June 29, 2017

Subtest Heritability, The G Factor, and Racial Intelligence Differences

The more heritable an IQ subtest is the bigger the difference between Blacks and Whites on that test tends to be. At least, that is what most of the evidence indicates.

Hu (2013-A) used sibling correlations as a proxy for heritability and found no relationship between a sub-test’s heritability and its racial gap size. However, Jensen (1973) used the same method and found a strong positive relationship between the heritability of a sub-test and its racial gap.

Hu (2013-B) and Hu (2013-C) used more direct measures of heritability and in both cases found a positive relationship between a sub-test’s heritability and the magnitude of racial differences on that sub-test. Nichols (1970) found a positive result as well. Chuck (2013) carried out two similar analyses one of which found a positive relationship, and one of which found a negative relationship.

1

Thus, 5 out of 7 studies found a positive relationship between sub-test heritability and the B/W gap size, 1 out of 7 found no relationship, and 1 out of 7 found a negative relationship.

What could explain this correlation? As will be seen, there are several possible explanations, but the best explanation is that the B/W IQ gap is significantly caused by genes.

Variation in Subtest Heritability

To being our exploration of this topic, it is important to remember that the heritability of a trait is the proportion of variance in that trait caused by genes. For the rest of this section, I am going to break variation down into “points” which are either caused by the environment or by genes. With that said, let’s look at the five possible causes of one subtest being more heritable than another.

2

One possibility is that test 1 may be more heritable than test 2 because differences in the relevant genes cause more absolute variation in test 1 than they do in test 2, and the amount of environmentally induced variation exhibited by each subtest is the same. For instance, suppose test 2 has 5 “variation points” caused by genetic differences between people and 5 variation points caused by environmental differences, while test 1 has 10 genetically caused variation points and 5 environmentally caused variation points. The heritability of test 1 will be 66% (10/15) while the heritability of test 2 will be 50% (5/10).

A second possibility is that test 1 is more heritable than test 2 because genes cause more total variation in test 1 than in test 2 while the environment causes more variation in test 2 than in test 1. Thus, if test 1 had 10 genetic variation points and 5 environmental variation point it would have a heritability of 66% (10/15), and if test 2 had 10 environmental variation points and 5 genetic variation points, it would have a heritability of 33% (5/15).

The third possibility is that both the environment and genes cause more variation in test 1 than in test 2. So long as the difference between the variation in the two tests caused by genes is greater than the difference in variation caused by the environment, test 1 will be more heritable than test 2. Thus, if test one has 10 environmental variation points and 15 genetic variation points and test 2 has 5 environmental variation points and 5 genetic variation points, test 2 will be 50% heritable (5/10) while test 1 will be 60% heritable (15/25).

Fourthly, it is possible that both the environment and genes cause less variation in test 1 than in test 2. So long as the difference between the variation in the two tests caused by the environment is greater than the difference in variation caused by genes, test 1 will be more heritable than test 2. Thus, if test 2 has 5 genetic variation points and 5 environmental variation points while test 1 has 4 genetic variation points and 1 environmental variation point then test 2 will be 50% heritable (5/10) and test 1 will be 80% heritable (4/5).

The fifth, and final,possibility is that the environment causes more variation in test 2 than in test 1, and genes cause the same amount of variation in each test. Thus, test 2 might have 5 genetic variation points and 10 environmental variation points, while test one has 5 genetic variation points and 5 environmental variation points. In this case, test 2 would be 33% heritable (5/15) while test 1 would be 50% heritable (5/10).

Competing Explanations

Which of these five scenarios is most likely to explain why more heritable subtests exhibit larger racial gaps? I am going to argue that the second explanation is the most likely, followed by the first, and that both explanations suggest that the Black-White IQ gap is significantly caused by genes.

The fourth and fifth explanation can be thrown out immediately because they both imply that more heritable subtests will exhibit less total variation. This is unlikely since the difference between Blacks and Whites is larger on these tests, and this implies that more heritable tests have more variance rather than less.

The first explanation posits that genes cause more variation on more heritable subtests while the environment causes the same amount of variation on each subtest. Intuitively, the first explanation would seem to be the most likely. There is no obvious theoretical reason to expect that the amount of variation in a trait caused by the environment would correlate positively, or negatively, with the amount of variation in that trait caused by genes. The default assumption would be that these two variables are independent, and thus the first explanation is the default assumption.

By contrast, the third explanation posits that there is a positive correlation between environmental and genetically caused variation while the second explanation posits a negative correlation between these variables. Thus, whether more heritable subtests have more or less than average environmentally caused variation is central to which explanation is most likely to be correct.

To explore this question, we need to first review some background about IQ testing.

General Intelligence

Meta-analyses of hundreds of studies show that a person’s scores across all IQ subtests correlate with one another (Deary, Spinath, and Bates, 2006). In other words, people who do well in one kind of test, say basic math, tend to also do better in other kinds of tests, such as vocabulary.

Most intelligence researchers think that the cause of these correlations is that most intelligence related cognitive traits involve an overlapping skill or set of skills called “general intelligence” or “The G Factor”

Subtests differ with respect to how well they, individually, measure general intelligence. How well a test measures general intelligence, also called a test’s G loading, is measured by how well it predicts scores on other sub-tests. The better one test predicts other tests, the more G loaded it is said to be.

The G Factor and Malleability

It turns out that the more G loaded a test is the less easy it is to change by manipulating the environment. This is shown by several lines of evidence.

First, Nijenhuis, Grimen, and Kirkegaard (2014) looked at how the G loading of a sub-test correlated with gains produced by Head Start.  Head Start is a program conducted by the American Government which attempts to put high risk kids (poor and low IQ children) into enriching environments early in life in order to increase their chances of doing well. The program provides education, nutrition, and positive social settings, that these children wouldn’t normally have access to. One goal of Head Start is to increase the intelligence of these children. Head Start is successful at this goal in the short term. The intelligence of children normally increases once they enter head start. However, these gains typically disappear before adulthood. Anyway, the Head Start kids gain more on some IQ tests than they do on others and the more Head Start produces gains on a sub-test the less G loaded that test tends to be.

Second, Nijenhuis and Flier (2013) carried out a meta-analysis on how a test’s g loading related to its role in the Flynn Effect. The Flynn effect refers to the fact that, over the 20th century, many nations around the world saw IQ scores rise by about 3 points per decade. Given that large changes to the human gene pool were not occurring every decade, this effect must have been environmental in origin. The paper found that the G loading of a sub-test negatively correlated with how strongly it was impacted by the Flynn effect.

Thirdly, Nijenhuis, Grimen, and Armstrong (2015) looked at how the G loading of a test related to how well scores on the test were improved by adoption. When children are taken from bad homes adopted into good homes they usually score better on intelligence tests than they would have if they were left with their biological parents. But their scores don’t show equal gains on all tests. Because of this, researchers have been able to show that the more a child’s score on an IQ test can be improved by adoption the less G loaded that test tends to be.

The G Factor and Subtest Heritability

IQ subtests also differ in terms of how heritable they are, and this tends to follow their pattern of G-loadings. Nijenhuis, Kura, and Hur (2014) meta-analysed data from 6 data sets on sub-test G loading and heritability among the Japanese and found a correlate of .38 between G loading and heritability. Similarly, Kan (2011) reported a correlation of .52 between G loading and heritability in western samples.

Thus, the more G loaded a test is the less malleable it is and the more heritable it is.

The G factor and Subtest Racial Differences

A subtest’s G loading also correlates with the size of its racial gap . Rushton and Jensen (2010) summarized this research thusly:

“With respect to the g factor, Jensen [22, pp. 369-379] summarized 17 independent data sets of nearly 45,000 Blacks and 245,000 Whites derived from 149 psychometric tests and found the g loadings of the subtests consistently predicted the magnitude of the Black-White differences (r = .62, P < .05). This was true even among 3-year-olds administered 8 subtests of the Stanford-Binet; the rank-order correlation between the g loadings and the Black-White differences being .71 (P < .05) [36]. In Hawaii, IQ differences between East Asians and Whites (favoring East Asians) were greater on the more g loaded of 15 subtests among people of Japanese, Chinese, and European ancestry [37]. In Zimbabwe, Rushton & Jensen [38] found 77% of the difference between Africans and Whites was due to g in a principal factor reanalysis of WISC-R data from 12- to 14-year-olds originally published by Zindi [39]. In South Africa, Rushton et al. [40, 41] found the differences between Black, South Asian, and White engineering students were greater on the more g loaded items from the Progressive Matrices.”

Conclusion

Taken together, this evidence suggests that the more heritable a subtest is the less easy it is to change, the more G loaded it is, and the more the races will differ on it.

Recall that before I had argued that the correlation between a subtest’s heritability and its racial gap size could be explained in one of three ways. All three explanations stipulated that the more heritable a subtest is the more total variation caused by genes there is going to be. Where they differed was in whether the amount of total environmental variation exhibited by more heritable subtests would tend to be greater than, less than, or equal to, average.

If the amount of environmentally caused variation in by a subtest test has no relationship, or a negative relationship, with the amount of genetically caused variation in a subtest, then the racial gap in IQ is probably partly genetic in origin. Why?

Because the only way that larger total variation caused by genes can cause larger variation between the races is if said genes cause some of the variation between the races. The only way that this data can be reconciled with the egalitarian viewpoint is if more heritable subtests also have exhibit more environmentally caused variation.

For example, if more heritable sub-tests tended to exhibit more variation caused by genes, but also more variation caused by wealth, then a positive relationship between subtest heritability and subtest racial gap size could be due to the gap being caused by genes or by wealth.

However, the negative relationship between a subtests G loading and its malleability, and the positive relationship between a subtests G loading and its heritability, as well as its racial gap size, suggests that the opposite is true. That is, more heritable subtests probably have less than average environmentally caused variation.

If you find this line of reasoning uncompelling, then you are left with the default assumption, explanation #1, which also implies a genetically caused racial IQ gap.

There is virtually no reason to think that more heritable subtests have more environmentally caused variation. And thus, the best explanation of all the data reviewed here is the Black/White IQ gap is partly caused by genes.

Facebook Comments