As you may have noticed, it is not popular to suggest that genetics is a possible cause of individual differences, and distinctly unpopular to even hint that it might be a cause of genetic group differences.

By way of background, when Arthur Jensen was first considering the causes of the black-white difference in scholastic attainment, he gave an entirely environment-based explanation. This was in a lecture in 1967, when he was first looking at the issue. You could not put a cigarette paper (remember those?) between him and any sociologist. It was only when he got far deeper into the literature that he changed his mind. Facts mattered to him, and should matter to everybody. He made an evidence-based alteration in his prior views, and on page 82 of his famous 1969 paper said:

The fact that a reasonable hypothesis has not been rigorously proved does not mean that it should be summarily dismissed. It only means that we need more appropriate research for putting it to the test. I believe such definitive research is entirely possible but has not yet been done. So all we are left with are various lines of evidence, no one of which is definitive alone, but which, viewed all to¬gether, make it a not unreasonable hypothesis that genetic factors are strongly implicated in the average Negro-white intelligence difference. The preponderance of the evidence is, in my opinion, less consistent with a strictly environmental hypothesis than with a genetic hypothesis, which, of course, does not exclude the influence of environment or its interaction with genetic factors.

A cautious statement which was met not with the proposed further research, but with outrage and the circling of defensive wagons which continues to this day. Indeed, any researcher who penned such a paragraph in a paper submitted to a main stream journal today would probably find that it was sent not to the usual two, but to three or four reviewers, the last two expressing severe reservations as to whether it should be accepted.

You know what happened over the last five decades: researchers cautioned, warned, excoriated, censured, censored, unfunded and more recently sacked for considering the hypothesis on an empirical basis.

So, what happens next? Well, if facts exist, someone will notice them and, if brave, find a way of letting people know what they have noticed.

A Multimodal MRI-based Predictor of Intelligence and Its Relation to Race/Ethnicity. Emil O. W. Kirkegaard and John G.R. Fuerst. The Mankind Quarterly · March 2023.
DOI: 10.46469/mq.2023.63.3.2
https://drive.google.com/file/d/1iVjNT1Uv9fC3uZzwBPpVQDRGWbmiPXoV/view?usp=sharing

What have they found?

They say:

We used data from the Adolescent Brain Cognitive Development Study to create a multimodal MRI-based predictor of intelligence. We applied the elastic net algorithm to over 50,000 neurological variables. We find that race can confound models when a multiracial training sample is used, because models learn to predict race and use race to predict intelligence.
When the model is trained on non-Hispanic Whites only, the MRI-based predictor has an out-of-sample model accuracy of r = .51, which is 3 to 4 times greater than the validity of whole brain volume in this dataset. This validity generalized across the major socially-defined racial/ethnic groupings (White, Black, and Hispanic). There are race gaps on the predicted scores, even though the model is trained on White subjects only.

This predictor explains about 37% of the relation between both the Black and Hispanic classification and intelligence.

So, by looking at all the MRI measures they can predict IQ better than by using brain size alone. So, we should stop using the brain size measure (about 0.28) and move to this better measure (0.51) and eventually the even better ones that may be found later.

First of all, some summary data:

These are adolescents, so things may change a bit with age, but these are good sample sizes. Black adolescents have a somewhat lower than expected low score, and a high standard deviation, the latter surprisingly so, since many previous black samples have a standard deviation of 13. I don’t know how to interpret this, but it might be due to the subtests used.

The intelligence tests used were not the best sample of skills (where were Maths, or Wechsler Vocabulary or Block Design?) and they over-represented working memory tests, which I think are weak measures, though fashionable. It may account for the large standard deviation finding. I predict that a more representative range of tests would lead to even higher predictive accuracy overall, and perhaps lower standard deviations.

The learning algorithm they employed was one suitable for use in the tricky setting where there are far more variables than individual subjects. When you apply the algorithm to the whole sample, leaving aside race, then the correlation with IQ is 0.60 which is very high. Using this technique, a brain image gives a good guide to the power of the brain as a problem-solving organ.

Using an MRI-based predictive equation the authors did a better job of predicting a person’s IQ than was possible from knowing their parent-described race, despite the racial differences in intelligence being large. These “social race” labels were redundant for the purposes of predicting intelligence.
The correlations of MRI prediction with actual intelligence test results within each social race were pretty similar: white 0.51, black 0.53, Hispanic 0.54 and other 0.58.

They tried to see if their algorithm could use MRI data to predict the social race of the adolescents, and found they could do so with 73% accuracy. The distinction between blacks and whites could be drawn with almost complete accuracy, only a 2% error rate either way.

They then trained a model to predict genetic ancestry. You might want to call this race.

They say:

We find that MRI-based predictions of genetic ancestry are very accurate. The three correlations of interest are .91, .89, and .61, for European, African, and Amerindian ancestry, respectively. As expected, the correlations were higher for European and African ancestry than for White and Black social race, respectively.

So, you can study a brain image and predict the person’s race with very high accuracy.
A possible counter-argument is that the model has learned to spot race, and then sharpens its intelligence predictors. The authors decided to test a predictive equation for whites only, so as to cut out this possible confounder. In fact, it only goes down from 0.51 on the full dataset to 0.48 for whites only.

As usual, I have left out some of the additional tests carried out to make sure the findings were robust.

the model learned to predict subjects’ social race/genetic ancestry based on the MRI data, and then used this information to predict intelligence. This finding is consistent with those of Gichoya et al. (2022), who report that machine learning can recognize self-reported race/ethnicity from a wide variety of medical imaging data. This is unsurprising because in the USA, socially identified race closely tracks continental genetic ancestry (Kirkegaard et al., 2021; Tang et al., 2005), which certainly is not “only skin deep”.

We further analyzed which aspects of MRI data were most useful in the prediction of intelligence. We found that functional MRI (fMRI) task data, which measures blood flow while performing tasks, had the highest validity for predicting intelligence. Additionally, MRI datasets which had more variables, which showed larger race differences, and which had higher correlations with polygenic scores, had higher validities.

So, what can we conclude here? Using a statistical technique to study the many scores produced by magnetic resonance images, it is possible to predict the subject’s intelligence and their race with high accuracy. The predictive equations work for all races, so they have not been damaged by some presumed test bias. Some of the MRI measures are more predictive than others, (these tasks are not used to calculate IQ scores), which are measuring something about brain flow in the brain as tasks are carried out.

All this is brain deep, not skin deep.

Is there anywhere left for blank slate-ists to hide? They had trouble accepting that intelligence was heritable within genetic groups. After many decades of research some (but by no means all) made a grudging admission that heredity accounted for something. Then they argued that heredity was weaker in lower socio-economic-status groups. That was eventually shown to be unlikely, though the debate has been long and hard, so not all researchers accept that finding. Throughout, blank slate-ists denied that genetics could explain genetic group differences. They argued that each race showed the effects of heredity within race, but none between races (or not more than a bare 5%). They claimed that observed differences must (95% of them) be due to environmental differences, including those which are hard to detect, but have powerful effects. Will this paper change their minds? I think not. Not for 50 years, anyway.

This paper makes a simple point: brains differ between races, and these differences relate to differences in intelligence.