The Unz Review - Mobile
A Collection of Interesting, Important, and Controversial Perspectives Largely Excluded from the American Mainstream Media
 TeasersGene Expression Blog
Hoping for High Heritability of IQ
🔊 Listen RSS
Email This Page to Someone

 Remember My Information



=>

Bookmark Toggle AllToCAdd to LibraryRemove from Library • BShow CommentNext New CommentNext New ReplyRead More
ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
AgreeDisagreeLOLTroll
These buttons register your public Agreement, Disagreement, Troll, or LOL with the selected comment. They are ONLY available to recent, frequent commenters who have saved their Name+Email using the 'Remember My Information' checkbox, and may also ONLY be used once per hour.
Ignore Commenter Follow Commenter
Search Text Case Sensitive  Exact Words  Include Comments
List of Bookmarks

twohigIQkidsThe new paper in PNAS, Common genetic variants associated with cognitive performance identified using the proxy-phenotype method, has resulted in a fair amount of reaction. One of the major things that people grasp onto is that the effects of the variants in question are extremely smaller (here’s an FAQ for the current paper). Each variant is associated with a 0.3 increment or decrement in IQ, where the average IQ is 100 and the standard deviation is ~15 points. These results are not surprising, as the problems with earlier attempts to fix upon a genetic region which explains a great deal of variation in intelligence in the normal range have not been successful (i.e., they fail replication, so probably just a false positive). Taking these results at face value many have wondered what the big deal is, as the associations here have such a small impact.

First, a small effect does not preclude important practical consequences. The locus HMGCR has been implicated in variation in cholesterol levels at 0.1 standard deviation, but that is the locus that statins target. Does this mean that we can make a “genius pill” in the future? I’m moderately skeptical, and obviously there are major ethical issues with this. But, this sort of research shows that it may be possible, and in this big wide world of ours knowledge is hard to keep under control. As a normative matter I’m in the always better to know category for almost everything. So big surprise I have no issues with this line of research.

There is a second issue of more practical relevance, and that is that many people wish to reject a heritable component for intelligence. To be clear it is robust science that intelligence is 0.3 to 0.7 heritable. That means that 30 to 70 percent of the variation in intelligence in the population is due to variation in genes. Because the trait is highly polygenic, on the order of thousands of loci controlling variation in intelligence, it is difficult to pick any particular signal. But very few scientists are under the illusion that intelligence is not at least moderately heritable. A good analogy here is height, which is highly heritable, and controlled by many genes of small effect (the genetic architecture here is moderately more tractable from what I can tell). But for many people, especially in the public, they “need a gene.” It makes the abstract, ratio of additive genetic variance over total phenotypic variance, concrete.

But I find it more interesting that some are spinning this as a support for the low heritability of IQ, and the importance of environment. Personally I wish for my children that environment was less important, not more. The reason is simple: in a behaviour genetic sense we really don’t know what we’re talking about when we say “environment.” The Invisible Gorilla has a lot of illustrations on how tools and techniques which make us “smarter” really don’t work (or, their efficacy has not been scientifically validated). The same for infants and children. Obviously malnutrition and abuse are going to cause problems in relation to development, but the sort of “enriching” activities and practices de rigueur among upper middle class parents probably are irrelevant to the final outcome of the trait in question (this is clear when you look at the high level of variation cross-culturally, with some “best practices” being contradictory, but the results are the same nonetheless).

The-Nurture-Assumption-Harris-Judith-Rich-9780684857077 The best way to think about it is that “environment” is just noise in your model. It is the genetic component you can control, or at least use to predict. Though heritability is a population wide statistic, it has some relevance for individuals. The mid-parent value of a trait for the parents can help you gauge your expectations for your offspring. When you standardize for sex the height of parents can tell you whether to expect tall or short offspring. This is not guaranteed, as there is a high standard deviation around the expected value, even for a highly heritable trait like height (the correlation between full-siblings for height is ~0.50). But, it does load the die. The correlation of IQ between full-siblings is also on the order of ~0.50. Remember here that environment, the noise parameter, changes your expected value. Since this isn’t heritable it drives the phenotype of the offspring back to the population mean. If IQ is less heritable, say 0.30, then if you and your spouse are deviated away from the mean, you can expect your children to regress back to the population mean, since they won’t inherit the magic mix of factors which resulted in high IQ. In contrast, if IQ is heritable on the order of 0.70, then you can update your expectations so that your children will be more likely to resemble you, assuming you are deviated from the norm.

Perhaps I’m a narcissist, but I want my children to be like me in cognitive profile. It makes it easier for me to understand where they are coming from. If I thought that I could as a parent control the environmental outcomes with a high degree of certainty I might be more sanguine about low heritability, but that’s not my hunch about this trait. Low heritability of intelligence to me connotes a flight back to mediocrity and a total lack of control. High heritability in contrast allows one to reclaim control, because you choose your spouse and you have a sense of their realized phenotype. Obviously this is conditional on where you stand on the distribution. So I emphasize the “I.” But many people at the higher end of the IQ distribution seem to want lower heritability, because they perceive that they can control outcomes through manipulation of environment. I’m not confident of this at all. Sometimes flighty academic abstractions can have real consequences in the choices we make in this world. This is one.

 
• Category: Science • Tags: IQ 
Hide 12 CommentsLeave a Comment
Commenters to Ignore...to FollowEndorsed Only
Trim Comments?
  1. >To be clear it is robust science that intelligence is 0.3 to 0.7 heritable. That means that 30 to 70 percent of the variation in intelligence in the population is due to variation in genes.

    I’m not a statistician, and I’ve been confused when I read debates on this topic. Aren’t you meant to square the heritability to get the variance? That would result in 9 – 49% of the variation explained.

    Is there a good writeup somewhere of the topic?

    • Replies: @Anonymous
    This is very accessible to the layperson as an intro to the subject:

    On the genetic architecture of intelligence and other quantitative traits by Steve Hsu:

    http://arxiv.org/abs/1408.3421



    And this is a lot of the same info in video form- a googletech talk- if you prefer.

    https://www.youtube.com/watch?v=62jZENi1ed8
    , @Anonymous

    I’m not a statistician, and I’ve been confused when I read debates on this topic. Aren’t you meant to square the heritability to get the variance? That would result in 9 – 49% of the variation explained.
     
    Heritability is a variance component, not a correlation. Heritability is the square of the correlation between genotype and phenotype. This is most easily understood when considering monozygotic twins reared apart (MZA twins). The phenotypic correlation between MZA twins is equal to the heritability, or genetic variance, of the trait. This is easy to see when depicted as a path model:

    P1
    /\
    |
    | a
    |
    A
    |
    | a
    |
    \/
    P2

    In this model P1 and P2 are phenotypic values (e.g., IQ scores) of MZA twin pairs, A is a latent variable representing genetic influences on the phenotype, and a is a regression weight representing the causal effect of A on P1 and P2.

    Path tracing rules tell us that the correlation between P1 and P2 is equal to a^2. Because a^2 is also equal to the variance that A explains in P1 and P2 (because it's the square of the correlation between A and P1 or P2), we conclude that heritability is equal to the phenotypic correlation between P1 and P2. The phenotypic correlation between MZA twins is used to infer the effect of a third variable (i.e., genes) shared by both twins.
  2. it is the proportion of total variation due to variation in genes. so heritability in the narrow sense:

    h2 = Va/Vp, where Va = additive genetic variance and Vp = phenotypic variance.

    read introduction to quantitative genetics if wikipedia etc. does not suffice. heritability is related to correlation (which is what i think you are talking about), but not the same.

  3. @ckp
    >To be clear it is robust science that intelligence is 0.3 to 0.7 heritable. That means that 30 to 70 percent of the variation in intelligence in the population is due to variation in genes.

    I'm not a statistician, and I've been confused when I read debates on this topic. Aren't you meant to square the heritability to get the variance? That would result in 9 - 49% of the variation explained.

    Is there a good writeup somewhere of the topic?

    This is very accessible to the layperson as an intro to the subject:

    On the genetic architecture of intelligence and other quantitative traits by Steve Hsu:

    http://arxiv.org/abs/1408.3421

    And this is a lot of the same info in video form- a googletech talk- if you prefer.

  4. @Razib Khan @StevieMac

    Thanks both of you!

  5. There seems to be good evidence that it’s highly polygenic, but I think Kevin Mitchell’s logic is right that we should expect most mutations to be deleterious, so it might be better to look for “genes for stupidity”. Greg Cochran has written about how merely removing de novo mutations in an attempt to revive Neandertals could put them multiple standard deviations above us. I’m not sure what relative frequency Mitchell would implicitly require for these mutations not to be “common” though.

  6. Beautifully said! Great post!

    But I find it more interesting that some are spinning this as a support for the low heritability of IQ, and the importance of environment.

    Three guesses, right?

  7. I was impressed by the FAQ for that paper. Somebody with very good writing skills put a lot of effort into it.

    I got the same idea from it that you had; that each variant adds about .3 points of IQ. So I did a little calculating. If you assume that a “plus” common allele frequency is 50% (which apparently they did, from the FAQ), it would take an excess of 50 of the “plus” alleles for a person to be one standard deviation above average intelligence. Leaning on Wolfram|Alpha to calculate probabilities, it looks like you would need around 2,000 of these genes to do it, if the effects are additive.

    Now, Imma gonna assume that they actually did find alleles that have a much higher effect on intelligence than average, but it is a pretty low percentage of the genes they tested, even though they identified the most likely candidates first. So it is really unlikely that there are 2,000 more common alleles lurking out there in the unexplored wilderness that can explain intelligence variation in an additive way.

    Psst! It’s not additive.

  8. “Personally I wish for my children that environment was less important, not more. ”

    First genuine belly-laugh of the day. I could not agree more.

  9. But many people at the higher end of the IQ distribution seem to want lower heritability, because they perceive that they can control outcomes through manipulation of environment. I’m not confident of this at all.

    Some people can be confident of this manipulation, if they have the money. I.E., their children´s outcomes in life can get a boost from Ivy league legacy admissions, family networks or even businesses, etc.

  10. Anonymous • Disclaimer says: • Website
    @ckp
    >To be clear it is robust science that intelligence is 0.3 to 0.7 heritable. That means that 30 to 70 percent of the variation in intelligence in the population is due to variation in genes.

    I'm not a statistician, and I've been confused when I read debates on this topic. Aren't you meant to square the heritability to get the variance? That would result in 9 - 49% of the variation explained.

    Is there a good writeup somewhere of the topic?

    I’m not a statistician, and I’ve been confused when I read debates on this topic. Aren’t you meant to square the heritability to get the variance? That would result in 9 – 49% of the variation explained.

    Heritability is a variance component, not a correlation. Heritability is the square of the correlation between genotype and phenotype. This is most easily understood when considering monozygotic twins reared apart (MZA twins). The phenotypic correlation between MZA twins is equal to the heritability, or genetic variance, of the trait. This is easy to see when depicted as a path model:

    P1
    /\
    |
    | a
    |
    A
    |
    | a
    |
    \/
    P2

    In this model P1 and P2 are phenotypic values (e.g., IQ scores) of MZA twin pairs, A is a latent variable representing genetic influences on the phenotype, and a is a regression weight representing the causal effect of A on P1 and P2.

    Path tracing rules tell us that the correlation between P1 and P2 is equal to a^2. Because a^2 is also equal to the variance that A explains in P1 and P2 (because it’s the square of the correlation between A and P1 or P2), we conclude that heritability is equal to the phenotypic correlation between P1 and P2. The phenotypic correlation between MZA twins is used to infer the effect of a third variable (i.e., genes) shared by both twins.

  11. […] like Razib Khan (who I think doesn’t identify with HBD now, but I think has in the past), who leans toward high heritability of many traits, but whose racial views are harder to read (even if he is conservative), or people […]

  12. […] Razib on heritability. Low heritability doesn’t make it easier to shape our children – once we’re above some minimum thresholds, we don’t really know what works. […]

Comments are closed.

Subscribe to All Razib Khan Comments via RSS