Race differences in IQ: a response to Colman 2016

Ion
11 min readMar 21, 2020

--

Introduction

Andrew M. Colman wrote a review of research concerning race differences in IQ called “Race differences in IQ: Hans Eysenck’s contribution to the debate in the
light of subsequent research”

As far as I know, no direct responses have been done to this literature review. It’s been cited only 12 times, half of which are by Kirkegaard et al. In sum, it’s hugely biased and simply wrong about the current (or what was current in 2016) research on race and IQ.

Scientific consensus

Coleman says that “Eysenck was exaggerating about the consensus of expert opinion”. Eysenck’s statement was mostly false in 1971, but became mostly true in the 1980s to today. It’s worth noting that Coleman cites Pettigrew 1965, a literature review to prove this, while there is much better evidence against Eysensck’s statement in Friedrichs 1973 which directly asked APA members what they thought about Jensen’s “unreasonable hypothesis”, and found that most members disagreed with it.

Snyderman and Rothman’s 1987 survey of expert opinion found that most experts agreed that the Black-White IQ gap is due to both genetic and environmental factors.

A priori assumptions

Coleman argues that racial differences in IQ related genes are a priori unlikely (at least he admits Gould-like arguments of an a priori impossibility are wrong).

He justifies this with the fact that humans are genetically relatively similar, and so are human populations. It’s unclear by what standard racial differences in IQ related genes are unlikely. If humans share 99.7% of their DNA with chimpanzees, and/or most of the variation is within groups (which depending on the measure it is), does that mean differences in IQ related genes between humans and chimpanzees are unlikely? Of course not.

This argument was also put forward by Lewontin and Ned Block. Sesardic 2005 p148–151 deals with it:

I think any argument of that form is logically flawed. For even if races
differ genetically only to a small extent on average (taking into account
a selected number of protein loci or genetic markers), this in itself is
not a good reason to think that the races will also differ genetically only
to a small extent with respect to a specific phenotypic trait. Block’s attempt
to use the low percentage of the between-races genetic variation (“only
7 percent) in his anti-hereditarian argument about IQ is fundamentally
misconceived. If, averaged over many genetic loci, the racial variation
constitutes only 7 percent of the total variation, it by no means follows that the proportion of racial variation in IQ genes will probably be
7 percent, or around 7 percent. It may well be, of course, but this cannot
be established a priori.

Notice that no evidence whatsoever is offered to support the first part of
(B), namely that practically the only genetic differences between groups
are in skin color and body form. This is in fact the very issue that Lewontin
was supposed to resolve by argument! The whole section is phrased as if
(B) is argumentatively linked to (A), but this is obviously not the case.
Moreover, although the main thrust of (A) is to prepare the ground for
a refutation of the idea that there are genetic differences responsible for
between-group differences in mental abilities, in the end Lewontin says
nothing that would speak to that issue.
Since Lewontin mentions differences in skin color, a good question
here is: do we expect that the component of inter-racial genetic variation with respect to that trait will be also around 7 percent? Certainly
not. Actually, according to a recent study (Relethford 2002) it is 88 percent. Now the issue we are addressing is the following: is the distribution
of genetic variance with respect to cognitive abilities more like (1) the
case of skin color, where between-race variation is comparatively high,
or like (2) genetic loci examined by Lewontin and others (Lewontin 1972;
Barbujani et al. 1997), where the average between-group component is
comparatively low (less than 12 percent), or perhaps (3) somewhere in
between?19 The honest answer is that we just don’t know. This is an empirical question, and drawing inference about cognitive abilities on the basis
of what we know about, say, blood groups is completely unjustified.

Furthermore, the amount of variance between groups was underestimated by Lewontin:

Yet the world had to wait until 2002 for someone to explain the basic problems with Lewontin’s famous 15 percent. It was Henry Harpending replying to a question from Frank Salter. Lewontin had noted that 85 percent of the genetic variability was among individuals within populations, and only an additional 15 percent was added when individuals in different populations were compared. However, this analysis omits a third level of variability — the within-individual one. The point is that we are diploid organisms, getting one set of chromosomes from one parent and a second from the other. To the extent that your mother and father are not especially closely related, then, those two sets of chromosomes will come close to being a random sample of the chromosomes in your population. And the sets present in some randomly chosen member of yours will also be about as different from your two sets as they are from one another. So how much of the variability will be distributed where?

First is the 15 percent that is interpopulational. The other 85 percent will then split half and half (42.5 percent) between the intra- and interindividual within-population comparisons. The increase in variability in between-population comparisons is thus 15 percent against the 42.5 percent that is between-individual within-population. Thus, 15/42.5 = 32.5 percent [This should be 35%], a much more impressive and, more important, more legitimate value than 15 percent.

Coleman makes the exact same fallacious argument 50 years later. There is no evidence that “race is only skin deep”. There is however evidence against it:

Coleman then says something extremely weird:

recent research has revealed that racial admixture has blurred whatever
genetic differences might have existed previously

It’s true that African Americans are, on average around 1/5 white. How this is an argument against Hereditarianism is unsupported, when Hereditarians are obviously aware of it. Under the Hereditarian hypothesis, the white admixture in African Americans would give them a boost of 5 IQ points compared to their African brethren, yet a large gap would still remain.

Heritability

Coleman states that the .8 heritability expressed by Jensen and Eysenck was too high:

One problem with the argument from heritability is that the estimate of
h2 (broad) = .80, on which both Jensen (1969) and Eysenck (1971) based
their argument, is now known to be too high

He hints at quantitative genetic methods (twin studies) being contradicted and outdated by molecular genetic methods (GWAS) yet quickly brushes this away by mentioning heritability estimates now range between .35 and .70. If molecular genetic methods are supposedly superior to quantitative genetic methods (at the moment, they’re not), why is he still using quantitative genetics? Why even mention molecular genetic methods in a section about heritability being too high if you then still trust molecular genetic methods? All these questions make me wonder what went through his head while writing this.

Coleman then mentions the Wilson effect:

depending on various features of the sample being investigated; for
example, it tends to be larger in older populations and smaller among children,

It is true that heritability consistently changes with age, reaching an asymptote at age 24. It’s unclear how this is related to heritability of .8 being too high. If we are to relate heritability to group differences, which we will touch on later, why is it relevant that heritability is lower at age 5? We can simply account for age and only compare adults. Furthermore, Jensen was fully aware of the Wilson effect, and wrote about the causes of it.

The Scarr-Rowe effect is also brought up:

it is close to zero at the lowest socioeconomic status levels
(Mackintosh, 2011, chap. 11; Nisbett et al., 2012)

What confuses me yet again is why the Scarr-Rowe meta-analysis is not mentioned, which only found a moderate effect in the US, and not anywhere else. We now know since then that Scarr-Rowe effects are insignificant, even in the US, and that there are no significant racial differences in heritability. I don’t blame Coleman for believing Scarr-Rowe effects exist. It was a somewhat reasonable belief at the time. I do blame him for exaggerating Scarr Rowe effects, not mentioning how Scarr-Rowe research is plagued with non-replications and not mentioning how significant racial differences in heritability have never been found.

Unsurprisingly, Colman repeats the 50 year old strawman made by Lewontin.

The most serious problem with the argument from heritability is
that it uses data about the heritability of IQ within populations to
draw conclusions about the genetic basis of differences between populations. The argument is now known to be invalid, as Lewontin (1970)
demonstrated

Lewontin’s argument is wrong both in theory and in practice.

First, Jensen never made the argument that high within group heritability (WGH) implies between group heritability (BGH), and it’s unsurprising that Colman does not quote Jensen. Jensen has this to state about Lewontin 1970:

The main thrust of Lewontin’s argument, as he sees it, actually attacks only
a straw man set up by himself: the notion that heritability of a trait within
a population does not prove that genetic factors are involved in the mean
difference between two different populations on the same trait. I agree. But
nowhere in my Harvard Educational Review discussion of race differences
do I propose this line of reasoning, nor have I done so in any other writings.
(Jensen 1976b: 103)

Jensen instead argued that high WGH is one of the factors that should be used in determining BGH:

So all we are left with are various lines of evidence, no one of which is
definitive alone, but which, viewed all together, make it a not unreasonable hypothesis that genetic factors are strongly implicated in the average
Negro–white intelligence difference. (Jensen 1969b: 82)

Flynn 1980 criticizes Lewontin’s thought experiment:

the real message of Lewontin’s example is that we can ignore high [heritability] only if there exists a highly specific and highly unusual set of circumstances. Therefore, it is absurd to say that high [heritability] estimates within black and white respectively are irrelevant. Their relevance consists precisely of this: they force us to look for a plausible candidate for the role of [X-factor].

In summary, Lewontin only proves that WGH, on its own, doesn’t determine BGH because BGH could be due to factors that are not related to WGH. Everyone knows this, and nobody has claimed otherwise.

This is however wrong on an empirical level. We know that BGH is caused by the same factors that cause WGH because the Black-White IQ gap is measurement invariant.

This paper goes in-depth about WGH, BGH and measurement invariance, and proves Lewontin wrong.

Colman also quotes Templeton, who says:

“Sometimes the argument is made that because a trait is heritable
within two different populations that differ in their mean trait value,
then the average trait differences between the populations are also influenced by genetic factors (e.g., Herrnstein & Murray, 1994)

Of course, Herrnstein and Murray never make this claim. They say otherwise:

The heritability of individual differences in IQ does not necessarily mean that ethnic differences are also heritable.

The Bell Curve, page 299

Admixture

Colman rightfully says that admixture studies can serve as evidence for or against the Hereditarian Hypothesis. He cites Mackenzie 1984 for this. This is interesting because Mackenzie 1984 also heavily criticises Witty and Jenkins, the study Colman mentions next.

Witty and Jenkins 1935 is cited by Colman as “devastating for the Hereditarian interpretation” because it showed no correlation between white ancestry and IQ in African Americans. Mackenzie 1984 says Witty and Jenkins’ study was unrepresentative, not only because of it’s small sample size (63) but also because the blacks studied were an elite, and had higher than average white admixture (which supports the Hereditarian hypothesis). An in-depth critical analysis of this study is given here. Its results are equivocal.

What’s intriguing is that Colman mentions Witty and Jenkins (and wrongfully states Witty and Jenkins is the first study of its kind) but misses out the 16 admixture studies done around that time, which were included in Shuey 1966’s meta-analysis on admixture. Shuey’s meta-analysis supports the Hereditarian interpretation. 7 studies show strong support for it, 6 show moderate support, 2 showed evidence against it. I wouldn’t consider these studies too important, as they’re all old, have small sample sizes and questionable methodologies. However, if Colman considered Witty and Jenkins to be “devastating” for the Hereditarian interpretation, then these studies are devastating for the Environmentalist interpretation. There are also more recent admixture studies, like Lynn 2002 and Rowe 2002 which support a Hereditarian interpretation. Why these studies are not included is questionable, since Colman quotes Rushton and Jensen 2005, which mentions them. Is Colman cherrypicking studies that support his position? I find it hard to think otherwise.

Colman then mentions Scarr 1977 and Loehlin 1973. Both of these studies do not serve as evidence against the Hereditarian Hypothesis, as they are methodologically flawed, or didn’t contradict it (Scarr 1977 is compatible with the gap being ~50% genetic).

Thus, we come to the conclusion that admixture studies done before Colman wrote his paper generally support the Hereditarian hypothesis, and those that do not are methodologically flawed. It’s also worth mentioning that since then, new admixture studies of a much higher quality that use genomic data (or indirect measures for admixture) and high sample sizes have emerged, and they strongly support the Hereditarian hypothesis.

Adoption

On adoption, I am heavily skeptical it can prove much. First, Adoption gains are not on g while the gap is. This means that adoption studies inherently can’t disprove the Hereditarian hypothesis. Secondly, adoption studies have a very, very low quality in terms of allowing us to make conclusions based on them. Small sample sizes, unrepresentative samples, no follow ups, no biological parents’ IQ, etc. Thirdly, adoption studies can’t disprove the Environmentalist hypothesis, because environmentalists can always fall back on colorism or some other ad hoc factor. This is true with admixture studies too, but some have controlled for colorism by comparing within and between family variance or by simply accounting for skin colour.

Colman mentions Moore 1986. He doesn’t mention any of the fatal criticisms of it:

  • Very small sample size
  • The Blacks raised in Black homes had a higher IQ than Whites typically do in the general population, meaning the sample was unrepresentative
  • Black children adopted in a White home raised IQ by 13.5, larger than the gap is at that age.
  • Children were between 7 and 10, when IQ is largely environmental, not heritable.
  • Biological parents’ IQ was not recorded.

Similar situation with Eyferth 1961:

  • The children were extremely young (5–13). Intelligence is much less heritable at this age.
  • 20 to 25 percent of the ‘black’ fathers were French North Africans (Caucasians). This explains why the mixed-race children had higher IQs.
  • Soldiers face selection for intelligence (30% of black soldiers get rejected and 3% of white soldiers).
  • There was a sampling error because White girls had 8 lower IQ than White boys
  • Biological parents’ IQ was not recorded.

Finally, Colman mentions the MTRAS just to criticise it. Weird how he only criticises the study that disagrees with his position.

The study was flawed, as the researchers acknowledged in their original
article, because the Black children had been adopted later in life and had
therefore spent less time in their adoptive homes when they were tested

Age at adoption does not matter. See Jensen 1998 also.

the researchers concluded that the similar
average IQ scores of these two groups “support the view that the social
environment plays a dominant role in determining the average IQ level
of black children” (p. 739)

The original paper had an environmentalist interpretation which later changed to an equivocal interpretation after Scarr was convinced by Jensen. Scarr conceded that the MTRAS supports a Hereditarian interpretation, but environmentalist interpretations aren’t proven wrong because of colorism, the same thing I talked about above. I applaud Scarr for her sincerity and dedication to the scientific truth, unlike Colman.

The Flynn Effect

Colman says the Flynn effect “may also explain the Black–White IQ gap, in part at least”. This is wrong because the Flynn effect is not measurement invariant and not on g, while the Black-White IQ gap is.

“Sometimes the argument is made that because a trait is heritable
within two different populations that differ in their mean trait value,
then the average trait differences between the populations are also influenced by genetic factors (e.g., Herrnstein & Murray, 1994)

Of course, Herrnstein and Murray never make this claim. They say:

The heritability of individual differences in IQ does not necessarily mean that ethnic differences are also heritable.

Bell Curve, page 299

--

--