A few friends have pinged me on this piece in Slate, Double Inanity: Twin studies are pretty much useless. The headline is bold, but the piece is just a sloppy mishmash. It’s really something amenable for a major “fisking,” but I generally don’t like doing that sort of thing, because it doesn’t seem a optimal allocation of time (though I have to note that the author seems to be implicitly using a colloquial form of the concept of heritability, which I think is going to confuse an already naive audience). A lot of the article is taken up with criticisms of political scientist John Alford’s behavior genetic findings on the heritability of ideology. I’ve had personal communication with other researchers in this area who actually are broadly critical of Alford’s methodology, but they’re still strongly invested in using genomics and behavior genetics to explore this issue. In other words, rejecting Alford’s conclusions does not entail that you just reject twin studies, as such.
Of course there’s a lot more to twin studies than behavior genetics. Luke Jostins at Genomes Unzipped had a long post up defending the methodology, with caveats, in estimating heritability. But what I find perplexing is that in many ways we are right now confirming the broad results of twin studies using genomic methods. I blogged a paper on the heritability of I.Q. a few weeks ago, but here’s a paper from 2006, Assumption-Free Estimation of Heritability from Genome-Wide Identity-by-Descent Sharing between Full Siblings:
…. Following the classic paper by R. A. Fisher in 1918, the estimation of additive and dominance genetic variance and heritability in populations is based upon the expected proportion of genes shared between different types of relatives, and explicit, often controversial and untestable models of genetic and non-genetic causes of family resemblance. With genome-wide coverage of genetic markers it is now possible to estimate such parameters solely within families using the actual degree of identity-by-descent sharing between relatives. Using genome scans on 4,401 quasi-independent sib pairs of which 3,375 pairs had phenotypes, we estimated the heritability of height from empirical genome-wide identity-by-descent sharing, which varied from 0.374 to 0.617 (mean 0.498, standard deviation 0.036). The variance in identity-by-descent sharing per chromosome and per genome was consistent with theory. The maximum likelihood estimate of the heritability for height was 0.80 with no evidence for non-genetic causes of sib resemblance, consistent with results from independent twin and family studies but using an entirely separate source of information. Our application shows that it is feasible to estimate genetic variance solely from within-family segregation and provides an independent validation of previously untestable assumptions….
The point is that before modern genomics one had to assume relatedness, and from these assumptions derived various estimates as to the heritability of a trait when taking into account patterns of genetic relatedness and phenotypic similarity. Today genetic relatedness can be pinned down on an individual basis with much greater precision. Two of my siblings for example have a relatedness of 0.42, instead of 0.50. This does not mean that all of the criticisms in the Slate piece are invalid, but there was a slapdash throw-the-kitchen-sink feel to it all.
Finally, I want to point out that just because a trait is not substantially heritable, it does not mean it is not substantially biological. This is the possible case with male homosexuality. I think as a scientific matter this is a huge issue, but when it comes to public policy it is much less important because the relevance of the genetic origin of a trait is that that necessarily means that it is biologically specified. If the biological specification is through random developmental processes or infection with pathogens, it means that it isn’t heritable and genetic, but it doesn’t mean that once the trait has fixed it is any less plastic.
Razib Khan’s degrees are in biochemistry and biology. He has blogged about genetics since 2002, previously worked in software development, is an Unz Foundation Junior Fellow and lives in the western US. He loves habaneros.
August 24th, 2011 at 5:26 pm
i want to note that this is one of those articles that you can do a line by line criticism. there wasn’t one or two points of contention, but one after another. so if you are asking me why i didn’t address a point, well, there’s too much to address (e.g., the name-checking of CNVs instead of epigenetics was at least novel in this genre).
August 24th, 2011 at 5:43 pm
Agreed… just a really crappy, almost polemical, article from Slate (even given the few valid points) that will unfortunately reach a very large audience and lead to a lot of misunderstanding… (luckily commenters at the article are mostly critical of the author).
August 24th, 2011 at 7:19 pm
Brian Palmer’s identical twin … is a moron.
August 25th, 2011 at 12:18 pm
Shecky R –
Yes, it’s encouraging to read the Slate commentariat. Most people wrote in to take issue with one or another of the points that Palmer didn’t quite grasp.
BTW, the trade e-mag GenomeWeb summarizes Palmer’s piece, and linked to this post as an instance of its reception in the blogosphere. Twin-Sanity!.
August 25th, 2011 at 6:49 pm
That was the worst article that I’ve read in a long time. I’ve only recently started to read about these studies, but his point regarding the copy length variations and mtDNA mutations struck me as a big “who cares?” rather than the devastating blow he was attempting to deliver. The correlation between genetic relatedness and general intelligence/various personality traits/etc have already been established in these studies, which suggests that the lack of completely genetic similarity among monozygotic twins leads to lower heritability estimates than what is accurate. He appears to have written the article for the apparent purpose of defending the environmentalist position, and yet he unintentionally does the exact opposite.
August 26th, 2011 at 12:23 pm
Twin studies underestimate heritability only if you subscribe to the crude notion that the effect of genes is additive, i.e., keeping “environments” the same, the more similar two people are genetically the more alike they will be. This ignores everything we now know about the way genes work.
Genes are not self-activating: they do not turn themselves on and produce traits. Genes do not, in fact, produce anything. Genes are turned on and off by the epigenome in response to environmental inputs. If you are inclined to doubt this, then consider: If all the cells in our body are supposed to contain identical DNA, how do you account for the existence of different tissues and cells types (answer: the epigenome). Twins have been shown to have significantly different epigenomes in the womb. Does this show that twin studies underestimate heritability? Of course not. It shows, rather, that these studies are out of touch with advances in molecular genetics.
You are all committed to a crude, 19th century conception of the relationship between genotype and phenotype. Twin studies are based upon a scientific paradigm that has remained unchanged since the time of Galton. Problem is, recent advances in molecular genetics are bringing about a paradigm shift in the science of genetics.
Twin studies make it seem all so easy: All you need is a large data set that has zygosity status and statistical software and any undergraduate can unravel the biomolecular basis of human behavior.
Would that it were so easy.
August 27th, 2011 at 10:12 am
@Andrea Cantor
The relationship between biological interaction and genetic addativity on the variance component scale is more complex than I think you realise, and the common terms between the two (additivity, interaction, etc) seem to have thrown off your intuition. The fact is that virtually all the assumptions of twin studies have been extensively tested, and they have all been found to be relatively good approximations (statistically) the behaviour of genetic risk in populations, in certain relatively well characterised circumstances. This doesn’t mean that biological systems aren’t full of interaction, or that epigenetic effects are not important, or that genetic effects are stationary over time, or that interactions between genes and environment do not exist. It just means that your intuitive feel that these effects result in statistical non-additivity on the genetic risk scale is wrong.