4 ms·

Among twin-twin studies of the effects of nature/nuture, I recall mathematical ability as being the most highly heritable at something like 50%. Even eye color

by Shinkei 12y ago

Among twin-twin studies of the effects of nature/nuture, I recall mathematical ability as being the most highly heritable at something like 50%. Even eye color is not that high, and so it seems to be true that intelligence (at least in maths) is very heritable.

bhouston 12y ago
> I recall mathematical ability as being the most highly heritable at something like 50%. Even eye color is not that high Actually, heritability of eye color is 98%. A somewhat complete listing of heritability according to SNPedia (which doesn't include intelligence): http://www.snpedia.com/index.php/Heritability http://www.snpedia.com/index.php/Heritability
- cariaso 12y ago
  http://www.snpedia.com/index.php/Eye_color http://www.snpedia.com/index.php/Eye_color does have extremely high heritability, but the mechanism is more complex than what we were usually mistaught in schools, which I think is the origin of Shinkei's comment. http://snpedia.com/index.php/Intelligence http://snpedia.com/index.php/Intelligence is not being well captured by GWAS. All current associations are probably false positives. I've not seen any evidence specific to mathematical ability.
  - gwern 12y ago
    > All current associations are probably false positives. Why do you say that? The recent results like Rietveld et al 2013 have used stringent false-positive rates precisely because of the early rash of false positives in the early '00s, and the hits seem to be replicating pretty well.
    - cariaso 12y ago
      I'm basing it on http://scholar.harvard.edu/files/laibson/files/most_reported_genetic_assoc_13dec2011.pdf http://scholar.harvard.edu/files/laibson/files/most_reported... but will concede that I misquoted and that 'Most' vs 'All' is an important distinction. The most current information I see is Rietveld published 10 days ago. http://pss.sagepub.com/content/early/2014/10/06/0956797614545132.abstract http://pss.sagepub.com/content/early/2014/10/06/095679761454... and that paper claims to have found an explanation for ≈2% of educational attainment and cognitive function. None of the reported snps (rs9320913, rs11584700, rs4851266) have been mentioned in any other publication. So I don't yet see a basis for "the hits seem to be replicating pretty well". I do not doubt that genetics plays a significant role in intelligence, but I expect that intelligence, and the proxy of educational attainment, is an overly broad phenotype. I also expect that epistatic effects may be important and would necessitate unrealistically massive sample size or a very different experimental design.
      - Shinkei 12y ago
        I appreciate the replies. I learned a lot just reading some of these sources posted. I think I misused heritability regarding eye color and my numbers are probably dated.
      - gwern 12y ago
        > I'm basing it on http://scholar.harvard.edu/files/laibson/files/most_reported.. http://scholar.harvard.edu/files/laibson/files/most_reported.... but will concede that I misquoted and that 'Most' vs 'All' is an important distinction. Yes, since people upped their game precisely because of that paper's findings... > None of the reported snps (rs9320913, rs11584700, rs4851266) have been mentioned in any other publication. So I don't yet see a basis for "the hits seem to be replicating pretty well". What do you mean? As I recall, Rietveld et al 2013 used internal replication to test its hits, which worked out well. Your link is itself a another replication of Rietveld et al 2013, is it not? (What exactly did you think the abstract meant when it talked about "A recent genome-wide-association study" or "The study"?) And then there's "Genetic Variation Associated with Differential Educational Attainment in Adults Has Anticipated Associations with School Performance in Children" , Ward et al 2014; and I believe "Educational attainment-related loci identified by GWAS are associated with select personality traits and mathematics and language abilities", Zhu et al 2015 is also relevant.
        cariaso 12y ago
        Rietveld replicating Rietveld doesn't do it for me in an area so lacking of success stories. Ward and Zhu are new to me and I am reading them now. Time will tell, I have no doubt.
        gwern 12y ago
        > Rietveld replicating Rietveld doesn't do it for me in an area so lacking of success stories. I think this is unreasonable. This is not a lone author working with a set he hand-collected and has had unlimited freedom to monkey with. This is the lead author on a project involving scores of authors analyzing vast datasets pooled together by like a dozen distinct organizations.

by Shinkei 12y ago

bhouston 12y ago

> I recall mathematical ability as being the most highly heritable at something like 50%. Even eye color is not that high Actually, heritability of eye color is 98%. A somewhat complete listing of heritability according to SNPedia (which doesn't include intelligence): http://www.snpedia.com/index.php/Heritability http://www.snpedia.com/index.php/Heritability

cariaso 12y ago
http://www.snpedia.com/index.php/Eye_color http://www.snpedia.com/index.php/Eye_color does have extremely high heritability, but the mechanism is more complex than what we were usually mistaught in schools, which I think is the origin of Shinkei's comment. http://snpedia.com/index.php/Intelligence http://snpedia.com/index.php/Intelligence is not being well captured by GWAS. All current associations are probably false positives. I've not seen any evidence specific to mathematical ability.
- gwern 12y ago
  > All current associations are probably false positives. Why do you say that? The recent results like Rietveld et al 2013 have used stringent false-positive rates precisely because of the early rash of false positives in the early '00s, and the hits seem to be replicating pretty well.
  - cariaso 12y ago
    I'm basing it on http://scholar.harvard.edu/files/laibson/files/most_reported_genetic_assoc_13dec2011.pdf http://scholar.harvard.edu/files/laibson/files/most_reported... but will concede that I misquoted and that 'Most' vs 'All' is an important distinction. The most current information I see is Rietveld published 10 days ago. http://pss.sagepub.com/content/early/2014/10/06/0956797614545132.abstract http://pss.sagepub.com/content/early/2014/10/06/095679761454... and that paper claims to have found an explanation for ≈2% of educational attainment and cognitive function. None of the reported snps (rs9320913, rs11584700, rs4851266) have been mentioned in any other publication. So I don't yet see a basis for "the hits seem to be replicating pretty well". I do not doubt that genetics plays a significant role in intelligence, but I expect that intelligence, and the proxy of educational attainment, is an overly broad phenotype. I also expect that epistatic effects may be important and would necessitate unrealistically massive sample size or a very different experimental design.
    - Shinkei 12y ago
      I appreciate the replies. I learned a lot just reading some of these sources posted. I think I misused heritability regarding eye color and my numbers are probably dated.
    - gwern 12y ago
      > I'm basing it on http://scholar.harvard.edu/files/laibson/files/most_reported.. http://scholar.harvard.edu/files/laibson/files/most_reported.... but will concede that I misquoted and that 'Most' vs 'All' is an important distinction. Yes, since people upped their game precisely because of that paper's findings... > None of the reported snps (rs9320913, rs11584700, rs4851266) have been mentioned in any other publication. So I don't yet see a basis for "the hits seem to be replicating pretty well". What do you mean? As I recall, Rietveld et al 2013 used internal replication to test its hits, which worked out well. Your link is itself a another replication of Rietveld et al 2013, is it not? (What exactly did you think the abstract meant when it talked about "A recent genome-wide-association study" or "The study"?) And then there's "Genetic Variation Associated with Differential Educational Attainment in Adults Has Anticipated Associations with School Performance in Children" , Ward et al 2014; and I believe "Educational attainment-related loci identified by GWAS are associated with select personality traits and mathematics and language abilities", Zhu et al 2015 is also relevant.
      - cariaso 12y ago
        Rietveld replicating Rietveld doesn't do it for me in an area so lacking of success stories. Ward and Zhu are new to me and I am reading them now. Time will tell, I have no doubt.
        gwern 12y ago
        > Rietveld replicating Rietveld doesn't do it for me in an area so lacking of success stories. I think this is unreasonable. This is not a lone author working with a set he hand-collected and has had unlimited freedom to monkey with. This is the lead author on a project involving scores of authors analyzing vast datasets pooled together by like a dozen distinct organizations.