“But there have been very widespread and harsh criticisms of these studies, suggesting that Cofnas’s empirical premises are perhaps not as sound as he argues.”
I disagree with the inference from “there have been very widespread and harsh criticisms” to the “empirical premises are perhaps not...sound.” If many people rail against X because they hate X for political reasons, never heard the argument for X, or are simply afraid to admit that they secretly agree with X, that is not evidence that X is wrong.
Race denial has never been based on serious arguments. In 1690, John Locke asserted that all races are identical as a philosophical principle. This idea became an orthodoxy among left-wing intellectuals who were often open about the fact that their belief was based on wishful thinking. When John B. Watson endorsed race denial in 1924, he said “I am going beyond my facts and I admit it, but so have the advocates of the contrary and they have been doing it for many thousands of years.”
In 1950, UNESCO published a Statement on Race declaring that “science” shows that all races are the same. Many of the world’s leading geneticists and evolutionary biologists (Mayr, Darlington, Fisher...) said that they disagreed, but they were ignored. UNESCO’s statement was the basis of a propaganda campaign that established the post-WWII race taboo.
Race realism is simply Darwinism applied to the human species. Our bodies and brains adapt to local conditions. One can object to the word “race,” but ultimately what matters is the underlying reality, not how we use a word.
“One serious criticism leveled against such studies is that they only measure ‘general intelligence’, or G, which is deeply culturally dependent: for instance, an American will likely have a far easier time answering questions on an assessment of G than someone from the Philippines, even if the two are equally intelligent.”
This is why we cannot make definitive conclusions about population differences in intelligence by comparing test results across time and place. An IQ score is only meaningful in a particular cultural context. In 21st-century America, the same pattern of race differences emerges even when people go to the same school or are literally raised by the same parents in adoption studies. That’s not like giving an American IQ test to an illiterate Filipino peasant.
“Additionally, it’s unclear that G is a reliable measure of intelligence in the first place, since it measures intelligence only in a very narrow sense; there are many types of intelligence for which G does not and cannot account.”
The scientific literature on g is much less unclear than many people think it is. IQ is the best single predictor of outcomes that exists in social science. That doesn’t mean IQ captures everything about intelligence, or that IQ differences between races tell us everything we need to know about race differences in cognition. But differences in g have real, far-reaching implications.
“Of course, some objectors may hold that it’s nearly impossible to take environmental and societal factors off the table, given the long history of racial discrimination in the United States. In fact, it seems implausible that we could ever fully address these things in our research on intelligence: racial prejudices are so pervasive that we simply cannot identify and eliminate all their effects as biasing factors.”
As James Flynn observed, just saying “it’s racism” isn’t an explanation. If racism lowers black IQ, it must do so via some mechanism. For example, it causes them to have lower self-esteem or to suffer from stereotype threat, and *that* lowers IQ. The problem is that no one has come up with a remotely plausible mechanism that stands up to logical or empirical scrutiny.
“Furthermore, it may be unclear why we take the subjects of these intelligence studies to be good representatives of their racial groups, on average. What if we are, even by mere accident, choosing particularly unintelligent subjects?”
Psychometricians go to great lengths to administer tests to population-representative groups of people. They control for geography, socioeconomic status, etc. When it comes to IQ differences in the US, all studies show the same pattern of differences: black << white < East Asian. What is “unclear” about “why we take the subjects of these intelligence studies” to be representative?
“Even if there do turn out to be empirically incontestable differences in intelligence across races, one might think that we still have sociopolitical reasons to ignore these differences. After all, acknowledging them could enforce pernicious racial hierarchies and biases. But Cofnas rejects this idea. If we can see racial disparities but not acknowledge genetic causes, he claims, we will start creating societal tensions where there should be none.”
I quibble with the phraseology “Cofnas...*claims*.” The fact that race denial creates societal tensions where there should be none is observable reality that we experience every day.
You're right that "sound" is likely not the best word -- perhaps a better choice on my part would have been "clear" or "obvious." As I'm not a scientist (social or otherwise) and don't feel comfortable validating empirical claims, it was important to me to include the information that these empirical findings have been contested, but it's true that I could have found a better word.
I appreciate your responding to the objections I brought up! I tried to include in the article all the major objections that were posed in the podcast itself, and I'm grateful for your additional replies to these ideas.
I use the phrase "[x] claims" in every article I write, for any idea that is not a) totally intuitive to the vast majority of people or b) common knowledge.
I'm so glad you were able to look at this article, and thanks for your very thoughtful response!
Biologically "race" is not real. Human genetic variation is continuous, not discrete. Treating a constructed category as if it were a natural kind because it happens to correlate with something measurable, is wrong.
Even if we granted Cofnas's premise that human populations form discrete biological races (which we shouldn't, but for the sake of argument), this would not establish that racial differences in intelligence are genetically caused, for several reasons.
i) Race-marking genes and intelligence-related genes are largely independent. Intelligence is extraordinarily polygenic. GWAS studies identify thousands of SNPs across the genome, each with minuscule effect sizes. These intelligence-related variants are distributed broadly across populations and are not physically linked to for example pigmentation genes. There's no mechanism by which selection on skin color would drag along intelligence alleles, or vice versa.
The genes for visible race and the genes for cognitive ability are different genes, subject to different selective pressures, distributed differently across populations.
ii) correlation is not causation
So, "Even if it turns out to be the case that intelligence is not influenced by race itself, but rather by some racially-correlated genetic factor, Cofnas argues that his position still holds: after all, race would still be correlated with intelligence on this account." is a spurious correlation logic.
Even if we could identify genetic predictors of intelligence, they would track geographic ancestry, not race.
African Americans, for example, average ~20% European ancestry. If intelligence alleles were differentially distributed by continental origin, we would expect African Americans with more European ancestry to have higher IQs, holding socioeconomic factors constant. Studies that have tested this (e.g., using skin color as a proxy for ancestry) find no such relationship. This is strong evidence against the genetic hypothesis
You report Cofnas's claim that IQ disparities emerge in the early 20s because "heritability of IQ increases and environmental factors become less causal." This is a common misconception.
Heritability measures the proportion of variance in a trait within a specific population that is associated with genetic variance. It tells us nothing about the causes of between-group differences and of individual differences in other populations, nor about to he malleability of the trait
To actually link race to genetically-based intelligence differences, Cofnas would need to show:
1. That the alleles associated with cognitive function differ in frequency between racial groups — not just that groups differ in phenotypic outcomes, but that the specific genetic variants purportedly causing those outcomes are differentially distributed.
2. That these frequency differences are not themselves due to recent selection, drift, or linkage with other selected traits, e.g., that genes for sodium retention (selected in hot climates) don't happen to sit near genes involved in neural development.
3. To hat the effect sizes of these variants are consistent across environments, i.e., that a variant associated with +0.1 SD IQ in Europeans has the same effect in Africans, Asians, etc. (GxE interactions routinely violate this).
4. That the genetic differences explain a substantial portion of the observed phenotypic gap; requiring heritability estimates derived from between-group variance, not just within-group variance.
None of this exists! The GWAS literature on intelligence has been conducted overwhelmingly in European-ancestry samples, and attempts to create polygenic scores for non-European populations show dramatically reduced predictive validity, not because those populations lack the variants, but because the statistical associations are population-specific (a problem known as "population stratification").
Cofnas proposes that "natural selection" produced different "types" of intelligence in different environments. What selective pressures? What alleles? Which specific genetic changes occurred, and when? What trade-offs? Where's the evidence?
> And even if it turns out to be the case that intelligence is not influenced by race itself, but rather by some racially-correlated genetic factor, Cofnas argues that his position still holds: after all, race would still be correlated with intelligence on this account.
Not sure what this sentence is trying to convey. No one would say "race itself" influences intelligence, since race is a categorization scheme/property and has no causal powers. By analogy, a cucumber isn't green because it's a cucumber. It's green because of its DNA. Its DNA is also what makes it a cucumber. But we wouldn't say "plant species itself" causes coloration.
Imagine thinking this hasn’t been asked. Here is the scientific consensus: (1) Race is arbitrary and not a meaningful grouping of human variation. (2) IQ can be useful in some situations, but lacks the necessary construct validity to be linked to any known or actionable biological process. (3) Studies that assess heritability of these traits (aka Twin questionare studies) do not show, suggest, or point to ,any known developmental processes underlying them. They are out of step with modern molecular and developmental approaches leaving only a core group of rather fringe actors within the dying field of behavioral genetics using them. Asking this question in 2026 is akin to asking about the luminiferous aether, good luck with that.
On measuring intelligence: see my post "Real measures of intelligence" and the Woodcock-Johnson technical manual linked there, particularly the early sections on test norming. It turns out IQ is not quite a measure of intelligence, but the best IQ tests are very sophisticated in measuring not only g but dozens of subfactors.
A proper measure of intelligence is a measure of the difficulty of problems one can solve. One can measure the difficulty of problems by giving them to many people, along with many other questions that many other people also have tried to answer. Make a matrix with questions in columns, people in rows, a 1 if that person got it right, a 0 if wrong, put it in triangular form and you've ranked questions by difficulty and people by ability, a bit more math and the difficulty and ability of each question and person is quantified.
Unlike IQ, this allows comparing the intelligence of kids of various ages, which for some reason no one obects to, unlike racial comparisons. The 1 standard deviation Black-White difference is the same as the difference between an adult and the same person age 14. A 2 s.d. difference, to the border of mental disability at 70 IQ, or the typical score of Blacks in Africa, or equivalently between the most and least intelligent in a representative group of six people, is the same as between an adult and a ten year old. Intelligence differences, (which is to say differences in ability to be right, or differences in the difficulty of problems one can solve) between adults and between ancestral groups are as real, significant and large as between adults and children.
“But there have been very widespread and harsh criticisms of these studies, suggesting that Cofnas’s empirical premises are perhaps not as sound as he argues.”
I disagree with the inference from “there have been very widespread and harsh criticisms” to the “empirical premises are perhaps not...sound.” If many people rail against X because they hate X for political reasons, never heard the argument for X, or are simply afraid to admit that they secretly agree with X, that is not evidence that X is wrong.
Race denial has never been based on serious arguments. In 1690, John Locke asserted that all races are identical as a philosophical principle. This idea became an orthodoxy among left-wing intellectuals who were often open about the fact that their belief was based on wishful thinking. When John B. Watson endorsed race denial in 1924, he said “I am going beyond my facts and I admit it, but so have the advocates of the contrary and they have been doing it for many thousands of years.”
In 1950, UNESCO published a Statement on Race declaring that “science” shows that all races are the same. Many of the world’s leading geneticists and evolutionary biologists (Mayr, Darlington, Fisher...) said that they disagreed, but they were ignored. UNESCO’s statement was the basis of a propaganda campaign that established the post-WWII race taboo.
Race realism is simply Darwinism applied to the human species. Our bodies and brains adapt to local conditions. One can object to the word “race,” but ultimately what matters is the underlying reality, not how we use a word.
“One serious criticism leveled against such studies is that they only measure ‘general intelligence’, or G, which is deeply culturally dependent: for instance, an American will likely have a far easier time answering questions on an assessment of G than someone from the Philippines, even if the two are equally intelligent.”
This is why we cannot make definitive conclusions about population differences in intelligence by comparing test results across time and place. An IQ score is only meaningful in a particular cultural context. In 21st-century America, the same pattern of race differences emerges even when people go to the same school or are literally raised by the same parents in adoption studies. That’s not like giving an American IQ test to an illiterate Filipino peasant.
“Additionally, it’s unclear that G is a reliable measure of intelligence in the first place, since it measures intelligence only in a very narrow sense; there are many types of intelligence for which G does not and cannot account.”
The scientific literature on g is much less unclear than many people think it is. IQ is the best single predictor of outcomes that exists in social science. That doesn’t mean IQ captures everything about intelligence, or that IQ differences between races tell us everything we need to know about race differences in cognition. But differences in g have real, far-reaching implications.
“Of course, some objectors may hold that it’s nearly impossible to take environmental and societal factors off the table, given the long history of racial discrimination in the United States. In fact, it seems implausible that we could ever fully address these things in our research on intelligence: racial prejudices are so pervasive that we simply cannot identify and eliminate all their effects as biasing factors.”
As James Flynn observed, just saying “it’s racism” isn’t an explanation. If racism lowers black IQ, it must do so via some mechanism. For example, it causes them to have lower self-esteem or to suffer from stereotype threat, and *that* lowers IQ. The problem is that no one has come up with a remotely plausible mechanism that stands up to logical or empirical scrutiny.
“Furthermore, it may be unclear why we take the subjects of these intelligence studies to be good representatives of their racial groups, on average. What if we are, even by mere accident, choosing particularly unintelligent subjects?”
Psychometricians go to great lengths to administer tests to population-representative groups of people. They control for geography, socioeconomic status, etc. When it comes to IQ differences in the US, all studies show the same pattern of differences: black << white < East Asian. What is “unclear” about “why we take the subjects of these intelligence studies” to be representative?
“Even if there do turn out to be empirically incontestable differences in intelligence across races, one might think that we still have sociopolitical reasons to ignore these differences. After all, acknowledging them could enforce pernicious racial hierarchies and biases. But Cofnas rejects this idea. If we can see racial disparities but not acknowledge genetic causes, he claims, we will start creating societal tensions where there should be none.”
I quibble with the phraseology “Cofnas...*claims*.” The fact that race denial creates societal tensions where there should be none is observable reality that we experience every day.
You're right that "sound" is likely not the best word -- perhaps a better choice on my part would have been "clear" or "obvious." As I'm not a scientist (social or otherwise) and don't feel comfortable validating empirical claims, it was important to me to include the information that these empirical findings have been contested, but it's true that I could have found a better word.
I appreciate your responding to the objections I brought up! I tried to include in the article all the major objections that were posed in the podcast itself, and I'm grateful for your additional replies to these ideas.
I use the phrase "[x] claims" in every article I write, for any idea that is not a) totally intuitive to the vast majority of people or b) common knowledge.
I'm so glad you were able to look at this article, and thanks for your very thoughtful response!
Biologically "race" is not real. Human genetic variation is continuous, not discrete. Treating a constructed category as if it were a natural kind because it happens to correlate with something measurable, is wrong.
Even if we granted Cofnas's premise that human populations form discrete biological races (which we shouldn't, but for the sake of argument), this would not establish that racial differences in intelligence are genetically caused, for several reasons.
i) Race-marking genes and intelligence-related genes are largely independent. Intelligence is extraordinarily polygenic. GWAS studies identify thousands of SNPs across the genome, each with minuscule effect sizes. These intelligence-related variants are distributed broadly across populations and are not physically linked to for example pigmentation genes. There's no mechanism by which selection on skin color would drag along intelligence alleles, or vice versa.
The genes for visible race and the genes for cognitive ability are different genes, subject to different selective pressures, distributed differently across populations.
ii) correlation is not causation
So, "Even if it turns out to be the case that intelligence is not influenced by race itself, but rather by some racially-correlated genetic factor, Cofnas argues that his position still holds: after all, race would still be correlated with intelligence on this account." is a spurious correlation logic.
Even if we could identify genetic predictors of intelligence, they would track geographic ancestry, not race.
African Americans, for example, average ~20% European ancestry. If intelligence alleles were differentially distributed by continental origin, we would expect African Americans with more European ancestry to have higher IQs, holding socioeconomic factors constant. Studies that have tested this (e.g., using skin color as a proxy for ancestry) find no such relationship. This is strong evidence against the genetic hypothesis
You report Cofnas's claim that IQ disparities emerge in the early 20s because "heritability of IQ increases and environmental factors become less causal." This is a common misconception.
Heritability measures the proportion of variance in a trait within a specific population that is associated with genetic variance. It tells us nothing about the causes of between-group differences and of individual differences in other populations, nor about to he malleability of the trait
To actually link race to genetically-based intelligence differences, Cofnas would need to show:
1. That the alleles associated with cognitive function differ in frequency between racial groups — not just that groups differ in phenotypic outcomes, but that the specific genetic variants purportedly causing those outcomes are differentially distributed.
2. That these frequency differences are not themselves due to recent selection, drift, or linkage with other selected traits, e.g., that genes for sodium retention (selected in hot climates) don't happen to sit near genes involved in neural development.
3. To hat the effect sizes of these variants are consistent across environments, i.e., that a variant associated with +0.1 SD IQ in Europeans has the same effect in Africans, Asians, etc. (GxE interactions routinely violate this).
4. That the genetic differences explain a substantial portion of the observed phenotypic gap; requiring heritability estimates derived from between-group variance, not just within-group variance.
None of this exists! The GWAS literature on intelligence has been conducted overwhelmingly in European-ancestry samples, and attempts to create polygenic scores for non-European populations show dramatically reduced predictive validity, not because those populations lack the variants, but because the statistical associations are population-specific (a problem known as "population stratification").
Cofnas proposes that "natural selection" produced different "types" of intelligence in different environments. What selective pressures? What alleles? Which specific genetic changes occurred, and when? What trade-offs? Where's the evidence?
> And even if it turns out to be the case that intelligence is not influenced by race itself, but rather by some racially-correlated genetic factor, Cofnas argues that his position still holds: after all, race would still be correlated with intelligence on this account.
Not sure what this sentence is trying to convey. No one would say "race itself" influences intelligence, since race is a categorization scheme/property and has no causal powers. By analogy, a cucumber isn't green because it's a cucumber. It's green because of its DNA. Its DNA is also what makes it a cucumber. But we wouldn't say "plant species itself" causes coloration.
Imagine thinking this hasn’t been asked. Here is the scientific consensus: (1) Race is arbitrary and not a meaningful grouping of human variation. (2) IQ can be useful in some situations, but lacks the necessary construct validity to be linked to any known or actionable biological process. (3) Studies that assess heritability of these traits (aka Twin questionare studies) do not show, suggest, or point to ,any known developmental processes underlying them. They are out of step with modern molecular and developmental approaches leaving only a core group of rather fringe actors within the dying field of behavioral genetics using them. Asking this question in 2026 is akin to asking about the luminiferous aether, good luck with that.
On measuring intelligence: see my post "Real measures of intelligence" and the Woodcock-Johnson technical manual linked there, particularly the early sections on test norming. It turns out IQ is not quite a measure of intelligence, but the best IQ tests are very sophisticated in measuring not only g but dozens of subfactors.
A proper measure of intelligence is a measure of the difficulty of problems one can solve. One can measure the difficulty of problems by giving them to many people, along with many other questions that many other people also have tried to answer. Make a matrix with questions in columns, people in rows, a 1 if that person got it right, a 0 if wrong, put it in triangular form and you've ranked questions by difficulty and people by ability, a bit more math and the difficulty and ability of each question and person is quantified.
Unlike IQ, this allows comparing the intelligence of kids of various ages, which for some reason no one obects to, unlike racial comparisons. The 1 standard deviation Black-White difference is the same as the difference between an adult and the same person age 14. A 2 s.d. difference, to the border of mental disability at 70 IQ, or the typical score of Blacks in Africa, or equivalently between the most and least intelligent in a representative group of six people, is the same as between an adult and a ten year old. Intelligence differences, (which is to say differences in ability to be right, or differences in the difficulty of problems one can solve) between adults and between ancestral groups are as real, significant and large as between adults and children.