>Did I use the word heritability? That is a total red
>herring. I was talking about the way that traits,
>including behaviors, are manifestations of genes in an
>environment. Heritability concerns a totally different
>question, the likelihood that the trait will show up
>in the same environment across generations. If you
>vary the environment significantly, you can't even
>formualte the question, How heritable Is It? because
>you aren't controlling for the environment -- which
>makes my point again. "Genetic" does not and cannot
>mean, even the cases of stuff that is "genetically
>determined" in a far more general way thana ny
>behaviorm sucha s height or eye color -- rigidly
>manifested the same way in all environments. And there
>is really no coherent way to ask, well, What % is it
>"genetically determined"? You can of course ask if
>traits are manifested in most environments, and as
>noted it is clear that in many cases the genetic
>contribution imposes limits on the extent of
>variation. That is all that you can say. And this is
>Genetics 101 -- if people didn't want to abuse science
>for ideological purposes, no one woul;d even prertend
>the question mde any sense.
>
I would like to second Andie's comment here. "Heritability" is one of
the most mis-cited and mischaracterized measures in genetics.
Some proponents of strong biological-determinist claims, like Arthur Jensen, tend to present a high heritability value as evidence of genetic fixity-- for example, if IQ had a heritability of 0.7 to 0.8, then they argue that most of one's IQ is set by one's genetic inheritance, and only the remaining 20-30 percent changes by environment. (Funniest example was in _The Bell Curve_, where Murray and Herrnstein present an accurate, one-sentence description of heritability-- and then argue for what it is precisely _not_.) But heritability is an estimate, drawn from distinct groups, and can't be applied that way.
Here's a quick description. In any species, traits are going to vary-- height at adulthood, tail length in mice, etc. And we have to estimate how much of this variance may be due to genetics, and how much is due to environmental factors (like nutrition). Heritability is an _estimate_ derived from that variance. It can only apply to populations, and more specifically, you can't derive useful heritability from populations in different environments.
I once interviewed William Tucker, author of _The Science and Politics of Racial Research_ (which is even better than Gould's book), and he gave me a good illustrative example. Let's say we have a population of humans living on a South Seas island, and we measure the incidence of cancer. This population eats well, lots of fruit, good climate, so their incidence of cancer's going to be low. And because their environment is more or less the same from person to person (same geography, similar socioeconomic status, same climate, same labor), then most of the variance in the cancer rates is probably due to genetics.
If we measure the rates of cancer in northern New Jersey, you'll get a different estimate. The rate of cancer'll probably be a lot higher-- but that, in itself, may not mean that the heritability is higher. There will be more variation in individual environments (income level, nature of work, whether their housing tract's built over a toxic waste dump), which means that more of the variation in the cancer rates is due to environments. The _heritability_ may still be high, but less so than in the more homogenous environment of the South Seas island.
Now, if we take _both_ of these populations together, and work out the heritability for them combined, you're going to see a much _lower_ heritability score. For one thing, the rates of cancer now vary much more widely-- which will skew the estimates more in favor of environmental factors. One _may_ get some kind of information out of this, but frankly, it's just too weird of a data pool.
The point is that heritability is an estimate derived from very specific groups-- and while one _can_ estimate heritability between groups (it's math, after all), it's not going to be a meaningful estimate, and will tell us very little about genetic or environmental influences.