Three genetic variants linked to IQ, and they vary by race

[Jokodo]

It is not wishful thinking, it is supported by your own data which shows that the relationships cancel out and the net effect is not different from zero. Random error differs most from zero with small samples, so the odds are much greater that the net effect will approach zero as the samples of alleles gets larger. It is quite plausible that the correlations with IQ differences between races are non-causal and spurious and that is why they are so unreliable and all over the spectrum from high to low to negative. IOW, even if those alleles have a causal impact on within group differences in IQ, they might have no causal impact on between group difference and the correlations for individual alleles are spurious products of third variable factors.

There is large random error given any proposed distribution for such a small set of alleles. If the results are expected from distributions of alleles equal to 0 IQ differences, that point would be hardly relevant if they are likewise expected from ANY distribution. It is night, and the closet is dark and black--exactly what we expect if there is a monster lurking in it.

So you're revoking your claim from the OP that "My analysis of these three SNPs makes it more plausible [that intelligence variations among the races are genetic]." - if the results are expected from just about any distribution, they can't be used as evidence for one or the other.

Or are you?

Also, from the OP again "After they are found, they will be mapped to the races just as easily as I have done with those three SNPs, and we need to expect that the results will tend to favor the races of high measured intelligence" is a case of circular reasoning. If there is a genetic basis for the correlations of measured intelligence and race, then yes, we'd also expect a correlation between the racial distributions of all genes involved in intelligence weighted by their effect and those measurements. But the existence of genes individual genes which contribute to intelligence and have an uneven geographical distribution does not make your premise more plausible.

 

[ApostateAbe]

ApostateAbe,

I will email Dr. Graves and ask for a source on population geneticists using the value of 0.25 as the threshold for classification of biological races.



The most relevant point for this thread is that genetic variants which account for the distribution of intelligence are not differentiated racially.

At least 1/3 of the 20,000 genes of the human exome are actively expressed in the human brain (Institute of Neurological and Brain Disorders, 2012). Given that brain function undoubtedly has something to do with cognitive performance, the notion that we can identify a small number of candidate genes which will capture the majority of the variation in the normal range of human intelligence is absurd. Clearly there are a number of mutations which occur in specific genes that reduce intelligence. These mutations have high penetrance and have pleiotropic impacts on a variety of physiological systems. Such pathological mutations have been well-studied and we know a great deal about how natural selection acts to reduce the frequency of such alleles. In this regard, the brain is no different from any other organ. However, comparing high penetrance mutations that are pathologic in character to genetic variants which account for normal variation in cognitive function is like comparing apples and watermelons.

Source: Race, Genomics and Intelligence: Slight Return by Joseph Graves

Graves' position as expressed in that quote (and I think we discussed it before) is that identifying the alleles that account for racial differences in intelligence would be impossible if they exist. Not that they don't exist. His reasoning seems to be that the relationship of the alleles to the phenotypes are too complex, because of such things as pleiotropy and linkage disequilibrium (LD happens when at least two alleles are related to each other and have increased phenotypic effect when working together but low effect when isolated from each other). Graves made his point using as an example a genome-wide association study (GWAS) related to human height. Height variations are thought to be 80% heritable, but the best GWAS could identify the alleles that together account for only 10% of the variance in height. I decided to investigate further. One of the studies listed in Graves' References is Yang et al's "Common SNPs explain a large proportion of the heritability for human height," 2010. The abstract is here and a discussion by some of the co-authors is here. In summary, the authors concluded that the lowness of 10% is due to LD, and the combined heritability estimate is increased to 45% when all the SNPs are analyzed together. It is a controversial finding, but it is mystery to me why Graves would proceed as though this finding does not exist while also citing this study to make the opposite point on page 9. Graves cited an example that seems to work against his point. It really is possible to identify a large portion of the SNPs for height in spite of the complexities, and, if it can be done for height, then it can be done for intelligence. And of course they can be found to vary by race.

 

[ApostateAbe]

There is large random error given any proposed distribution for such a small set of alleles. If the results are expected from distributions of alleles equal to 0 IQ differences, that point would be hardly relevant if they are likewise expected from ANY distribution. It is night, and the closet is dark and black--exactly what we expect if there is a monster lurking in it.

So you're revoking your claim from the OP that "My analysis of these three SNPs makes it more plausible [that intelligence variations among the races are genetic]." - if the results are expected from just about any distribution, they can't be used as evidence for one or the other.

Or are you?

Also, from the OP again "After they are found, they will be mapped to the races just as easily as I have done with those three SNPs, and we need to expect that the results will tend to favor the races of high measured intelligence" is a case of circular reasoning. If there is a genetic basis for the correlations of measured intelligence and race, then yes, we'd also expect a correlation between the racial distributions of all genes involved in intelligence weighted by their effect and those measurements. But the existence of genes individual genes which contribute to intelligence and have an uneven geographical distribution does not make your premise more plausible.

The argument is as follows: a small number of the total alleles that code for intelligence have been found, and each of them are unevenly distributed among the races. If so, then it is probable that almost ALL alleles for intelligence are each unevenly distributed among the races. There are many possible phenotypic outcomes that can follow from this.

(1) Races of higher measured intelligence tend to have more alleles that code for higher measured intelligence than races of lower measured intelligence.
(2) Races of higher measured intelligence tend to have fewer alleles that code for higher measured intelligence than races of lower measured intelligence.
(3) Races of higher measured intelligence tend to have an exactly equal amount of alleles that code for higher measured intelligence than races of lower measured intelligence.

If outcome #1 is true, then there is a match between genotype and a phenotype among the races, where the phenotype is known to be mostly heritable within groups. If outcome #2 is true, then there is a mismatch between genotype and phenotype among the races. If #3 is true, then I take it to be akin to throwing a hundred thousand or more quarters on the ground and they all land on heads. Why would we expect exact equality, besides the obvious ideological interest?

 

[Jokodo]

So you're revoking your claim from the OP that "My analysis of these three SNPs makes it more plausible [that intelligence variations among the races are genetic]." - if the results are expected from just about any distribution, they can't be used as evidence for one or the other.

Or are you?

Also, from the OP again "After they are found, they will be mapped to the races just as easily as I have done with those three SNPs, and we need to expect that the results will tend to favor the races of high measured intelligence" is a case of circular reasoning. If there is a genetic basis for the correlations of measured intelligence and race, then yes, we'd also expect a correlation between the racial distributions of all genes involved in intelligence weighted by their effect and those measurements. But the existence of genes individual genes which contribute to intelligence and have an uneven geographical distribution does not make your premise more plausible.

The argument is as follows: a small number of the total alleles that code for intelligence have been found, and each of them are unevenly distributed among the races. If so, then it is probable that almost ALL alleles for intelligence are each unevenly distributed among the races. There are many possible phenotypic outcomes that can follow from this.

(1) Races of higher measured intelligence tend to have more alleles that code for higher measured intelligence than races of lower measured intelligence.
(2) Races of higher measured intelligence tend to have fewer alleles that code for higher measured intelligence than races of lower measured intelligence.
(3) Races of higher measured intelligence tend to have an exactly equal amount of alleles that code for higher measured intelligence than races of lower measured intelligence.

If outcome #1 is true, then there is a match between genotype and a phenotype among the races, where the phenotype is known to be mostly heritable within groups. If outcome #2 is true, then there is a mismatch between genotype and phenotype among the races. If #3 is true, then I take it to be akin to throwing a hundred thousand or more quarters on the ground and they all land on heads. Why would we expect exact equality, besides the obvious ideological interest?

If you have two players throw a die 1000 times, the chance that the sum of their eyes will be exactly equal is near zero (less than half a %, to be more precise). But if player A gets a dollar from the bank everytime he throws a four or higher, and player B gets a dollar everytime he hits 5 or higher, you cannot conclude that player A must have been luckier because he has a higher pile of money in front of him. That is, in effect, what you're attempting to do here.

I never said that I expect an "an exactly equal amount of alleles that code for higher measured intelligence" - if you believe otherwise, I challenge you to find a quote saying so. Indeed, the chance that they are exactly equal is negligible - just like the chance that the amount

What I did say is that the measured distributions tells us nothing about in which direction those differences are going to be tilted. As with the dice and money piles above.

 

[ApostateAbe]

The argument is as follows: a small number of the total alleles that code for intelligence have been found, and each of them are unevenly distributed among the races. If so, then it is probable that almost ALL alleles for intelligence are each unevenly distributed among the races. There are many possible phenotypic outcomes that can follow from this.

(1) Races of higher measured intelligence tend to have more alleles that code for higher measured intelligence than races of lower measured intelligence.
(2) Races of higher measured intelligence tend to have fewer alleles that code for higher measured intelligence than races of lower measured intelligence.
(3) Races of higher measured intelligence tend to have an exactly equal amount of alleles that code for higher measured intelligence than races of lower measured intelligence.

If outcome #1 is true, then there is a match between genotype and a phenotype among the races, where the phenotype is known to be mostly heritable within groups. If outcome #2 is true, then there is a mismatch between genotype and phenotype among the races. If #3 is true, then I take it to be akin to throwing a hundred thousand or more quarters on the ground and they all land on heads. Why would we expect exact equality, besides the obvious ideological interest?

If you have two players throw a die 1000 times, the chance that the sum of their eyes will be exactly equal is near zero (less than half a %, to be more precise). But if player A gets a dollar from the bank everytime he throws a four or higher, and player B gets a dollar everytime he hits 5 or higher, you cannot conclude that player A must have been luckier because he has a higher pile of money in front of him. That is, in effect, what you're attempting to do here.

I never said that I expect an "an exactly equal amount of alleles that code for higher measured intelligence" - if you believe otherwise, I challenge you to find a quote saying so. Indeed, the chance that they are exactly equal is negligible - just like the chance that the amount

What I did say is that the measured distributions tells us nothing about in which direction those differences are going to be tilted. As with the dice and money piles above.

Is it your counterposition that some alleles count for more intelligence than others? That is most certainly true, and I didn't mean to imply that all alleles for intelligence confer the same amount of intelligence--it is just more difficult to express the points understandably and I took a shortcut. Is it your position that, given equal environments, that every race would have either exactly equal or close to exactly equal intelligence? If so, why? If not, then we are on the same page.

 

[Jokodo]

If you have two players throw a die 1000 times, the chance that the sum of their eyes will be exactly equal is near zero (less than half a %, to be more precise). But if player A gets a dollar from the bank everytime he throws a four or higher, and player B gets a dollar everytime he hits 5 or higher, you cannot conclude that player A must have been luckier because he has a higher pile of money in front of him. That is, in effect, what you're attempting to do here.

I never said that I expect an "an exactly equal amount of alleles that code for higher measured intelligence" - if you believe otherwise, I challenge you to find a quote saying so. Indeed, the chance that they are exactly equal is negligible - just like the chance that the amount

What I did say is that the measured distributions tells us nothing about in which direction those differences are going to be tilted. As with the dice and money piles above.

Is it your counterposition that some alleles count for more intelligence than others?

No, that's not my counterposition, since I have no reason to think you believe otherwise.

That is most certainly true, and I didn't mean to imply that all alleles for intelligence confer the same amount of intelligence--it is just more difficult to express the points understandably and I took a shortcut. Is it your position that, given equal environments, that every race would have either exactly equal or close to exactly equal intelligence? If so, why?

No, I don't believe they're exactly equal. In the same way that I don't believe that the average intelligence of brown-eyed and blue-eyed, or left-handed and right-handed people within a "race" is going to be exactly equal. In the same way I don't believe that the average intelligence of people born in February and people born in September is going to be exactly equal - and saying so doesn't give credence to astrology; because such a state of affairs would be statistically highly improbable given finite sets and random variation.

I do believe that your sacred "measurements" tell us exactly nothing about the underlying distribution of genes contributing to higher intelligence. I have good reason to do so.

Because (a) we know that environmental factors play a major role; (b) we know that the environments of blacks and whites (in the US or globally) are not equal; therefore we know that there's going to be a large environmental factor in whatever measured differences we find (and don't fool yourself, Lynn's data doesn't count as "measured differences" for a good number of reasons); we don't know exactly how large the difference predicted from environment should be (too many parameters, basically), so it's anyone's guess whether there is a positive or negative residue.

In other words, your data are perfectly compatible with a scenario in which the "races" with "lower measured intelligence" actually would perform better given equal environments. Your statement that "we need to expect that the results will tend to favor the races of high measured intelligence" is no different from someone saying that we need to expect that player A had better luck with his dice because we see a larger pile of money in front of him, when we all know that there were different payout matrices for the two player: Wishful thinking at best.

 

[ApostateAbe]

Jokodo, do you believe that, given equal environmental distributes, that every race would have APPROXIMATELY equal average intelligences? As in: any small difference in the averages we find would be attributable to random error?

 

[Jokodo]

Jokodo, do you believe that, given equal environmental distributes, that every race would have APPROXIMATELY equal average intelligences? As in: any small difference in the averages we find would be attributable to random error?

I have no reason to believe otherwise and neither have you. There could even be systematic differences, but they might still not correlate with the measured differences. You have given no reason to believe they do.

 

[ApostateAbe]

Jokodo, do you believe that, given equal environmental distributes, that every race would have APPROXIMATELY equal average intelligences? As in: any small difference in the averages we find would be attributable to random error?

I have no reason to believe otherwise and neither have you. There could even be systematic differences, but they might still not correlate with the measured differences. You have given no reason to believe they do.

OK. To me, differences in phenotype would make a big difference in the probabilities of each prediction, if the phenotypic variations are shown to be mostly heritable per twin studies. The more heritable it is, the more likely it is that the racial differences are mostly due to genes not environment. There are confirmed differences in environment, and we may expect that would count for a lot based on a gut feeling, except in part for the transracial adoption study of Scarr and Weinberg, showing average adult IQs of black adopted children close to the IQs of blacks in the surrounding population and a standard deviation below their white siblings. And intermediate-race children had an intermediate average IQ. See Weinberg, Scarr and Waldman's "The Minnesota Transracial Adoption Study: A follow-up of IQ test performance at adolescence," 1992, and Richard Lynn's "Some Reinterpretations of the Minnesota Transracial Adoption Study," 1994. We know that environments play a minority part in IQ, but we do NOT know that such environmental components between American blacks and American whites play a part, and in fact the claim has seemingly been falsified.

A great analogy that Joseph Graves brought up is the set of genes for height. Variations in height are about 80% heritable within groups, but the 20% remainder could still count for a lot between groups. Average human height has grown significantly for all groups within the last hundred years, much like IQ. But, genotypic differences to match the phenotypic differences is still the most probable position until proven otherwise, in my opinion. We should not be defaulting to equal genotypic height when the phenotypic reality is inequality of height. If we don't directly observe the genotypic differences, then we should be defaulting to the best indicators of genotypic differences, which are phenotypic differences. If you disagree, then make your prediction: using the 1000 Genomes Project Browser, I will examine the racial frequencies of at least twenty alleles with the greatest magnitude of beta coefficients reported in Supplemental Table 1 of Allen et al's "Hundreds of variants clustered in genomic loci and biological pathways affect human height," 2010. I predict a positive correlation. Is it just as likely that there will be a negative correlation with average height as a positive correlation, in your opinion?

 

[EgalitarianJay]

I got a reply from Dr. Graves:

Dear EgalitarianJay, This is a common error I have dealt with many times in the past. Wright discusses the level of variation in Fst and its meaning on page 85 in the chapter entitled: Genetic Variability in Natural Populations: Methods. On that page he does not mention the terms "race" or "subspecies". Instead he talks about F = 0.25 as an arbitrary value above which there is very great differentiation.


Sewall Wright was clearly a racialist (one who believed that biological races existed within our species). This is demonstrated by his discussions of Racial Differentiation in Mankind in chapter 10. He recognized that Fst in humans was pretty small, for the genes he examined in that chapter Fst = 0.1248, and he understood the principle of discordance (see discussion on 449--450.) While I did not know Wright personally, I know and work closely with many people who knew him and worked closely with him. The determination that Fst = 0.250 for the boundary of racial/subspecies identification is really a post-Wright phenomenon.


Sincerely,


Dr. Joseph L. Graves Jr.

 

[ApostateAbe]

Looks like he didn't answer your question, Jay. Bummer.

 

[ApostateAbe]

Graves wrote correctly about Sewall Wright. Here is the first page of his chapter about human races.

Show

 

[EgalitarianJay]

Graves wrote correctly about Sewall Wright. Here is the first page of his chapter about human races.

Show

Your link isn't working for me but I saw the picture in your Dropbox. I have some other questions for Graves that I want him to answer. When he does I will ask him for sources on the threshold for biological races.

 

[Trodon]

Jokodo, do you believe that, given equal environmental distributes, that every race would have APPROXIMATELY equal average intelligences? As in: any small difference in the averages we find would be attributable to random error?

I have no reason to believe otherwise and neither have you. There could even be systematic differences, but they might still not correlate with the measured differences. You have given no reason to believe they do.

The reason to believe otherwise is that racial differences in average intelligence are durable across time and space in ways that give more evidence of genetic causation, than of environmental differences.

In the United States every effort to bring blacks up to white levels of performance has failed. The most recent failing effort has been No Child Left Behind.

One can always claim that No Child Left Behind was poorly designed, that not enough money was spent on it, etc., etc. What no one can do is to point to a single effort that has succeeded in closing the race gap.

Members of some races tend to score higher than members of other races in all of the mental aptitude tests, however they are designed. Different test scores correspond to different performance levels academically and economically.

 

[Don2 (Don1 Revised)]

... closing the race gap ...

Flynn Effect.

 

[Trodon]

... closing the race gap ...

Flynn Effect.

The Flynn Effect has not closed the race gap in IQ scores. It has raised them for every race fairly equally.

SAT scores have risen somewhat since 1986 - 87, but the gap between white and black averages has increased slightly.

https://nces.ed.gov/fastfacts/display.asp?id=171

 

[Canard DuJour]

Flynn Effect.

The Flynn Effect has not closed the race gap in IQ scores. It has raised them for every race fairly equally.

Yes, by more than the initial difference between them in under a century. So it is not true that "racial differences in average intelligence are durable across time and space in ways that give more evidence of genetic causation, than of environmental differences."

 

[ApostateAbe]

The Flynn Effect has not closed the race gap in IQ scores. It has raised them for every race fairly equally.

Yes, by more than the initial difference between them in under a century. So it is not true that "racial differences in average intelligence are durable across time and space in ways that give more evidence of genetic causation, than of environmental differences."

The secular rise in IQ lends some weight to the environmental hypothesis, but nevertheless it is true that the DIFFERENCES have remained durable across time and space. Height is much the same way. I think rational people take it for granted that racial height differences are largely genetic--or are Asians short because of all the rice they eat?--but, if there was a strong ideological interest in denying the genetics of racial height differences, the same argument would apply. There has been a large secular rise in height among all races over the last hundred years, greater than the differences between races.

 

[Don2 (Don1 Revised)]

Yes, by more than the initial difference between them in under a century. So it is not true that "racial differences in average intelligence are durable across time and space in ways that give more evidence of genetic causation, than of environmental differences."

The secular rise in IQ lends some weight to the environmental hypothesis, but nevertheless it is true that the DIFFERENCES have remained durable across time and space.

It sounds like you are saying by differences that there are differences not the measurements of the differences. The measurements have changed. They have also changed along with various other things such as better world literacy, better world health, better income. There have also been specific studies finding one or the other very correlated to the iq gap: such as pathogens vs iq gap and literacy change versus gap. There are still many gaps in these things correlated to iq difference. So one cannot discount them, if one is assuming iq is a real thing anyway.

Height is much the same way. I think rational people take it for granted that racial height differences are largely genetic--or are Asians short because of all the rice they eat?--but, if there was a strong ideological interest in denying the genetics of racial height differences, the same argument would apply. There has been a large secular rise in height among all races over the last hundred years, greater than the differences between races.

I read that when American colonials encountered Native Americans, they saw them as a tall race and there were measurements that went along with that. After a century of living in America, it was the colonials' descendants who had become tall.

Regarding rice, don't forget soy. Soy is related to production of estrogen which could be related to other observations of racial difference. In fact, diet is very related to estrogen and testosterone. Diet is also related to cognitive development, even during gestation.

 

[Canard DuJour]

Yes, by more than the initial difference between them in under a century. So it is not true that "racial differences in average intelligence are durable across time and space in ways that give more evidence of genetic causation, than of environmental differences."

The secular rise in IQ lends some weight to the environmental hypothesis, but nevertheless it is true that the DIFFERENCES have remained durable across time and space.

Not, however, "in ways that give more evidence of genetic causation, than of environmental differences," but in ways that leave the question open.

Height is much the same way. I think rational people take it for granted that racial height differences are largely genetic--or are Asians short because of all the rice they eat?--but, if there was a strong ideological interest in denying the genetics of racial height differences, the same argument would apply. There has been a large secular rise in height among all races over the last hundred years, greater than the differences between races.

And if that weren't almost entirely down to a single easily controlled for factor (childhood nutrition), but multifactorial with umpteen confounders and counterexamples, then the cause of a persistent differential would indeed remain an open scientific question. The idea that we can extrapolate from height to any other persistent differential, however, remains pseudoscience.