Is Genius Fragile?

One of the subjects people care most about in ev psych and related disciplines is intelligence. Teachers would love it if all of their students suddenly began scoring in the 90th %; so would parents, of course. Tons of psychological studies have been done on subjects like “Do people score better on tests after thinking about famous scientists,” (without finding much useful,) not to mention millions of dollars spent on education reform without, as far as I can tell, much real change in school performances.

Since “IQ”–our best attempt at measuring and quantifying intelligence–appears to be at least 50% genetic, genes are a good spot to look when attempting to unravel the mystery of genius.

One of my theories on the subject is that if there are two kinds of dumb, perhaps there are two kinds of smart. Obviously dropping someone on their head is probably not going to result in genius, but perhaps there are some people who are smart due to having the good luck to have a variety of genes that generally code for things leading to high IQ, while other people are smart because they have a few particular genes or mutations. The folks with the generally IQ-boosting all-around genes are people who come from a background of parents and extended families with similar IQs to themselves, but folks with rare, particular, or novel mutations/genes would likely stand out even from their families. Such genes might have deleterious side effects or only confer genius in one or two particular arenas, resulting in, say, the stereotypical absent-minded professor or idiot savants.

If genius is fragile–my definition of fragile, not necessarily anyone else’s–then it is easily damaged; the difference between high-IQ and low-IQ in a particular population will be related to the possession of deleterious mutations that damage IQ. If IQ is not fragile–that is, if it is robust–then we would find rare, beneficial genes that boost IQ.

Environmentally, it is already obvious that genius is fragile–that is, it is much easier to drop someone one their head and subtract 40 IQ points than to find any intervention that will reliably add 40 points, but this does not necessarily preclude a variety of interesting genetic findings.

Perhaps I am thinking about this all wrong, but that’s the structure I’ve got worked out so far.

Anyway, so people have been searching for genes linked to IQ. Will they find specific IQ-boosting genes that highly intelligent people have, but dump people don’t? Or will they find specific IQ-damaging genes that dumb people have but intelligent people don’t? (Or maybe a combination of both?)

So, Neuroscience News recently covered a study published in Molecular Psychology that looked at genetic differences between highly intelligent people and the general population.

Now, I’m going to have to stop and point out a potential design flaw, at least according to the article:

“Published today in Molecular Psychiatry, the King’s College London study selected 1,400 high-intelligence individuals from the Duke University Talent Identification Program. Representing the top 0.03 per cent of the ‘intelligence distribution’, these individuals have an IQ of 170 or more – substantially higher than that of Nobel Prize winners, who have an average IQ of around 145.”

Duke TIP is aimed at middle schoolers, based largely on their elementary school test scores Anything that starts out by comparing the IQs of elementary school kids to people who’ve already won Nobel Prizes may not be saying much.

Second, I’d just like to note that while the article is unclear, they are probably not claiming that all Duke TIP participants have IQs over 170, since they don’t–Duke TIP’s own website states that they only require IQ scores over 125. Rather, I suspect they used the test scores submitted to the TIP program to select students with IQs over 170. If some confusion has occurred and they actually used people with 125s, well, results may not be as claimed.

Quick rough calculations indicate that 1,400 people in the top 0.03% is not an unreasonable number, since it would only require 4.667 million people, and there are about 4 million kids per grade level in the US, TIP takes from multiple grades, and they could have used multiple years’ worth of participants. But I don’t know how many kids TIP takes each year.

Anyway, results:

“The study focused, for the first time, on rare, functional SNPs – rare because previous research had only considered common SNPs and functional because these are SNPs that are likely to cause differences in the creation of proteins.

“The researchers did not find any individual protein-altering SNPs that met strict criteria for differences between the high-intelligence group and the control group. However, for SNPs that showed some difference between the groups, the rare allele was less frequently observed in the high intelligence group. This observation is consistent with research indicating that rare functional alleles are more often detrimental than beneficial to intelligence. …

‘Rare functional alleles do not account for much on their own but in combination, their impact is significant.

‘Our research shows that there are not genes for genius. However, to have super-high intelligence you need to have many of the positive alleles and importantly few of the negative rare effects, such as the rare functional alleles identified in our study.’

Or as the abstract puts it:

We did not observe any individual protein-altering variants that are reproducibly associated with extremely high intelligence and within the entire distribution of intelligence.* Moreover, no significant associations were found for multiple rare alleles within individual genes. However, analyses using genome-wide similarity between unrelated individuals (genome-wide complex trait analysis) indicate that the genotyped functional protein-altering variation yields a heritability estimate of 17.4% (s.e. 1.7%) based on a liability model. In addition, investigation of nominally significant associations revealed fewer rare alleles associated with extremely high intelligence than would be expected under the null hypothesis. This observation is consistent with the hypothesis that rare functional alleles are more frequently detrimental than beneficial to intelligence.

*What does “and within the entire distribution of intelligence” mean in this sentence?

To be honest, I’m not sure about the interpretation that only genetic differences between high IQ and low IQ people is that the low-IQ have more deleterious mutations and the high-IQ don’t. For starters, we observe ethnic variation in IQ scores, and I find it difficult to believe that vast swathes of the planet, some of which have very different marriage patterns, have abnormally high levels of deleterious, fitness-reducing mutations that other swathes of the planet don’t.

I certainly can believe, though, that there are deleterious mutations that reduce IQ.

What do you guys think?





8 thoughts on “Is Genius Fragile?

  1. …TIP is one of the “Talent Searches.” These are now US-wide and happen in some regions of some other countries as well (such as Ireland). The first one (CTY) was started by JHU in the ’70s and now covers the US coasts. Duke then started TIP covering the American South, and Northwestern started CTD covering the Midwest. JHU also runs CTY-Ireland in association with (IIRC) Dublin City University.

    The Talent Searches take 7th and 8th graders with IQs over about 125, and have them take the SAT or ACT at around age 12 (instead of the usual age of 17-18). The ones whose age-12 SAT/ACT scores map to 99.5th+ percentile (for age 12, based on the Talent Searches’ by-now-huge database) get into the legacy summer programs for ages about 12-16. (There have been additional summer programs since created.)

    SET, initially called SMPY, is the study (based at JHU, but selects from all Talent Search participants worldwide), that’s also been going on since the ’70s, of those Talent Search participants who have scored +2 sds (for 12th graders) on an SAT subtest before age 13. This probably maps to about 99.99th percentile (for 12-year-olds). Australian Terry Tao is a participant. So is Eugene Volokh (the legal scholar). Among others.

    The study you’re discussing does not include all summer program qualifiers, but does include more than just SET members, mapping to about the 99.97th percentile–based on those same age-12 SAT/ACT scores:

    Individuals were selected from the United States for participation in the HiQ study on the basis of performance on the Scholastic Assessment Test (SAT) or American College Test (ACT) taken at age 12 rather than the usual age of 18 years. A composite that aggregates verbal and mathematics SAT and ACT scores correlates >0.80 with intelligence tests and it is estimated that the TIP program recruits from the top 3% of the intelligence distribution. For this study (HiQ), cases were selected from the top 1% of these TIP individuals, representing approximately the top 0.03% of the intelligence distribution.

    And they do mention the issue of adolescent IQ being less heritable than adult IQ.

    (Actual study:

    Now, you mentioned the study’s IQ estimates…

    The issue with IQ estimates is that g is not actually normally distributed. It looks normally distributed near (where “near” is “within 2 standard deviations of”) the mean, which is why IQ tests try to map their raw scores into a standardized score based on a normal distribution. But that raw score distribution has fat tails.

    Example #1: Ratio IQ. If you take a representative sample of the population of children of a given age, give them a mental test and assign them scores of the form ((mental age)/(chronological age))*100 –IOW, ratio IQs–their scores have a mean of about 100 (of course) and, if you use the formula for calculating standard deviation, it outputs the number 16. And ratio IQ scores look normally distributed between about 68-132. But…well, most people interested in IQ already know what happens at the high end. There have been multiple kids with ratio IQs above 220. Multiple kids with IQs 7.5 formula-generated standard deviations above the mean. In a normal distribution, such a score would occur at a rate of 1 in 31 trillion. So…it’s obviously *not* a normal distribution. There are more high scorers than a normal distribution would predict.

    Example #2: Consider the CTY/CTD/TIP “out of level testing” (giving smart kids tests designed for older kids so as to test the limits of their ability). Based on their decades-long databases, it’s pretty certain that SET members really are 1 in 10,000. For 12-year-olds. But…there exist people who qualify for SET at age eight. (Such as poor Terry Tao, who if he’s reading this must be getting pretty annoyed at me using him as a fricking example instead of the human he is. Sorry, man.) The normal curve doesn’t predict this; it predicts that nobody exists who could ever qualify for SET as young as that. (That last sentence was an extreme oversimplification, but true; see “1 in 31 trillion” above.)

    So, do the people in this study have IQs over 170? Well…where do you want to anchor your extrapolation? At the high end, as you increase rarity, ability to handle cognitive complexity (g) just…increases faster than a normal distribution predicts.

    If you take a representative sample of the population (such as an IQ test’s norm sample), calculate the mean and standard deviation of its raw score, and use these to generate “standard score” IQs (anchoring the extrapolation at the 50th percentile)…you will get “too many” high scores. And this is true even if you take a sample not of the entire population, but just of those who you know do better than 99 out of 100 test-takers (anchoring the extrapolation at the 99.0th percentile)–as the SMPY/SET researchers did in creating the “IQ-estimating” procedure also used by this study. No matter whether you anchor your extrapolation at the 50th percentile or the 99th, more high scorers will appear than the normal curve predicts.

    So the question is *why* isn’t it normally distributed?

    …damned if I know. ;)

    (Personally, I also don’t care. I care about helping kids who don’t fit standard curricular expectations. Which includes those with extreme IQs–as well as those with specific learning disabilities or who come from a very different culture than the curriculum expects.)


    • Thanks for the details; very helpful.
      My first response is simply that either the tests are bad at measuring very high end IQs, or humans just aren’t normally distributed. After all, lots of things aren’t.

      Hitting the right balance for those kids–who often mature at different rates than their peers, too–is really tricky.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s