Heritability is a fraught topic. It comes up repeatedly on this weblog, but even long time readers can be confused as to its implications, as evidenced by the incorrect inferences made from their own understanding of the concept. The most common problem is that too often people think that heritability is just a scienced up version of the colloquial idea of some traits being "more genetic" or "less genetic." It's not that at all. Traits totally specified in their details by genetic pathways can be non-heritable. That's because heritability looks at the association between parents and offspring on a trait and attempts to separate the population-wide proportion of the variation attributable to genes and not attributable to genes. When you have a genetically specified trait, like the number of human fingers, you have no real variation within the population to work with (with some rare exceptions). It doesn't make sense to talk about the heritability of the number of fingers, because this is a fixed trait in the human species. In contrast, height is a perfect trait to illustrate heritability. Unlike behavioral or cognitive traits its measurement is clear, distinct, and uncontroversial. Additionally, there's a normal distribution of the trait. By that, I mean that there is a bell curve from tall to short, with a median at the peak of the distribution. Not only does height vary within the population, but it varies across populations, and it varies within families. When considering the heritability of height there's a lot to grapple with and ponder for a relatively easy to measure characteristic. Intuition will tell you that parents and offspring tend to correlate positively in terms of height, but the trait is imperfectly correlated. Comparisons between identical and fraternal twins can allow us to partition the rough effects of genes vs. non-genes on the trail value. Identical twins share ~100% of their genes, while fraternal twins share ~50% of their genes, so the correlation in trait value should diminish in proportion to the decreased genetic similarity and the effect of genes on the trait value. In plain language, if genes don't account for any of the variation in the trait value then identical and fraternal twins should exhibit the same correlations on a trait. In contrast, if genes control almost all the variation on the trait value then identical twins should be rather close in value, while fraternal twins far less, because the twins in the second case differ a great deal genetically while they do not in the former case. But even with a trait like height environment matters, more or less. Again, intuition tells you that if you starve a child during a critical period they may never attain their height as might be inferred from their genetic potential. Therefore heritability is informative in light of background conditions. In the developed world the heritability of height is ~0.80 to ~0.90, which means that 80-90% of the variation in height can be explained by variation in genes. This is not so in other regions of the world, where environmental variables such as nutritional stress loom larger. There is no fixed proportion by which height is "more genetic" or "more environmental." There's much more that could be said. But much of it has been said, and by those better able to say more. So I'll move to my value-add: what is the profile of those who know and don't know what heritability is? In the survey I asked: "Do you know that narrow-sense heritability is the proportion of the phenotypic variance due to the additive genetic variance?" The terminology may seem opaque to some of you, but to those who knew what heritability was it wouldn't be. ~33% of you out of ~550 (the survey is still live) recognized what narrow-sense heritability was by the definition I offered. I was frankly nicely surprised, since I know many of you skim for topics of particular interest to you rather than "deep dive" into quantitative genetics, which I don't talk much about in depth in any case. The results from the survey will be analyzed in three parts: - In the first knowledge of heritability is a function of another variable. In other words below you a see table which shows that 34 percent of males recognized the definition of narrow-sense heritability and 29 percent of females did. The average is ~33 percent because 80 percent of respondents were male vs. 20 percent female. Remember that the proportion in the aggregate pool is ~33 percent. - The second section focuses on the differences between people who know what narrow-sense heritability is, and those who do not. So in this case you have other factors as a function of heritability as a categorical variable. - A final section examines some scatter plots of estimated height vs. IQ heritabilities. An interesting point to remember here is that many people who weren't totally clear on what narrow-sense heritability was nevertheless offered up a heritability estimate, probably because they got the gist of what the value was.
In the plots below red = those who know what narrow-sense heritability is, and blue = those who do not.
Knows what “narrow-sense heritability” is
Male34
Female29
Some university16
Bachelors30
Masters30
Advanced degree53
Supports abortion on demand32
Rejects abortion on demand33
Accepts supernatural25
Rejects supernatural34
Mathematics background34
Computer science background30
Physical sciences background27
Life sciences background60
Psychology background38
Economics background37
Anthropology background48
Other social sciences background19
Philosophy background36
Other humanities background23
Law background45
Medicine background50
Has engaged in recreational genomics47
Has not engaged in recreational genomics28
Can program34
Can not program32
Is interested in transhumanism29
Is not interested in transhumanism34
Need to always be on guard against genetic determinism35
Genetic determinism is a concern, but often overblown34
We don't have to be concerned about genetic determinism34
Biological sex differences in behavior trivial6
Biological sex differences in behavior very modest29
Biological sex differences in behavior somewhat significant32
Biological sex differences in behavior very significant42
Biological race differences in behavior trivial25
Biological race differences in behavior very modest32
Biological race differences in behavior somewhat significant41
Biological race differences in behavior very significant40
Biological within population differences in behavior trivial18
Biological within population differences in behavior very modest27
Biological within population differences in behavior somewhat significant39
Biological within population differences in behavior very significant46
Biological across sibling differences in behavior trivial21
Biological across sibling population differences in behavior very modest30
Biological across sibling population differences in behavior somewhat significant38
Biological across sibling population differences in behavior very significant40
Has made a non-trivial edit to Wikipedia38
Has not made a non-trivial edit to Wikipedia30
Now let's look at the scatter plot of height and IQ heritabilities, with shading proportional to the N within a bin.
The correlation between the two estimates is ~0.52, ~0.60 for those who know what narrow-sense heritability is, and ~0.47 for those who don't. It's clear that a lot of people who didn't know the definition for heritability just guessed 0.50 for both traits (reasonable I think). Just to be clear, in developed nations the heritability for height is pretty well known to be in the ~0.80-0.90 range. That explains the peak there, especially for those who know what narrow-sense heritability is. For IQ there are a large range of numbers, from ~0.30 up to ~0.80.
