Genetics of IQ: not so simple
IQ is not controlled by one or two genes
Posted December 19, 2011
You have probably noticed the trend of newspaper articles describing the genetic basis of some particular trait. "Now we know the genetic basis of (altruism)(risk-seeking)(happiness)(religiousity)!" These are all published in legit science journals and come with the stamp of approval of Real Science.
But behind the scenes has been a big scientific debate about whether these types of studies are scientifically legitimate. And now it's starting to seem like a lot of them are wrong.
It all comes down to statistics. Let's take IQ. When you compare someone's IQ to their genes, you scan all their genes (a large number) against their IQ and look for genes that correlate. By chance, some will correlate, and to deal with that you have to add a correction term to the statistical comparison. That is no so easy, since it depends critically on some key and debated assumptions.
The most controversial studies are focused on single nucleotide polymorphisms (SNP). You know how DNA is a code of millions of ACTGGTCACAATG... The SNP is a place where the DNA is changed by one letter.
There are numerous studies that have found SNP's that predict IQ. Well, a group of scientists led by Christopher Chabris has gone back and looked at 12 of the most famous ones, and looked in some of the giant DNA databased that have sprung up. They failed to replicate. Failure to replicate is Science Language for "uh oh., this might all be untrue.".
The title of the paper is "Most Reported Genetic Associations with General Intelligence Are Probably False Positives." This is pretty direct language for the title of a paper. It is science Language for "You screwed up majorly and we are here to fix it."
I find this result disturbing because it suggests that these earlier papers, many published in Big Famous Journals were just plain wrong. The authors helpfully point out in the Conclusion that "candidate gene studies of complex traits were [recently] commonplace in medical genetics research. Such studies are now rarely published in leading journals." It's pretty clear the pendulum has swung away from these simple genetics results.
More broadly, it suggests that other linkages between cognitive traits and SNPs are wrong. Even more broadly, it suggests that maybe we have to revisit the way we do statistics in an age where computers can help us collect millions of data points. The old statistics tools invented by Fischer, Neyman, Pearson, and others, might not scale. I see this statistical problem express itself in a different way in neuroimaging studies, which collect thousands of voxels of hemodynamic activity all at once. That debate is continuing to rage.
Importantly, this doesn't mean that IQ isn't genetic. It just means that IQ isn't going to come down to a single gene or 12, the way that rounded vs smooth peas worked for Mendel. Instead, it seems that some portion of your IQ is determined by the sum of about a zillion genes interacting, but no single one makes much difference. And of course, some of IQ is determined by environment, and some is a gene-environment interaction. How much is each part? After we sequenced the human genome, we thought we would soon understand stuff like this. Turns out genetics is a lot harder than we hoped.
UPDATE: More coverage on the topic from neuroskeptic.
Thanks to @markgbaxter for the tweet!