[tptacek]

I mean, start with the fact that it's a correlation statistic and not a causal statistic and just work your way back from that. Heritability --- of the kind with a research literature cite record --- is simply the ratio of phenotypical variation to genetic variation. The number of fingers on your hand is not, in that statistic, highly heritable.

I don't think there's a Wikipedia cite that's going to get you over that speed bump in your argument.

[Dylan16807]

A lot of those studies work very hard to split apart the genetic (and sometimes including epigenetic) factors at point of sperm meets egg, the environmental factors in the womb, the parenting factors, and/or the rest of the post-birth environment. They're not doing a simple ratio.

And to the extent that research successfully isolates the genetic factors, we know it's not some outside factor causing a correlation, and we know it's not IQ causing genes. Anything isolated there is genes causing IQ.

[tptacek]

IQ absolutely can cause genes! It's called assortative mating, and it's super potent and difficult to control for.

[Dylan16807]

Was it not clear that I was talking about the IQ and genes within each single person.

[tptacek]

I think you should take a beat and read up on assortative mating; also on heritability, which is a population statistic, not a measure of any single person. As it stands, this response you just wrote doesn't make sense.

[Dylan16807]

If you just measure genes and IQ, then assortative mating screws up that statistic very badly.

When researchers do studies specifically designed to isolate different factors, it's different. Please stop ignoring this part of my argument!!

Assume the most extreme case of assortative mating possible. Every child inherits two genes that list the exact IQ of its parents.

Do those genes correlate with the IQ of the child? If they do, there's only a few ways for that to happen. The causal factor could be how they're parented, or the environment, or the genes themselves. If you correct for the first two across a statistically large sample, and still see an effect, then it must be the last one.

> not a measure of any single person

I know. I'm saying each data point inside the statistic is arrived at in a specific way. Cause and effect only go via certain paths. It's easy to make mistakes about cause and effect by forgetting about paths, but you can categorize them and only some paths are possible. Any statistically resilient correlation has a cause somewhere.

[tptacek]

Molecular and behavioral geneticists absolutely, in modern studies, attempt to deconfound heritability statistics. Within-family and sibling regression are two of those techniques. When you use those study designs, heritability plummets for IQ, but not for traits like height.

I'm not trying to stake out a position about whether IQ is in any sense genetically causal or fixed. I'm saying that it's much more complicated than the Wikipedia page on "Heritability of IQ" would suggest. That's the only reason I dipped into this thread. You can believe whatever you want to believe, but this is an actively (indeed, furiously) studied open question, and the answer is definitely not "twin studies from 20 years ago set a heritability number that resolves the question".