Over the past week I’ve been asked via email and on message boards about about David Shenk’s new book, The Genius in All of Us: Why Everything You’ve Been Told About Genetics, Talent, and IQ Is Wrong. Since I haven’t read the book I can’t really comment, but I did finally listen to Will Wilkinson’s interview with Shenk on bloggingheads.tv. It seems to me that Will exhibited more clarity and precision in one sentence in relation to the term heritability than Shenk did in 10 minutes. It is true there are many people who don’t understand that 80% heritable does not mean that a trait is “80% genetic.” In fact, I really don’t know what a trait being “80% genetic” means in a precise sense, but I also know that long time readers of this weblog do fall into this trap.
Instead of reading Shenk’s book I strongly suspect that people might gain some more genuine insight about heritability and the genetics of complex traits by looking at what we know about height. We don’t know much in terms of the underlying genes; height seems to be controlled by many genes of small effect. But, we do know that in the developed world, where nutritional intakes have saturated, height is about ~80% heritable. That is, most of the variation in the population can be accounted for by variation in genes. There are probably gene-environment interactions in regards to the trait of height. For example, there may be individuals whose genotypes are more sensitive to nutritional deprivation than others, so that changing uniform nutritional intakes across a population may not change just the median of the distribution, but also the general shape. But those interaction effects are obviously not as important today in the developed world where malnutrition is very rare.
At least judging by the conversation with Wilkinson, and the title of the book, Shenk seems to want to spotlight people who are many standard deviations from the norm. For example, Mozart and Michael Jordan are arguably not 1 in 100, or even 1 in 1,000, in regards to their domains of virtuosity. I think that focusing this far out to the tails is interesting, and makes for good narrative as one can populate it with illustrative anecdotes, but on any given quantitative trait most people are going to be much closer to the median. Variation on the margins of the normal are very significant, and all too often ignored. In here that I think that the simplest models have the most utility. So you want to complexify, just focus on the outliers….