The most race-obsessed people, paradoxically, are those who claim race does not exist. Academics in particular are prone to insist on race being a fiction invented by “racists”, people who can’t do proper statistics. These academics become DIE-maniacs, bending all honors toward those races, that do not exist, which they love most, blaming the white race, which does not exist, for all that ails the world.
Let us think about some statistics of race. In that vein, you might have heard of “Lewontin’s Fallacy,” named for an argument made by Richard Lewontin in his attempts to show there was no such thing as race. That one is well known (and if you don’t know, you can see the link). Today I want to focus on another of Lewontin’s arguments, here about race and “IQ” (by which everybody means intelligence).
We won’t speak of race nor intelligence until the end. Instead, we first want to understand how causes of differences in any biological measure can be known. This will ask a lot of you, because it is not simple. But since race is so contentious, it will pay to be as careful as we can, so that in the end we are precise in any claim.
Lewontin’s Race and Intelligence
Here is Lewontin with an influential example from his paper “Race and intelligence“. Here is the start:
Let us take two completely inbred lines of com. Because they are completely inbred by self-fertilization, there is no genetic variation in either line, but the two lines will be genetically different from each other. Let us now plant seeds of these two inbred lines in flower pots with ordinary potting soil, one seed of each line to a pot. After they have germinated and grown for a few weeks we will measure the height of each plant. We will discover variation in height from plant to plant. Because each line is completely inbred, the variation in height within lines must be entirely environmental, a result of variation in potting conditions from pot to pot. Then the heritability of plant height in both lines is 0.0.
By “completely inbred” he means each line is genetically identical. One line has (say) gene set A and the other line gene set B. After planting and measuring heights, he then says something which sounds strange, but which is common in genetics. He says “the heritability of plant height in both lines is 0.0.”
Now, in plain English, the kind we commoners use, this is weird because we might take the “heritability” to be 100%.
If we held with the machine metaphor of life, and say all organisms are naught but gene machines, we’d say each plant’s height (in each line) has inherited its heightness, or height capabilities, and inherited it exactly. We’d also believe, with Lewontin, that any differences in height within each line (A or B) was because of differences in the pots, or whatever happened to those pots (differing amounts of sun, say, or evaporation). If, somehow, we could control the environments of the pots exactly (including through time), then we’d believe all heights in each line would come out identical. Which seems to be a perfect kind of heritability.
But this is not the heritability of genetics, which is instead a bit of statistical jargon.
Heritability is the amount or proportion of variance in a measurable trait, like plant heights in a line, that is believed to be caused by genetic and not environmental differences. Since here, accepting the machine metaphor, there is no variability in plant heights we assume was caused (“attributed” is the usual euphemism) by genes, there is no heritability.
Heritability is an on-average concept. Like a lot of statistical buzzwords it, speaks of variation and correlation. And like all correlations, it can be tricky to interpret, because correlation isn’t known causation.
Add to that that the machine metaphor of life is false (see this and this). Take what is called “epigenetics”. It could be, though identical genetically, those plants might express their genes differently because of differing environments (conditions) or even through causes inside individual plants themselves (and not the pots per se). Suppose one or more, but not all, pots in a genetic line had some event, perhaps internal to the plant, which caused different genes to be “activated”, and these in turn caused differences in heights. Were the differences in observed heights in that line caused by genes or conditions?
Well, both. Because, do not forget, we take the Aristotelian definition of cause to be the full explanation for a thing, the complete why, when, how, etc. The efficient cause is only part of the explanation. Heritability, the statistical concept, is thus not fully causal.
If all we measured was heights and assumed identical genetics in each line, heritability, even in the face of epigenetics, would still be 0 in this case even though some of the differences in heights was caused by genes. Or genes-environment, rather. Once again, I must warn us about the dangers of one-number summaries of complex behaviors.
All right. That much (alas) is just throat clearing.
Let us suppose the machine metaphor is right, and that the only differences within a line are caused by differing conditions. Each line will have a distribution of plant heights. I don’t mean some theoretical probability distribution. I mean an actual set of different height measurements, one for A and one for B. It is logically possible the distribution of height will be identical, in frequency of heights, between the two lines. However, this is almost never observed. Instead, what is observed are differences in those distributions, both in the entirety of the frequency distribution, and in functions of it, like averages (means).
Functions of distributions are more likely to be the same or similar than the distributions themselves, because functions like means or variances compress information. Two distributions that vary markedly can have the same or similar means, or even the same variances, or even both. As always, we must always be wary compressed-information summaries.
All right, suppose line A and line B had different distributions, and even different means. The “heritability” of both lines, assuming the machine metaphor, is 0. What caused the differences between the lines?
We don’t know. Not yet. It could be the differing conditions in the pots, or it could be the difference in genes between A and B, or it could be both, and even more so if we don’t accept the machine metaphor. Just looking at the statistics and observing the differences cannot tell cause. It could even be measurement error (the device used to measure may have problems), which we’ll here ignore.
As always, it’s we who bring cause to data. I don’t mean in only genetics: everywhere.
Enter control. We suspect the difference in distribution is caused mostly by the gene-environment makeup of the plants. But we know that because there are differences in height in the same line, assuming things like epigenetics are minor here, that difference must be caused by differing conditions. So, if we can, we redo the experiment, this time controlling all the conditions in the pot to make them as similar as possible. Maybe even putting seeds from each line into the same pot. Conditions can still vary inside pots, and these could be causal differences. But we hope the better we control, the more likely the differences we see are caused by the plants themselves.
Even if we’ve controlled the conditions as perfectly as we can, there could still be things we haven’t thought of, like that measurement error. But if we assume those things are minor or nonexistent, in the end we have the distributions of heights for both lines, and any differences we put down to the plants themselves. We don’t need probability, we don’t need “hypothesis testing”, we just need those observations, and a glance tells us if there are differences in distributions. We only need probability and models if we want to make predictions of differences we didn’t measure; we could even use them to account for differing conditions.
The Next Step
Go back and re-read Lewontin’s example set up above. After that, read these, his very next words:
But there will be an average difference in plant height between lines that arises entirely from the fact that the two lines are genetically different. Thus the difference between lines is entirely genetical even though the heritability of height is 0!
The first sentence we now understand. That second sentence only makes sense knowing the jargon of heritability. Another way to put it, in plain English, is that the difference is entirely due to inheritance in each line. Which might have sounded opposite until we understood the jargon.
We further confirm that if we want to think about causes of differences in those plant heights, heritability is the wrong term. Because we know (up to the caveats mentioned above) the cause is ineradicable innate genetic difference.
Lewontin continues (with my paragraphifications):
Now let us do the opposite experiment. We will take two handsful from a sack containing seed of an open-pollinated variety of com. Such a variety has lots of genetic variation in it. Instead of using potting soil, however, we will grow the seed in vermiculite watered with a carefully made up nutrient, Knop’s solution, used by plant physiologists for controlled growth experiments.
One batch of seed will be grown on complete Knop’s solution, but the other will have the concentration of nitrates cut in half and, in addition, we will leave out the minute trace of zinc salt that is part of the necessary trace elements (30 parts per billion).
After several weeks we will measure the plants. Now we will find variation within seed lots which is entirely genetical since no environmental variation within lots was allowed. Thus heritability will be 1.0.
However, there will be a radical difference between seed lots which is ascribable entirely to the difference in nutrient levels. Thus, we have a case where heritability within populations is complete, yet the difference between populations is entirely environmental!
Let me rephrase. The seeds, handful A and handful B, are both filled with genetically different specimens. Handful A is grown in an ideal condition. Handful B is grown in a poor condition. Inside A, the variation in plant heights is ascribed to genetic variation on the assumption the conditions for each seed are identical, and ignoring any epigenetics etc. Same inside B.
The (jargon word) heritability is now 1 (some say 100%) inside each handful. Even though, in plain English, we’d say the differences inside each handful were caused by different genetics.
There is also, presumably, a difference in distributions (actual, not theoretical) between the handfuls. How much of that difference is caused by the difference in growing conditions and how much was caused by the difference in genetics of the seeds (again ignoring epigenetics)?
As above, we don’t know. It could be that the handfuls have very different distributions in genetics: after all, we did not measure anything about the seeds. We can assume that the distribution (the actual, not some theoretical model) of genetics differences were the same in each handful; if we do, then we can ascribe the difference between handful distributions to the differing environments. But this is only an assumption.
A good one, given experience, since in order for there to be a difference in the handfuls, something must have happened to cause us to grab different distributions of genetics. If the seeds are all mixed up in a bag, this seems difficult. Not impossible, though. The differences in genetics might cause differences in seed size, say. If you’ve shaken a jar of peanuts, you’ll have noticed the small bits cluster on the bottom, with the large ones shifting to the top. Same kind of thing can happen here. Our handful assumption much be checked.
Bad Chemistry
Lewontin continues:
But let us carry our experiment to the end. Suppose we do not know about the difference in the nutrient solutions because it was really the carelessness of our assistant that was involved. We call in a friend who is a very careful chemist and ask him to look into the matter for us. He analyzes the nutrient solutions and discovers the obvious–only half as much nitrates in the case of the stunted plants. So we add the missing nitrates and do the experiment again. This time our second batch of plants will grow a little larger but not much, and we will conclude that the difference between the lots is genetic since equalizing the large difference in nitrate level had so little effect. But, of course, we would be wrong for it is the missing trace of zinc that is the real culprit.
First, we note that the “very careful chemist” wasn’t that careful, because he missed the zinc, that “necessary”, Lewontin said, trace element. Well, we all make mistakes.
It is true that if we assume there is no difference in conditions, but we then observed differences in distributions in heights, then we can assume the difference is due do genetics of the handfuls. Which brings up the question how those differences arose, since we just grabbed handfuls.
Was our assumption correct that the genetics of the handfuls was the same in distribution? Or, like with peanuts, was it wrong? Should we check? Or just let it go and keep our sameness assumption? After all, Lewontin goes on to assume we ought to keep checking conditions. So it seems only fair to also check the genetics.
We could hire other chemists, also very good, and they, too, could miss the tiny hidden difficult-to-spot singular cause all the chemists before missed. They could, even, all miss a causal condition that nobody even knew existed. As I said above, we infer cause based on our assumptions.
But we could also measure the genetics, because, hey, why not? If we did, we might discover they were the same (in distribution) or different. If they were the same, then we would assume we missed some environmental condition, even if we could not identify what it was. We might then re-run the experiment as we did above, say putting handful A seeds and handful B seeds in the same pots and so forth. Then test the conditions of the pots that gave the smallest and largest heights and so on, see if we could nail what that condition or conditions were.
If the measured genetics were different, then we might jump to the conclusion (assumption) that the height differences were caused by genetics. But since it would also be possible that differences in environment might also be causal (as above), we again might rerun the experiment, this time being careful to put handful A and B in the same conditions, as best we could.
Then, after this repetition (or repetitions), we still see differences, we would say the genetics were causing the differences, as before not paying too much attention to epigenetics or assuming its causal influence was negligible.
And this is exactly how science works. Not so easy, with many perils. Yet not impossible, either, which we know because scientists got real good at genetics, especially in plants, where experiments are (relatively) cheap and easy.
Live Wire
In Lewontin’s examples, the heritability concept was not really helpful, and was even a tad misleading. It certainly did nothing for us in understanding what caused observed differences. Indeed, it would be best to eschew it altogether, when you can, and just speak about cause plainly, or probabilistically when causes are known imperfectly. Far less confusing for everybody.
Here’s where it gets interesting. Lewontin continues his example, but switches from seeds to IQ, reifying it with intelligence, as everybody does (The Deadly Sin of Refication is the most common failing in science). You cannot inherit IQ (a score on a test): you can only inherit intelligence. Let that pass.
He argues, in effect, that there must be hidden or uncontrolled environmental conditions which cause observed differences in intelligence between races. He says:
The article under discussion began with the observation, which he [Arthur R Jensen, coiner of the original fallacy name] documents, that compensatory education for the disadvantaged (blacks, chiefly) has failed. The explanation offered for the failure is that I.Q. has a high heritability and that therefore the difference between the races is also mostly genetical. Given that the racial difference is genetical, then environmental change and educational effort cannot make much difference and cannot close the gap very much between blacks and whites. I have already argued that there is no evidence one way or the other about the genetics of inter-racial I.Q. differences.
He wrote this in 1970 when the genetics evidence was really only just starting. It is now richer and robust. There are now many known differences between races. Of course, there is much left to learn.
But it is no longer a question whether the genetics in our handfuls are different: they are. Yet Lewontin’s followers are still searching for that missing zinc. They won’t likely find it, especially considered the experiments have been repeated time and again, in many varying conditions, even those in which lavish attention and money are spent on favored races.
Of course, this isn’t conclusive proof. There is always the chance that the zinc is there and as yet undiscovered, and that the differences in genetics turns out to be non-causal for intelligence. Again, this is less and less likely given the widely varying “experimental” conditions.
Now all of this is different than saying which genes under what conditions are important for intelligence. We can no longer ignore those epigenetics, and must abandon the machine metaphor. If we don’t, then all we’re left with are bunches of correlations with this and that allele, which can shift and be vague (or, Lord help us, principal components, which are vaguer linear combinations of measures). This is because intelligence itself is difficult to measure, environments can be hard to quantify, and because we know except in rare cases there are not dedicated genes “for” things like intelligence (see this and this). This, of course, does not mean that we cannot know; only that we have to eschew crude statistical practices (see the Class).
Lewontin’s other fallacy is thus to suppose we haven’t measured those genetic differences, that we’re operating blind about those handfuls of “seeds”, which he insists without evidence are equal in distribution. Of course we have, and of course they are not. Even if we cannot know with absolute certainty which genes are causing which differences in measurements like IQ, because we have seen so many varied conditions, especially in DIE-“affirmative action”-regimes, and because we know with certainty there are genetic differences, we cannot dismiss, as many would, that innate differences are causal.
This is an unhappy result for those who insist on Equality, i.e. those who consider the “pots” to be the strongest causes, and that the known differences in distributions of genes between races are somehow not causal. They allow the possibility that different genetics in individuals can be causal, but paradoxically dismiss the idea that those differences cannot be causal when the differences are different between races.
None of this would be that important, except for one or two things of interest. As I’ve said many times, our culture assigns moral value to intelligence, but only for man. The other is that we, all of us, cannot help believing we are at the high end of the distribution of the group to which we claim membership, or that we are somehow responsible for the top, while simultaneously believing the guy in the group to which we don’t belong is at the bottom. We are all above average, and the other guy is always below. And we do this even after we have received evidence of any individual’s intelligence. And then, of course, there is Equality, which tries to fix this with its own disastrous results.
Video
All my work is free and user-supported. Here’s how to help:
- Subscribe at Substack (paid or free!)
- Cash App: $WilliamMBriggs
- Zelle: use email: matt@wmbriggs.com
- Buy me a coffee
- Paypal
- Other credit card subscription or single donations
- Hire me
- Subscribe at YouTube
- PASS POSTS ON TO OTHERS
Discover more from William M. Briggs
Subscribe to get the latest posts sent to your email.
