At least a dozen papers have been published that appear to show links between general intelligence and specific genes. But the authors of a forthcoming paper, to be published in Psychological Science, contend that those findings are all most likely false. I asked one of those authors, Christopher Chabris, who co-wrote the best-seller The Invisible Gorilla, a few questions via e-mail, and he was kind enough to answer:
So general intelligence is to some degree heritable, but previous attempts to pin down the particular genetic variants involved were flawed because their sample sizes were too small. Is that right?
This is pretty much correct. But I wouldn’t call those studies “flawed” in the sense of “designed erroneously.” When researchers started looking for the genes that contribute to differences in cognitive ability, it was thought, or at least hoped, that there would be specific genetic variants that each accounted for a meaningful portion of those differences, say one to two IQ points. It was also very expensive to assay genes, making it impractical to collect really large samples. So samples were used that in retrospect, given what we now know about the genetics of attributes like height and obesity, seem much too small, and when samples are too small, some of the positive results that turn up are more likely to be accidents than true successful finds. Our article shows that out of the 12 published genetic associations with intelligence that we were able to examine, none consistently replicated in our three samples that totaled almost 10,000 people. This doesn’t mean at all that there are no genes associated with intelligence. It just means that we haven’t found them yet, and this is probably because the common genetic variants that have been studied each have at most tiny effects on intelligence—not the large effects that had been expected when the enterprise of finding those genes began.
This seems like bad news. Does your study mean that, when it comes to figuring out which genes are associated with intelligence, we actually didn’t know what we thought we knew, and now we’re back to square one?
Not at all. We know more about what’s called the “genetic architecture” of intelligence. It turns out to be similar to that of height, which is interesting, since height is a trait that’s easy to measure accurately, not based on mental performance, and not very controversial. For both height and intelligence we now know that there are probably no individual genetic variants that are both common in the population (as opposed to rare mutations that are peculiar to particular families) and responsible for large differences between people in the trait. The field has to look for more genes, each with much smaller effects, which requires different methods from the ones that have been mostly used to date. So we have replaced some conclusions that turned out to be premature with some knowledge that can form a basis for the next steps in working out the genetics of cognitive abilities. I think we have also gained an instructive example of how the scientific literature can wind up—through no one’s fault—incorporating results that turn out to be unlikely to be correct.
The concept of general intelligence has had its critics, like the late Stephen Jay Gould. Obviously you think it’s worth studying. Why?
Perhaps part of the confusion involved stems from the fact that “general intelligence” means something specific to psychologists that it may not mean to others. In psychology, it’s a well-established fact that most tests that seem on their face to measure mental ability —the ability to figure out the right answer to an objective question or puzzle—correlate positively with each other. That is, people who do well on some of the tests tend to do well on the others. The correlations aren’t at all perfect—being a vocabulary whiz doesn’t guarantee that you will also be a math wwiz—but general intelligence refers to the fact that, all else equal, you should bet on the person with a better vocabulary to be better at math than the person with a smaller vocabulary, and bet on the person who is good at math to have a better vocabulary than the person who is bad at math. Understood this way, the concept of general intelligence is just a morally neutral statistical observation about human behavior. But it’s an incredibly consistent, universal, and powerful observation, made more powerful by the fact that one’s score on a fairly short cognitive test correlates with a diverse array of other life outcomes, such as educational attainment, income, occupation, susceptibility to mental illness, and even lifespan—even when many other factors that affect these outcomes are controlled for. I think that understanding the reasons for this state of affairs is a compelling scientific goal, and genetics is an important category of reasons that must be explored if social science is to reach a correct understanding.