Genes and Intelligence: My Anti-Story

In the latest issue of Scientific American, I have a feature on the biology of intelligence. (Read it online at or I’ve been fascinated by the subject for a long time, and I decided recently that the time was right to put together an article.

What’s the news? That there is no news.

Allow me to explain…

A lot of experts on intelligence were very busy late last year trying to clear up some misconceptions about the nature of intelligence in the wake of Nobelist James Watson’s remarks about race and IQ. (The best example I came across at the time was an hour-long discussion that aired in December 2007 on the show On Point. You can listen to it here.) It would be a mistake, however, to conclude that there is no such thing as intelligence–by which I mean something that scientists actively study, testing hypotheses and finding significant results. Intelligence tests do identify a difference among people that has predictive power, and that difference can be linked–in part–to differences in people’s genes.

The news about intelligence is that now scientists have new tools for probing intelligence, from brain scans to gene chips that can search for variations in half a million genetic markers at a time. But so far, those tools are yielding some pretty scant results. For example, just a handful of genes show much sign of influencing intelligence, and yet each one accounts at best for a fraction of one percent of the variation in test scores.

It may not be fashionable to write about the lack of breakthroughs, but in a case like that of intelligence, that’s what fascinates me most. I’ll be curious to hear what others think.

Update: Where are my manners? For those who want to head for the primary literature, here are two key papers:

Butcher et al, Genome-wide quantitative trait locus association scan of general cognitive ability using pooled DNA and 500K single nucleotide polymorphism microarrays Genes Brain Behav. 2008 June; 7(4): 435–446. (Free full text)

Jung and Haier, The Parieto-Frontal Integration Theory (P-Fit) of intelligence: converging neuroimaging evidence, Behav. Brain Sci. 2007 Apr;30(2):135-54 (Download free pdf here)

The Parieto-Frontal Integration Theory (P-FIT) of intelligence: converging neuroimaging evidence.

23 thoughts on “Genes and Intelligence: My Anti-Story

  1. I know its fashionable to talk about our recent “dumbing down”, and I think its true, but… or better yet, AND, I think that the possibility for good intelligence seems to be in most people. I still believe that education is the silver bullet that will bring about the “Great Enlightenment” . Take that you post modernist.

  2. But a lack of breakthroughs is what *usually* happens in science. Judging by headlines it’s one big discovery after another, and that’s simply not true. Kudos for writing about a non-breakthrough!

  3. “Intelligence” is just a metaphor.

    Though genes may set the stage for how an individual perceives their environment, “intelligence” is learned. It helps to think of intelligence as a behavioral repertoire, a set of skills one has, instead of some sort of physical entity within an organism.

  4. When it’s so well established that environment is overwhelmingly the determinant of so many forms of intelligence, what is the rationale for allocating such expensive tools to address this question? The cost and opportunity cost really bothers me. Maybe I’ll have to read your piece.

  5. For example, just a handful of genes show much sign of influencing intelligence, and yet each one accounts at best for a fraction of one percent of the variation in test scores.

    When I fit a statistical model that is significant, yet accounts for only a trivial portion of the variance, I strongly suspect that 1) the data is junk, or 2) I have made incorrect assumptions in my model. This suggests we need to rework the basic concept of intelligence in a different way (as Matt already noted).

  6. I no expert on statistical genetics, but it is likely the data is the best available – I did not mean to imply these are junk studies. I sometimes see statistical models with strong significance yet almost no predictive power, and it generally indicates a bad model fit. In other words, my model isn’t considering the data in the correct way. I am suggesting something analogous with intelligence: We have learned some key pieces of the puzzle, but it not being considered in the right sort of way.
    I will go out on a (very short) limb and suggest that intelligence tests really aren’t measuring the important qualities very well.

  7. <B<"What’s the news? That there is no news."

    With all due respect, I disagree.

    From The American Psychological Association:

    “The observed increase in average IQ scores of 3 points per decade has been reported by many as proof
    that intelligence is not stable but is flexible with regards to environmental influences. However, none of the
    authors in this extensive review of
    the data, believes that intelligence is increasing at a rate greater than can be
    attributed by eugenic means (breeding smarter people). Instead, they are looking at
    the mechanisms of tests and how they are administered and interpreted in different time periods as people
    become exposed to differing environments. That is, as the humans change their environment they change the
    context in which their intelligence comes into play, and the testing methods must take this change into
    account. Obviously they have not because the generational change in IQ results has increased,
    without any evidence that today’s children are any smarter than their great-grand-parents. So it is back to
    the drawing boards for the psychometricians who design the tests.
    Neisser states that, “Whatever g may be, we at least know how to measure it. The accepted best measure,
    which has played a central role in analyses of the worldwide rise in test scores, is the Raven Progressive
    Matrices. This test, devised by Spearman’s student John C. Raven, was first published in 1938 and is now
    available at several levels of difficulty. Arthur Jensen has said that Raven’s test ‘apparently
    measures g and little else’ and that it ‘is probably the surest instrument we now possess for discovering
    intellectually gifted children from disadvantaged backgrounds’. The Raven is of particular interest because it
    shows such large IQ gains over time. In The Netherlands, for example, all male 18-year-olds take a version of
    the Raven as part of a military induction requirement. The mean scores of those annual samples
    rose steadily between 1952 and 1982, gaining the equivalent of 21 IQ points in only 30 years! This amounts to
    a rate of no less than 7 points per decade — a figure confirmed by data from many other countries. What can
    these increases mean?”
    And later he states, “However, one may choose to interpret it, the fact that (unknown) environmental
    factors are raising the mean IQ of Americans by 3 points per decade certainly shows that the environment
    matters! The second proposition has quite a different status. Within a given population and a given range of
    environments (e.g., those that are characteristic for White American males in 1998), genetic factors do make a
    major contribution to individual differences. This has now been shown beyond a reasonable doubt
    by the methods of behavior genetics, a discipline that is primarily concerned with variability. The
    individuals in a given population differ on almost any measure one is likely to care about: their heights,
    weights, Raven scores, IQ scores, or anything else. Every such measure has a distribution, often a bell-shaped
    normal one. . . . Unfortunately, no one knows what it is about the environment that makes
    this contribution to differences in IQ scores. Some obvious possibilities, such as the economic and
    intellectual quality of children’s home situations, may be less important than was once believed. The
    surprising fact is that when biologically unrelated children are raised in the same home (as in many cases of
    adoption), the correlation between their IQ scores is unimpressive in childhood and near zero as they grow up!
    This finding is important, but it is still negative: The aspects of the environment that do matter for the
    development of intelligence have not yet been identified.”
    Well, it doesn’t mean that people are getting any smarter but rather that the environment is impacting
    how the results of IQ test scores are interpreted. That is, the expression of intelligence is not the same as
    intelligence. If children today have a high level of exposure to visual-spatial stimulus such as computer
    games, and IQ tests used 50 years ago used the same visual challenges to interpret intelligence, then the
    tests may no longer be valid from one generation to the next for comparisons. Intelligence hasn’t changed, the
    means to test intelligence has not kept up with human contextual flexibility to deal with a changing
    “Flynn states this fact succinctly, “Moreover, data whose quality cannot be challenged have posed the same
    question. The Dutch military data, like those of Israel, Norway, and Belgium, are near exhaustive; but even
    better, Vroon compared a sample of the total population of Dutch examinees with the scores of their own
    fathers. There is simply no doubt that Dutch men in 1952 had a mean IQ of 79 when scored against 1982 norms.
    Has the average person in The Netherlands ever been near mental retardation? Does it make sense to assume that
    at one time almost 40% of Dutch men lacked the capacity to understand soccer, their most favored national
    Of course not, and that is why the Flynn effect is not taken seriously as an increase in real
    intelligence, because we just do not see one generation as more intelligent than previous ones, on a myriad of
    social indication scales. One would have to assume that the Greek philosophers were all mentally retarded, and
    yet wrote with such elegance that we still read and try to interpret their works today. It is absurd. And not
    one scholar in this book believes that real intelligence is changing but ever so slightly over time from
    environmental effects.”

    I have written extensively on this topic. Start at

  8. Gene-hunting has already found over 800 Mendelian genes of major effect that influence human intelligence. Thing is, they all reduce it. They’re called genes for mental retardation. Chances are all of the DNA microarray results will look similar –lots of quantitative trait loci will emerge that predict slightly impaired intelligence in certain populations. But those QTLs will basically be evolutionarily recent mutations — things like single nucleotide substitutions, deletions, insertsions — that reduce the precision of brain growth and the efficiency of brain function. Since they’re mutations, they’re selected against, and don’t spread across populations, and don’t replicate across samples.– Geoffrey Miller

  9. Geoffrey, you’ve brought up a point that I’ve often considered. A few years back I was working in Oxford, close to one of a group involved in analysis of the genetics of Fragile-X syndrome. I’m sure you know the details of this condition but for others unfamiliar the important point is that, while it frequently results in highly retarded males, its expression in females can be much less severe. One of the societal points of this condition as pointed out to me at the time was that the mildly retarded fragile-x females were able to function in society, they just had a lower IQ – and consequently were not that good at birth control – they frequently had much more children than average, thereby increasing the numbers of carriers and thus the overall incidence in the population.
    Obviously this is just one condition but one wonders if there are any other similar factors that may become exposed by studies such as the haplomapping project once we begin to compare things like reproductive rates to the data.

  10. A very fascinating summery on IQ research until now. Perhaps the most telling research, I would think, is from Turkheimer et al. that debunked gene-centric view based on identical twins.

    It would be nice to have another summary of the all the researches that look at environmental factors that influence IQ. viz.,
    – Parental economic conditions
    – Parental verbal ability
    – Food habits
    – Temperate/Tropical environment

  11. I feel like the anti-story here is larger than IQ, and is really about what genome-wide association studies have been able to reveal about continuous traits in general at this stage. I doubt intelligence is very unique compared with other quantitative genetics traits. Height and intelligence, for example are similar in a lot of ways (e.g. similar heritabilities), and we’re seeing the same stuff with the genetics of height.

  12. Given what we know about the complexity of the nervous system–the number of different types of neurons, the complexity of connections, the multitude of neurotransmitter receptor subtypes with subtle functional differences, it would be astonishing if there were not a huge number of alleles impacting intelligence. Of course, there is no meaningful distinction between a gene for higher intelligence and a gene for lower intelligence–it just depends upon which one you define as “normal.” It is likely that most of them have effects too subtle to be detected in isolation, and that some of them interact in nonlinear ways.

    It is notable that some transgenic manipulations of mice result in animals that exhibit greater intelligence in some behavioral tests. Some of these are pretty crude manipulations, such as deleting or overexpressing a particular gene, so it seems unlikely that the apparently greater intelligence results in a benefit in a natural environment–if it did, the mutation would likely already be common. Very likely the improved intelligence in some contexts is associated with deficiencies in behaviors that are not tested in the laboratory. However, it supports the common wisdom that intelligence is multifactorial, and cannot be adequately defined by a single score.

  13. Another comment: rather than attempting to correlate general measures of intelligence such as IQ with SNPs, it might be more productive to look for families in which a high capability in one kind of intellectual activity is linked with low capability in another kind–for example, people with high mathematical skills but difficulty with facial expressions. This might be a way of finding genes with larger effects, which might lead to genuine insights into the nature of intelligence.

  14. Carl Zimmer is not my favor science writer, being quite overrated. Too often Carl Zimmer distorts facts for a particular premise he seeks to pursue. Here, science and truth get quite distorted.

    Intelligence and genes is certainly a case in point.

    First, Carl Zimmer never answers the question why the brain in the first place.

    All organic life is cognitive because it is a Responsive Design. Carl Zimmer is still on the shore where Evolution is guided by Darwinian conception through Natural Selection, saying nothing!

    Roy D. Schickedanz
    Refuting Charles Darwin in the case of Life’s Responsive Design

  15. [i]For example, just a handful of genes show much sign of influencing intelligence, and yet each one accounts at best for a fraction of one percent of the variation in test scores.[/i]

    i think the question is how we could objectively average variables without even controlling them par avance, as seems to be the case with ‘intelligence’.
    could there be such a thing as global intelligence which can be measured by IQ, i highly doubt it.

    even if IQ marks were 100% accurate, how could we possibly differentiate genetic factors from environment ones which affect different parts of IQ?

  16. Hi Carl. The two articles you provide do not actually establish your point.

    One is a study of SNPs with no proven chemical effect on anything, and thus no meaningful conclusion can be drawn from them. For all we know, they are intron polymorphisms (the paper fails to specify!), in which case it’s absurd to link them to anything behavioral.

    The other is a meta-analysis, enough said. If I had a dollar for every time some statistical hand-waving allowed a meta-analysis to reach an opposite conclusion from every other study on the same topic, I’d be wealthy enough to pay poor kids to write this comment for me. From the perspective of scholarship, they are not worth much. Double so in the case of biology or medicine.

    Also both articles beg the question they seek to discover, namely whether “intelligence” is a meaningful description of human behavior. In general, studies on this topic consistently fail to eliminate sufficient variables from environment to make a meaningful conclusion about a genetic basis for variations in intelligence, not including obvious cases of mental retardation or Down syndrome or something like that. The “in otherwise healthy adults” category still supports a null hypothesis.

  17. I’m absolutely amazed at how many people still argue that intelligence is not a valid description of human behavior. Anyone with more than a dozen acquaintances can tell you which of them are the more intelligent. On first meeting someone, most of us quickly evaluate their intelligence, so I’d be very surprised if intelligence is not something that can be objectively measured, with at least as much accuracy and meaning as how good looking someone is. (OK, I actually think IQ can be evaluated a bit more objectively.)

    It has been my experience that neither social background nor education has any real impact on intelligence. One’s occupation does not seem to predict their intelligence either, though performance in any occupation does. Of course, how long a particular occupation will retain someone of high IQ definitely depends on how interesting that person finds the work and that is quite unpredictable. And, some occupations will be quite limiting for someone without a high enough IQ to perform. For example, you don’t find many theoretical physicists with an IQ below 120 though they do come in a wide range of social skill.

    Since education, social class and most other environmental factors seem not to have any real effect on IQ, an assertion based only on personal experience, it seems quite reasonable to look for genetic markers. And, since no race is excluded from producing brilliant people those genetic markers must be more basic to humans than are the markers for race. I think the search for such genes is well worth the effort. Note however, that being born with particular traits cannot be blamed entirely on genes. It seems that a great deal of what we are born with is a matter of chance. I’m not even implying environment during development, but rather just chance. I suspect you are born with a particular IQ or g for the same reasons you are born with particular fingerprints.

    If I’m right, then a more important question is how to educate people so that we, and they, get the most from what they are born with. I think “no child left behind” should be replaced with “no child’s potential left underdeveloped”.

    -Troy Stark

Leave a Reply

Your email address will not be published. Required fields are marked *