On my recent post on the genetics of g – really the genetics of group differences (and especially racial group differences) in mean g – colnago80 raised in a comment some work on Panda’s Thumb summarizing certain research about intelligence, intelligence testing, g, and genetics. You should certainly read it if you have a mind to do so, and you can find it here. It was written recently, published yesterday, and intended to be a contribution to the current debates closely related to the discussion Murray and Sam Harris had on Harris’ podcast: do liberals irrationally reject a genetic contribution to g? For Panda’s Thumb, the current version of this discussion began with a post there 3 weeks ago that was based on research by PhD candidate Emily Willoughby.
The long and short of it is that neither the post linked by colnago80 nor the discussion-originating post on Willoughby’s research did anything to challenge the basic fact that IQ testing confounds are not distributed evenly across racial groups and that because of this, even though IQ tests do, in part, test something real that could be labeled g, the points identified as gaps between different racial groups average scores do not necessarily indicate a difference in average g. And even where they do, the differences in confound-distribution make it impossible to use these points of difference to make straightforward estimates of a gap in racial genetic potential for g.
I’m all for truth in science. I’m all for stating what we know candidly. But we must also state candidly the limits of what we know. We currently know that, for example, different children have access to vastly different educational opportunities. We currently know that, for example, access to regular meals before schools affects learning and access to meals specifically on test days affects standardized testing scores (and scores on class exams). We currently know that stereotype threat affects test scores, which in turn affect acceptance to universities, colleges, and other programs, access to scholarships, and other aspects of educational opportunity. We currently know that lead toxicity is damaging the brains of our children, leading to impulse control problems that indirectly affect learning as well as direct effects that injure intelligence.
Given the state of this knowledge, we can confidently state that our governments and other public institutions should be feeding children and actively working to eliminate racial, gender, illness/disability, and other prejudices. We should be constantly striving to improve schools. We should be actively cleaning up lead and other environmental toxins. And we should be taxing ourselves more if that is needed to accomplish these goals because the long term economic benefits clearly outweigh the current economic costs – and even if they didn’t, who wants to morally justify poisoning children because it was going to cost you five or twenty trips to Starbucks?
We do NOT know that there is a difference between the population genetics of racial groups that is responsible for any substantial part of the observed difference in IQ test results. There might be. If people want to research that, that’s fine. But because of the size of the confounds on such tests, because of the proportion of IQ testing results that are indirect measurements of factors other than g, because these other factors/confounds are distributed unequally across racial lines, and because environmental impediments to realizing an individual’s full genetic potential for g differ between racial (and other population) groups, at the current time we cannot know for sure that there is a g gap between any two racial groups or even which group would be found to be superior, much less can we know for sure a specific, quantized deficiency in IQ performance that is determined by racially-different genetic factors promoting g.
Given that we can’t know for certain that a gap exists (or even its sign) without dramatically more research on the exact nature and quantity on the different distribution of confounds, Murray and Harris are prematurely assuming we can make a socially or scientifically useful estimate of the genetic contribution to such a hypothetical gap in g.
Right now, we have a gap in IQ testing. What we don’t know for sure is whether we have a gap in g.
That ignorance is not addressed or ameliorated in any way by the two posts identified on Panda’s Thumb, nor by the current state of psychological literature. Murray has to make too many unwarranted assumptions in his model – in particular he has to assume that the accuracy of IQ testing in measuring both g and the genetic contributions to g are precisely equal across racial groups – to arrive at the answers he seems to prefer. These assumptions are in some cases directly in conflict with the evidence (if lead poisoning is different across racial groups then the ability to reach one’s genetic potential g is different across racial groups, which means that the racial contribution to g of genetics vs environment will be different across racial groups).
Murray has not addressed these criticisms. Harris’ interviewing seems not to understand or to (adequately) address these criticisms.
Given all this, Murray’s conclusions are neither socially nor scientifically useful. Given the history of racism in the US and the erroneous science, fallacious science, and pseudoscience that gave additional power to racism in the US, Murray’s conclusions and his discussions of his conclusions are not brave: by their nature they risk serious harm, and given the nature of large populations, have almost certainly done real harm in at least some lives.
Nothing written on Panda’s Thumb or elsewhere can change that.
Note: About 20 minutes after posting this article, I realized that the title needed changing. I added the word “Racial” to the beginning of the title. Other than that the title was not changed, but without that change, I realized that the title could imply that I believe that differences between persons could not be established. In fact, while there is a significant realm of uncertainty when scores on IQ tests are sufficiently close as to whether the gap is solely or partly on the basis of g, it is possible to have a sufficiently large gap between individuals that known confounds cannot explain it and a difference in g is, therefore, the only currently valid scientific explanation for (at least part of) that large gap.
Great post. Addresses the problems without a lot of name-calling. Congratulations!
One of the things which occurred to me after reading several blog posts and associated commentary on this subject is the question as to what is meant by inheritability. As I understand it, inheritability means that a child inherits 1/2 of it’s genetic makeup from mom and 1/2 from dad. However, suppose that mutations occurred after fertilization which affected one or more of the genes that determine g/IQ. If that is the case, then I would argue that the genes in question were not inherited from the parents but are, in fact, a result of environmental factors of one sort or another. As far as I can see, this issue is not addressed by any of the commentary and is, in particular, ignored by the genetic determinism proponents.
This is good critical thinking. I went with the example of lead poisoning because in addition to showing how a parent could have a beneficial gene and can give that gene to their offspring without their children actually benefiting from that gene, what we know about lead poisoning has practical implications for policy and collective action.
Your example of mutation is perhaps more on point – just because something is heritable doesn’t mean it is actually inherited – but with our current state of knowledge, there’s not much we can do (and it’s not even clear that there’s anything that we should do) about post-fertilization mutations.
Thanks for the contribution.
And that doesn’t even touch the question of whether “g” actually exists in a meaningful way. One of the first things you learn about intelligence is that it cannot be measured directly. I can measure height and weight and if my instruments are well calibrated and I do it well, this will lead to objective, reliable and valid results.
Everything else we measure in the area of cognition, education, IQ etc. only always measures behaviour and from that behaviour we draw conclusions. Depending on the subject and the result, we can draw very reasonable conclusions. Or not.
I just graded an English as a foreign language class test. One of our topics was stative and dynamic verbs, so we had an exercise, 10 sentences, where I gave the infinitive and they had to put it into the simple present or present progressive.
Some kids got 10 out of 10. It’s reasonable to say they understood the subject. Some kids got 6 out of 10. What does that tell me? Did they understand it in general but erred on the specific? Or didn’t they understand it at all and were lucky guessing?
Another kid didn’t use “be” in front of the present participle, which told me something completely different I hadn’t even intended to measure in that exercise.
Now I know that IQ tests are highly standardized and calibrated and they put a lot of effort into them, but given that the construct they’re measuring is indefinitely more complex and fuzzy than “knows how to use stative and dynamic verbs” it would be more than foolish to treat their results with anything but great caution.