Dec 22 2015

The Genetics of Intelligence

DNA1A new study published in Nature Neuroscience explores genes that correlate with intelligence in healthy individuals as well as those with epilepsy and cognitive disorders. They looked at the brains of patients undergoing surgery for epilepsy, which means that a sample of their brain tissue was available. This allowed them to find which genes were highly expressed in brain tissue.

They then combined this data with other data sets including genetic analysis and IQ testing on healthy individuals and those with developmental disorders. They analysed thousands of genes to find those that correlate with differences in IQ.

They found that there are two sets of genes, which they are calling M1 and M3, that highly correlate with intelligence. M3 contains 150 genes that are tightly developmentally regulated. Keep in mind these are genes that they know are highly expressed in the brain.

What this means exactly is not yet known, but there are some obvious hypotheses. The researchers believe that these two gene networks are each regulated in a coordinated way and direct brain development. A certain set of gene variants is associated with optimal brain development and therefore function. The more variants differ from this optimal arrangement, the lower IQ, and not only for the range of healthy individuals, but the same genes seem to be involved in at least some individuals with cognitive impairment and epilepsy.

As an aside, epilepsy is often (not always) a marker for disorders of brain development. Do not think of this as people with epilepsy have low IQ. Rather, people who have severe disorders of brain development tend to have epilepsy.

The researchers further speculate that there may be master control genes that coordinate the functioning of the M1 and M3 sets of genes. This is not yet known, but is likely given what we know about gene regulation.

The next steps will including finding the master control mechanisms for these gene sets. Also, researchers would like to discover what these genes do, exactly. What proteins do they code for, and what brain developmental pathways do they influence?

Of course, science news stories must contain a section in which the practical applications of the new discovery are discussed. It is an obvious question and is often interesting, but I do think this speculation often gets more attention than it deserves, and sometimes goes wildly beyond the evidence. In reality we can’t always predict how basic science knowledge will pay off in the future. The knowledge is important unto itself.

News reports on this story have mentioned two possible applications. The first is that these genes could be tweaked in order to make people more intelligent. With the rapid progress of CRISPR technology, this is becoming more feasible. We are a long way from this type of application, and it is not known if altering these genes in an adult will have any effect or if they have to be changed in the embryo. Both situations are plausible, in the former case if genes affect neuronal functioning and not just their development.

Another study earlier this year examining standardized exam scores and genes from 12,500 twins found that between 54 and 65 percent of differences in the exams could be explained by genetic differences. This is always tricky data, and the very concept of using this type of twin analysis is not universally accepted, but it does suggest a strong contribution to overall intelligence from genes.

Genetics, however, are not destiny, as many are quick to point out. This still leaves a large portion that is down to environment, and that is the component we can most readily affect (at least for now).

The other possible application, as reported by the Telegraph, is this:

Report author Professor Robert Plomin believes that children should be genetically screened at the age of four so that an individualised curriculum could be tailored to their needs.

“Understanding the specific genetic and environmental factors influencing individual differences in educational achievement – and the complex interplay between them – could help educationalists develop effective personalised learning programmes, to help every child reach their potential by the end of compulsory education,” he said.

I found that suggestion a bit surprising. It certainly raises fascist images of GATACCA, even though the intention is completely benign. That aside, I don’t find the suggestion very compelling.

First, while genetics may predict a portion of overall intelligence, specific genetic variants, because they do interact with each other and the environment (and probably epigenetic factors), may never have sufficient predictive value to offer sufficient insight into a specific individual. It may be that genetic information is never as useful as regular aptitude testing.

Second, I am not convinced that personalized learning programs are of any value. The evidence does not support the conclusion that children have their own learning style, and that catering to that style is of any advantage.

Again, if you want to individualize a program you are probably better off using direct testing of academic ability and potential, and ongoing assessment of how each student is doing. Individualized attention and going at a rate that is appropriate for each student is advantageous (if resource intensive), and I don’t see how genetic information will help with this.

Using genes as a dominant determining factor in education may in fact be detrimental. It will punish students and their parents who provided an excellent learning environment, so that their children will have made good use of their potential. They will be judged not for their hard work but for their bad genes.

In any case – it just seems much better to use actual ability rather than using genes as a marker for potential ability. The latter approach does treat genes as destiny, which we know is not the case.


The new study provides insight into the genetics of brain development and function, which of course is tied to overall intelligence. It is interesting that the correlations found hold up not just for the spectrum of healthy individuals, but for those with developmental disorders and cognitive impairment.

This type of information is interesting enough on its own, but it is made more provocative as we are apparently on the cusp of the gene modification revolution. It does seem likely that we will face a not-so-distant future in which we will have to decide on the ethics of tweaking human genes to create more intelligent children (not necessarily super-intelligent, but at least at the upper end of current human variation).

25 responses so far

25 thoughts on “The Genetics of Intelligence”

  1. This would be just fine, if “IQ” were substituted for “intelligence” in the title, and many other places.

    My view is that anyone who thinks that vast complexity of human abilities can be summed up in a single number needs to have their head examined 🙂

  2. BillyJoe7 says:


    “This would be just fine, if “IQ” were substituted for “intelligence” in the title”.


    Here are the title of the articles referenced:
    “Systems genetics identifies a convergent gene network for cognition and neurodevelopmental disease”

    Here is the title of this post:
    “The Genetics of Intelligence”

  3. Logan Blackisle says:

    “A new study published in Nature Neuroscience explores genes that correlate with intelligence in healthy individuals…”

    “They analysed thousands of genes to find those that correlate with differences in IQ.”

    “They found that there are two sets of genes, which they are calling M1 and M3, that highly correlate with intelligence.”

    Are you saying that they’re equating intelligence with IQ? Doesn’t that create a pretty big error margin?

  4. The data involved IQ. Whenever I refer to “intelligence” in the article, it only refers to as it is measured by IQ. Obviously, IQ only captures a slice of all the various things that can be called intelligence.

  5. 5i5i says:

    I see no ethical reason to avoid tweaking genes to improve intelligence :
    We have no problem in doing all manner in interventions to bring people up from below average, whether that’s to fix a disease, or whatever. So why not to push above average?
    Increasing intelligence is simply is taking control of our evolution rather than trusting to blind chance.
    Having more intelligent people is most likely to be a boon to us all

  6. pandadeath says:

    It seems as though it would be useful to really determine if IQ is a useful measure of desirable traits before attempting to tweak genes to conform to the IQ test. We are only barely one hundred years away from the times when intelligence tests were widely used to justify racist and sexist prejudices. There is a lot of cultural bias to overcome when designing such tests. Not the least of which are the issues that these tests tend to measure only a small subset of populations in industrialized nations. To me it seems as though the IQ test is based upon measuring traits that were defined by a subset of a particular culture to be “important”. It is as Novella says, a small slice of human intelligence, but is it even a relevant one?

    For example, a hundred years ago, one could have measured the patterns of the iris of all the success and educated people. And if we then studied the genetics of these people with these iris patterns, we would most likely find some genetic link. Then we could speculate about inserting the genes into our children to give them the needed advantage, which could be done. But it doesn’t mean that it was ever useful because the iris pattern was never proven to be relevant to success. And all we managed to do was to limit our genetic diversity unnecessarily.

    That being said, if the majority was society was convinced that a certain gene was advantageous to intelligence, without any prove of how intelligence is measured, I would not hesitate for a second to insert it into my embryo. Just like how I got a perfectly nice education at public universities, but if my child has a chance to go to an ivy league school, I would take it. I think that is the real caveat to altering genes of future children. I don’t personally worry about people choosing a particular appearance, but I think people will choose genes that are promised to promote intelligence or athleticism (or even religiousness/non-religiousness) based upon faulty, or at least preliminary, measures of the desired trait. And I for one, would probably do so in spite of other risks. Another example, if I was given the choice to give my child the “smart” gene, knowing full well the limitations, but was informed also that it would make say the MMR vaccine 5 % percent more likely to be ineffective, I would still do it. Then if most people made the same choice, we are then in danger of losing herd immunity. And I would still take that risk knowing full well that altering a few genes could never turn my child into the next Einstein, even if those genes were satisfactorily shown to affect say scientific reasoning ability. I guess my real worry if that people will take unnecessary risks to overall genetic populations based on rubrics that are inadequately tested.

  7. daedalus2u says:

    Steve, have you seen this article? The author (Elliot M. Tucker-Drob) has a copy at his website for those without access.

    What this shows is that in the US, the the “genetics of intelligence” is really the “genetics of high socioeconomic status”.

    Another way of stating what they found. The prior studies (by Plomin) used the ACE method, where “intelligence” is assumed to be due to a linear and strictly additive sum of genetic and environmental effects. Interactive effects (genetic X environment were not allowed in the model). To the extent that there are genetic-environmental effects, they all have to be ‘taken up’ by the “genetic factor” (even if they are purely interactional). This results in phantom heritability, where the model wrongly imputes genetic effects when they are really environmental.

    A trivial example would be an ACE kind of study on slavery in the Antebellum South. If your mother was of African origin, then you were a slave. Slave status would appear to be 100% heritable, and hence 100% “genetic”, even though slavery was purely a social construct.

    Large Genome Wide Association Studies can’t find “genes” for intelligence. What they find are SNPs of unknown function. I am pretty sure that these SNPs are simply markers for different ethnic groups that either get SES privilege, or SES anti-privilege (aka discrimination).

    In the US, ethnicity correlates with socioeconomic status, the SNPs also correlate with socioeconomic status and anything else that correlates with SES (like health status, intelligence and educational attainment).

    A non-genetic causation of intelligence differences also completely resolves the Flynn Effect, and explains why large GWAS can’t find “genes for intelligence”. Intelligence is a property of a phenotype; the property of a brain comprised of many cells which have to work together “in sync”. If cells can’t work together “in sync”, they can’t instantiate thinking (which is an emergent property of the entire brain). For individual cells to work together “in sync”, they must auto-tune their signaling. That auto-tuning is more important than the genes or SNPs. The paper of a year or so ago that showed large CNVs in 10-45% of neurons would (seemingly) preclude genetics being singularly important in cognitive abilities. If a brain has to cope with 25% of cells having large scale (and different) CNVs, the SNPs that Plomin finds are pretty trivial.

  8. mumadadd says:

    “A trivial example would be an ACE kind of study on slavery in the Antebellum South. If your mother was of African origin, then you were a slave. Slave status would appear to be 100% heritable, and hence 100% “genetic”, even though slavery was purely a social construct. ”

    This bothers me, though not because I think it’s wrong. I remember a similar example; can’t remember where I picked it up: in a hypothetical society, only women wear earrings, thus earring wearing is 100% genetically determined (or maybe “heritable” was the term used, can’t recall).

    This seems like a weird way to look at it. It seems like the environment is assumed to be constant, and genes considered as the only variable.

  9. daedalus2u says:

    The problem is if you don’t know what actually “causes” something, a genetic association can wrongly be imputed to be causation.

    What actually causes differential intelligence remains unknown. Researchers like Plomin are measuring stuff and then imputing that it “proves” genetic causation because they used different training sets and testing sets. Shared correlations across training and testing sets don’t “prove” causation. Shared association is still only correlation.

    The most parsimonious hypothesis to me is that it is shared discrimination leading to stunted cognitive growth in people with the “wrong” SNPs due to being in the “wrong” ethnic group.

    Assuming the environment is “constant” and doesn’t interact with the genome is what they did assume. I agree is is an invalid assumption, but that is what they did, and as a result they get invalid results.

    In the ear ring example, if you assume that there is no gene X environment interaction, then the hypothesis that there is an environmental effect (wearing an ear ring) that depends on genetics (presence of XX chromosomes) is not something that the model can consider. It has already been ruled out by the assumption of linear and strictly additive genetic and environmental components. An ACE-type approach would “show” genetic causation for ear ring wearing.

  10. mumadadd says:

    Please can you tell me what ACE stands for so I can Google it.

    Don’t twin studies invalidate the idea that intelligence is largely (deliberately vague but in one direction) environmentally determined?

    How far back are you pushing the “environment” of the phenotype’s development?

    Doesn’t it eventually become a bit of a semantic argument about how and when we draw distinctions and categories?

  11. mumadadd says:

    Btw, I can accept the premise that IQ tests are flawed and biased towards factors not related to raw “intelligence”, and that we don’t have a proper handle on what constitutes intelligence.

  12. daedalus2u says:

    To me, “environment” is everything that isn’t coded for by DNA. The most important “environment” is the “environment” that affects how DNA is expressed and how cells proliferate and differentiate. Development is non-linear and coupled, so it exhibits the “butterfly effect.”

    An example of what that means is the example of Abby and Brittany Hensel. Their phenotype is due to “environmental effects”, not genetic effects. While they were in utero, they experienced a level of cellular adhesion that was intermediate between the level that would give a normal set of MZ twins and a normal single individual. Because there is no telelology, there can only be “normal as process”, not “normal as outcome”. Their phenotype is not typical, but there is nothing non-normal about their process of development.

    That people want to impute pathology is about human hyperactive agency detection and the need to impute teleology. There is no teleology in development, there is only the process of development working as it evolved to work in the environment that it finds itself working.

    Abby and Brittany Hensel seem to be completly neurotypical, but there is a large phenotype difference that is not genetic.

    The discussion of the ACE model is in the supplementary material of this paper.

    1 Variance component models

    We consider a general partitioning of a quantitative phenotype Y (either reading or mathematics ability in our study) into five components Y = A+D+I+C+E, where A, D and I correspond to additive, dominance and interaction genetic effects over the whole genome and C and E are shared and individual environmental effects, respectively. We assume that these components are defined to be uncorrelated with each other and thus the phenotypic variance is also partitioned into five components VY = VA+VD +VI +VC +VE (see e.g. [15] p.85-87, [22]). We apply two models, a twin model and a population level model, to estimate some of these variance components (VA, VC and VE) together with some of their covariances across reading and mathematics ability.

    1.1 Bivariate twin analysis

    We consider the traditional ACE twin model which makes a strong assumption that dominance and interaction effects are zero (D = I = 0). Under this model, twins reared together resemble each other only due to the additive effects of shared genetic variants (A) and shared environmental factors (C). In particular, the ACE twin model assumes that the environmental correlations are similar for both MZ and DZ pairs.


    They impute that the variance due to interaction effects is zero. So anything that has a “genetic component” gets imputed to be solely a “genetic effect”.

  13. hardnose says:

    IQ tests measure the ability to do well on IQ tests.

  14. hardnose says:

    They could have done this research with animal subjects, and they could have measured abilities on specific tasks. Then it might have been scientific.

  15. Sylak says:

    I don’t disagree wirh HN for once. You are party right ( and than take the reasonable part of your idea and drive straight off to road and hit the anti science telephon pole). IQ is really not perfect, and are extremely limited, but not completely useless. And there’s no way really to assess intelligence in accurate way. Making people do a specific task, like you propose, would just assess their capacity at doing that specific task. At least IQ test has diverse stuff, mathematics, coordination etc.

    Anyway, that’s interesting, og course with brain plasticity,. Thr brain change a lot after is initial state. Even people with thr good gene could end up less. Intelligent of they never use their brains, have no education etc. And thr opposite is true, but the “bad genes” persons might only need more time. Like drumming, i knew people who could drum like a guy who being drumming for 10 year in only 3-4 years! I wish I started playing drums in my teenage years, but I started 10 years ago at 26, and I have a life so I probably have 5 year of experience im real playing time because the lack of time and. Space . Anyways slice of life lol.

  16. Pete A says:

    I fully expected the usual triple axel from the Disco ‘Tute shills, especially on Christmas Eve. However, I do understand that the festivities of the season should lead us to expect only a double axel performance. You are such persistent clowns! It must really suck at this time of year to accept the harsh reality that people across the globe are becoming increasingly disinterested in your ‘creation science’ ideology and your plethora of other abject bullshit: so thinly veiled under the guise of Intelligent Design.

    Fortunately, I live in an area of Europe where your abject bullshit and your modus operandi are totally rejected by legislation.

  17. edamame says:

    “Obviously, IQ only captures a slice of all the various things that can be called intelligence.”

    This is pretty important, without it we risk falling into journalism hype about the study. Imagine finding genes in mice correlated with performance in the Morris Water Maze. It’s actually really hard to interpret the meaning of such correlations. Calling it “intelligence” is tendentious.

  18. BillyJoe7 says:

    Please ignore my incomprehensible post above.
    I don’t know what I was drinking thinking.

  19. Pete A says:

    While on the subject of ‘intelligence’: Neither an IQ test nor The Genetics of Intelligence can adequately explain “How the internet misled you in 2015”

  20. KeithJM says:

    Even if we don’t know exactly what IQ tests are testing (other than the ability to do well on IQ tests, which seems to correlate with academic success to some degree), we can agree that specific brain development disorders should be less strictly associated with socioeconomic class, right? Even if poor people have a higher incidence of this sort of disorder, you’d expect wealthier children to be more likely to be diagnosed and included in this study (at least in the US, where health care is perhaps less universally available).

    If these genes show a correlation to IQ and a negative correlation with brain development disorders, that steps outside of the socioeconomic origin of IQ.

  21. steve12 says:

    “Another way of stating what they found. The prior studies (by Plomin) used the ACE method, where “intelligence” is assumed to be due to a linear and strictly additive sum of genetic and environmental effects. Interactive effects (genetic X environment were not allowed in the model). To the extent that there are genetic-environmental effects, they all have to be ‘taken up’ by the “genetic factor” (even if they are purely interactional). This results in phantom heritability, where the model wrongly imputes genetic effects when they are really environmental.”

    It’s more accurate to say that this COULD result in phantom heritability, right? Why is it more justifiable to say that interactions are evidence for environment than it is to say they’re evidence for genes?

    The interactions are of special interest here since by design, much of what’s coded in the genes is meaningless without X environmental feedback. The feedback is often as simple as gravity, light, presence of other beings, etc. It’s simply more efficient to NOT store information that will be provided by the environment – it’s a compression algorithm of sorts.

    All of the phenotypes brought about by this process would be interactive. Considering this, is it correct to say that all of these interactions are environmental? I don’t think so.

  22. daedalus2u says:

    Plomin ignores the environmental aspects as they interact with genetics. He makes unjustified and unjustifiable assumptions and then proclaims that ignoring gene X environment interactions are valid. They are not valid. That approach does result in phantom heritability. It can’t not result in phantom heritability. He assumes that there is no gene X environment interaction. If there is gene X environment interaction, and that “phenotype effect” must end up in one of the variance bins that he has allowed. The only “bins” his analysis allows are “genetic” and “environmental” and “noise”. Since he never measures “environmental”, he can’t allocate degrees of freedom to the environment.

    What is the phenotype of somatic cell nuclear transfer clones from the same donor? They are all different, and more different from each other than they are from unrelated wild-type conspecifics. Yes, that “phenotype” difference is largely due to the “environment” of the SCNT cloning process as mediated through the genome (coding and non-coding).

    In other words, clones with identical coding and non-coding DNA have phenotypes that are more different than do unrelated conspecifics.

    If you made 10^6 SCNT clones, every single one of them would be different at some level. How does it make sense to consider any those differences to be “genetic”?

  23. steve12 says:


    Was that last post a response to me?

  24. SWattoo says:

    I know this is an old post. However, I have two questions that relate to this topic:

    1.) How far away are we from genetic selection (via PGD) and/or genetic editing of children in order to optimize cognitive ability? From what I understand, researchers have had a lot of difficulty finding genes that account for intelligence.

    2.) What impact will this have on other people? If, for example, we find that most of the genes that code for cognitive ability are associated with structural aspects of the brain and adults consequently cannot avail themselves of cognitive enhancements, then these technologies will create massive social problems. I know you mentioned in a previous post that we ought not to hold the technology responsible for these social and cultural implications of its use and instead try to address the cultural and social factors that bring about the aforementioned negative consequences (I think you said this in your post about CRISPR). However, I’m not sure there’s any straightforward way of dealing with a lot of the social forces at work vis a vis this particular technology. The best we might do is create robust social security nets. However, even then you’ll probably see a kind of crowding out effect where un-enhanced are slowly driven out of various spheres of public and economic life because they can’t compete.

    To be clear, I’m not against enhancement per se. I think it will be a very good thing for society in the long run and it would be unfair to future generations to deprive them of cognitive advantages that they would enjoy if we made full use of this technology. That being said, I think this technology will inevitably create substantial short-term problems that wont be eliminated until pre-enhancement generations die out (or until we find alternative forms of enhancement that adult humans can use to level the playing field).

Leave a Reply