic.

I think something similar explains a good amount of the IQ difference between African countries and first world countries. Sub-saharan africans are raised in an environment that encourages learning to deal with concrete reality (how do we make enough money to eat, avoid getting malaria, etc) and does not encourage abstract reasoning. Their IQ test results reflect those changes. As their societies change, IQ scores should come up, as people suffer less from disease and harsh current realities, and have more time to spend on early education and abstract thinking. This is why I am very supportive of the Gates Foundation's plan to vaccinate all of Africa. If they can pull it off, their society could transform the way ours did in the first half of the last century.

In summary, I think environmental effects magnify genetic differences as people seek out environments they prefer to the extent that they are able. Genetic differences then appear to loom large when the genetics are very similar (as in twins), even though a large amount of the difference is environmental force-multiplication. Do genes play an important roles -- of course! A critical role. But the story is much, much more complicated than "look, correlations imply that genetics play a defining role in IQ or anything else".