What would settle the IQ hereditarianism debate?

In August 2026 the Triggernometry podcast published a nearly two-hour interview with the philosopher Nathan Cofnas, under the title “The Controversial Race Scientist Who Exposed Jason Arday.”¹ Roughly half of it concerns the Arday affair at Cambridge. That half makes no IQ-science claims, and this page does not cover it. The other half is the most careful popular statement currently available of the hereditarian position on group differences in intelligence — the view that genetic differences between populations play a non-trivial role in average differences in measured IQ.

That position is either suppressed pseudoscience or suppressed science, depending on who is talking, and the shouting usually starts before anyone states what the evidence actually is. This page does the slower thing. It lays out who Cofnas is — including the parts of his record that fit neither side’s caricature — then takes each of his load-bearing claims, pairs it with the published research behind it, and sets the strongest published counters beside it, steelmanned rather than strawmanned. It ends where the evidence ends, which is not where either side’s loudest voices say it does.

Two ground rules, stated once. Every reference at the foot of this page was verified against the publisher’s or journal’s own record on 27 August 2026; anything the underlying research could not verify to that standard was excluded, and the exclusions are named. And this page issues no verdict on the empirical question. Debby’s job is to record the strongest case and the strongest challenge on file, under the same rubric for both.

Who Nathan Cofnas is

Cofnas is a philosopher of biology: a doctorate from Oxford, then a Leverhulme Early Career Fellowship in Cambridge’s Faculty of Philosophy (2022–2025), and now a postdoctoral researcher in the Department of Philosophy and Moral Sciences at Ghent University.² On the podcast he is explicit that he speaks for himself, not for his university.

The part of his record that neither side’s caricature includes: his first academic mobbing came from the right, not the left. In 2018 he published a peer-reviewed critique — the first academic one, by his account — of the evolutionary psychologist Kevin MacDonald, whose trilogy argues that Judaism is a “group evolutionary strategy” adapted to undermine gentile societies.³ Cofnas’s paper argues that the trilogy rests on systematically misrepresented sources and cherry-picked facts, and that the data are better explained by what he calls a default hypothesis — Jewish overrepresentation in intellectual and political movements as a product of above-average intelligence and urban concentration, no group strategy required. As he puts it on the podcast:

my first mobbing actually was by the right not the left. It was in 2018 when I published a critique of the evolutionary psychologist Kevin McDonald [MacDonald]¹

Two things follow for an honest map. A portrayal of Cofnas as a generic race-science advocate is incomplete against his own record: debunking anti-Semitic pseudo-scholarship is part of what he does, with follow-up replies published through 2023. And the asymmetry is worth noticing — his alternative to MacDonald is itself a hereditarian-flavored explanation, applied in a direction the far right dislikes. His framework does not cut only one way.

The paper he is best known for is not an empirical claim at all. In December 2019, while a doctoral student, he published “Research on group differences in intelligence: A defense of free inquiry” in Philosophical Psychology.⁴ Its argument is meta-level: refusing to investigate possible genetic contributions to group differences has its own costs — policy built on a false premise, scholarship left unprepared, a public that learns to distrust the referees. Truth, he argues, has intrinsic value, and suppression backfires. The paper does not claim that any gap is genetic, and reading it as an empirical hereditarian paper is a category error — as is citing it as evidence for one. The response was a petition and boycott call demanding retraction, with an open letter arguing the paper had failed peer review’s own norms,⁵ and, months later, the resignation of Cees van Leeuwen, one of the journal’s two editors, after twenty-five years — in protest at the journal’s decision to publish a critical reply⁶ while bypassing editor moderation.⁷ (Cofnas’s podcast telling adds that the other editor apologised and was fired. The apology is on the record, in a narrower form: Mitchell Herschbach published an editorial note acknowledging flaws in the paper’s peer review and promising procedural improvements — Cofnas, writing in 2020, called it “a groveling apology.”⁸ The firing is on no record: no outlet reported one, and the editorship changed hands in an announced succession effective January 2022.⁹ This page does not rely on the unverified detail.)

He had already made the sociology-of-science version of the argument in 2016: a paper in Foundations of Science documenting cases, from the 1970s onward, in which moral and political commitments visibly drove the acceptance or rejection of intelligence research — concluding that “science is self-correcting” is not reliably true in this area.¹⁰

The fellowship controversy, stated neutrally and once: in 2024, after a blog post on race and intelligence, a student petition and formal complaints triggered investigations by the university and the Leverhulme Trust; Emmanuel College ended its research association with him;¹¹ and the university’s own inquiry concluded that his published views, “while seen by many as offensive, did not breach the law and did not contravene University regulations designed to uphold free speech.”¹² That sequence is context for the free-inquiry question. It is not evidence in the empirical one, in either direction.

What he argued on the podcast

The definition is narrower than the reputation. Hereditarianism, as Cofnas states it:

hereditarianism with respect to group differences refers to the idea that genetic differences play a non-trivial role in average differences in socially relevant traits such as intelligence. That’s it.¹

Note what that definition does not contain: no claim that races are discrete, no claim that environment is powerless, no policy conclusion. On race itself he is explicit that the categories have no sharp lines — his analogy is languages and dialects, overlapping and blending, classification partly a matter of interest — “but… the underlying reality which is that there are patterns of differences that really exists in the world.” The entire fight is over “non-trivial” and over what the evidence shows. That is a much smaller target than the one usually shot at, and this page aims at the smaller target.

The magnitude claim — and the hedge that follows it. Asked how much of the group difference is genetic:

I think it should be the majority of differences.¹

Pressed — “So over 50%?” — he answers: “Yeah… so intelligence is over 50% heritable and the differences between groups are a reflection of that.” Then, seconds later, he hedges: heritability “is tied to a particular environment and particular population. It can change over time.” The answer places within-group heritability next to between-group causation as if the first supported the second; the hedge concedes the very point that severs them. That inference is the debate’s oldest crux, and the counter-section below shows where it now stands. The lineage behind the claim runs from Arthur Jensen’s 1969 Harvard Educational Review paper — IQ roughly 80% heritable, environment as a “threshold variable,” class and racial variables that “cannot be accounted for in terms of environmental differences alone”¹³ — to Rushton and Jensen’s 2005 review, which argued a 50/50 genetic-environmental model across ten categories of evidence and then, in a rejoinder, moved to an 80/20 “default hypothesis.”¹⁴

The cross-context claim. The same ordering of group means, he argues, appears across very different histories and locations:

when raised under comparable conditions, you will tend to get certain patterns of differences between population representative groups of different races… the same patterns of sub-Saharan African, European, East Asian… will manifest all over the world… and that’s due to genetic differences.¹

This is the worldwide-consistency leg of Rushton and Jensen’s scorecard, and behind it sits Richard Lynn’s 2006 Race Differences in Intelligence, which reviews hundreds of studies across ten populations and explains the pattern by ice-age selection pressure — the “cold winters” theory Cofnas gestures at while flagging it himself as the speculative part: “this is the kind of thing that academics should be working on.”¹⁵ A structural fact belongs here, because it cuts both ways: much of the recent hereditarian empirical literature appears in non-mainstream venues (Lynn’s publisher is linked to the National Policy Institute), which is a real credibility cost — and the suppression argument of Cofnas’s own papers is precisely that mainstream venues declining this work is part of the mechanism, not independent evidence against it. Both halves of that sentence have to be held at once.

The numbers. He gives the standard figures:

in the United States the African-American IQ is between 82 and 85. The average white IQ is 100. The average East Asian is a few points above the white mean.¹

Reports that the gap has closed, he says, are “untrue. That’s based on achievement tests, which are not IQ tests”; the actual reasoning tests “continue to show approximately one standard deviation or 15 point gap between blacks and whites in the United States.” He puts Hispanic Americans intermediate, and the best Ashkenazi Jewish studies around 110 to 112 — explicitly distancing himself from higher figures circulating online. The numbers match the 1994 “Mainstream Science on Intelligence” statement and Rushton and Jensen.¹⁶ The claim that gap-closing evidence rests on achievement tests, though, is inaccurate as stated: the main convergence study used IQ standardization samples, not achievement tests. What those samples actually show is a live dispute, and it is in the counter-section.

The tests themselves. IQ tests, he argues, are not trivia and cannot be made culture-fair — “we know that’s impossible. It doesn’t work” — but for anyone schooled in a country like the US they measure something real, predicting outcomes from car crashes to criminality. That is consistent with the mainstream position on predictive validity: the 1994 statement’s point 5, and the APA’s 1996 task-force report.¹⁷ The strongest counter is about a different property of the tests entirely, and conflating the two is its own crux — see the measurement subsection below.

The height analogy, and a large concession. Heritability does not mean fixity:

height is a good example of a trait which is highly heritable but is still malleable¹

North Korea versus South Korea is his illustration: where nutrition is class-graded, environment explains much of the variance; where nutrition is uniform, genes explain most of what remains. And he concedes the Flynn effect outright — “our grandparents would have scored lower… Isaac Newton would have scored like average on an IQ test” — framing it as unfamiliarity with scientific thinking rather than genetic change. This is a real concession, and it matters: large, environmentally caused, population-wide IQ gains over the twentieth century are a datum, not a disputed finding. What it refutes is an inference — “high heritability means environment cannot move the trait” — and the formal model that shows exactly that is in the counter-section.

Fadeout, and the adoption study. The longest empirical passage:

this is the most studied question in the history of social science… early childhood interventions will raise IQ in the black population… significantly… there’s what’s called a fade out effect. IQ shoots up in early childhood and then… in early adulthood, mean IQ after intensive intervention reverts to the African-American mean.¹

(“Most studied question in the history of social science” is unquantified rhetoric; the fadeout claim itself is not.) He then cites “a cross-racial adoption study, the Minnesota study of twins reared apart,” in which black adoptees raised by middle-class white families had much higher childhood IQs, “and by adulthood the fade out effect occurs.” Two things need separating. The fadeout of early-intervention IQ gains is accepted by the counter-side’s own review — which adds that achievement and life-outcome effects can persist after the IQ gains fade.¹⁸ And the adoption study he is reaching for is real but misnamed: it is not the twins-reared-apart study, and the difference matters enough to get its own section below. He also argues an adoption ceiling: above a middle-class threshold, differences in parenting style or income have no “detectable effect” on adult IQ — while conceding that genuinely deprived environments may matter. The counter-side’s own number here — adoption from working-class to middle-class homes raises individual IQ by 12 to 18 points¹⁸ — is about individuals, not the group gap. Both sides need that discipline.

Regression to the mean. The claim he calls impossible to explain environmentally:

African-Americans regress to mean 85 and white Americans regress to mean 100. So even the children of high socioeconomic status black Americans tend to get lower scores on the SAT than white Americans or Asian-Americans from low socioeconomic status. That is impossible to explain on the assumption that this is some intergenerational cultural inheritance because that’s just the biological thing.¹

The descriptive pattern — gaps persisting within socioeconomic levels, and differential regression — is real and appears in the 1994 statement (point 23).¹⁶ The word doing the work is “impossible.” Whether environment covarying with social position has been adequately controlled is exactly what the two sides dispute, so the pattern is shared ground and the inference is the fight.

The polygenic concession. The host reads out an AI-generated summary: no specific genes found; polygenic scores built on mostly European samples do not reliably predict outcomes across populations. Cofnas calls the framing an attempt “to bamboozle people” — and then concedes the substance:

polygenic scores… within Europeans… often produce invalid results or at least less valid results when applied to Africans or Asians.¹

He adds that “we know that there are genetic variants that are expressed in the brain that vary between racial groups,” and when asked which ones, declines: “I can’t comment on the details of the science. That’s not my area.” The concession is load-bearing for both sides, and it is worth slowing down on. Polygenic scores are the only existing tool that could turn this debate into a direct genomic measurement. If they do not port across populations, then naive hereditarian claims built on them fail — and so do null results built on them, including the strongest published genomic counter below. Both sides need portability. Neither currently has it.

The ancient-DNA paper. His most recent citation:

just a few months ago David Rice [Reich] laboratory at Harvard published a paper in the journal Nature… showing based on an analysis of ancient DNA that there has been selection for genetic variance [variants] associated with intelligence and educational attainment in West Eurasians in the last several thousand years. So the idea that evolution stopped… 200,000 years ago… is just not consistent with mainstream science.¹

The paper is real and verified: Akbari and colleagues, from David Reich’s laboratory, in Nature, April 2026.¹⁹ What it shows and what it does not show are different enough to get their own section below — including the caveat printed in its own abstract.

What he concludes from it. So that the map records what he actually argues for, attributed and without evaluation: that under pure meritocracy, “at the very highest level at some place like Harvard it would be under 1% representation. There would still be brilliant black scholars… because there are outliers within every group”; that he is a self-described Burkean conservative whose lodestar is meritocracy, who accepts a competing value of visible representation and would accept modest quotas while rejecting DEI as an ideology of historical redress; and, as his stated thesis rather than an empirical claim: “without hereditarianism it’s impossible to have an intellectually coherent right.”

The strongest counters

Each of these is stated at its strongest, with its limits attached. That is the rubric, and it applies in both directions.

The logical wall: Lewontin to Schraiber and Edge

The oldest and most important counter is a thought experiment. Richard Lewontin’s 1970 seed-corn example: take genetically variable seed, plant one handful in rich soil and one in poor. Within each plot the height differences are entirely genetic; between the plots the difference is entirely environmental. A trait can be 100% heritable within every group while the difference between groups is 0% genetic. Within-group heritability, by itself, licenses no conclusion whatever about between-group causes.²⁰ Jensen conceded the logical point immediately and shifted to a probabilistic connection; Ned Block’s 1995 “How heritability misleads about race” became the standard statement of why that shift needs more than intuition.²¹ The strongest philosophical counter-counter is Neven Sesardic’s Making Sense of Heritability (Cambridge, 2005), which argues that the anti-heritability consensus rests on conceptual confusions and that heritability estimates can carry causal information.²² A more recent critique argues that human subpopulations are not randomly formed, so the seed-corn analogy’s premise fails empirically; it was published on a blog, not peer-reviewed, and this page marks it as such rather than silently upgrading it.

In 2024 the logical point became a theorem. Schraiber and Edge, in PNAS, prove mathematically — with simulations — that within-group heritability provides no information about between-group heritability, that the same holds for the usual workarounds (Pst, local-ancestry admixture methods), and that even a correctly estimated between-group heritability leaves the direction of the genetic and environmental components undetermined.²³ What this settles and what it does not: it closes the deductive path from “IQ is highly heritable within groups” to “the gap is partly genetic.” It does not close a probabilistic, convergence-of-evidence case — which is why the actual fight moved to the scorecards.

The scorecard fight: Rushton and Jensen versus Nisbett

Rushton and Jensen’s 2005 review is the strongest statement of the convergence method: ten categories of evidence — worldwide scores, g-loadings, heritability, brain size, transracial adoption, admixture, regression, life-history, human origins, environmental variables — scored against two models, “culture-only” (0% genetic) and “hereditarian” (initially 50/50), with the conclusion that no culture-only model fits the whole pattern.¹⁴ Nisbett’s reply, in the same journal issue, is the strongest statement of the direct-evidence method: the ten categories omit or misread the most direct tests — five admixture designs, convergence over time, intervention, adoption — and those, he argues, point one way: “the genetic contribution to the Black–White IQ gap is nil.”²⁴ This is a genuine same-question rivalry, which is rarer in this literature than it should be: two scorecards, opposite weightings of the same studies, and the dispute is precisely over which evidence is direct.

The narrowing dispute

Dickens and Flynn’s 2001 Psychological Review model is the analytical core: if IQ and environment cause each other reciprocally — a slightly better environment raising IQ, which earns a slightly better environment, multiplied across a society — then high heritability and enormous environmental effects are fully compatible, and kinship studies systematically mask how potent environment is.²⁵ This is a possibility proof, and its target is an inference (“high heritability means environment cannot produce large gaps”), not the hereditarian conclusion itself; hereditarians accept the multipliers and reply that the gap’s pattern is the evidence. Then the empirical claim: Dickens and Flynn’s 2006 analysis of IQ standardization samples found the Black–white gap had narrowed,²⁶ and the counter-side’s 2012 consensus review states the reduction as 0.33 standard deviations.¹⁸ Rushton and Jensen’s 2010 reply is that the gains are not on g — the general-intelligence core where they take the gap to live.²⁷ This is a real, instrument-specific, unresolved dispute, and it is exactly the one Cofnas’s “achievement tests” remark misdescribes: the convergence evidence comes from IQ tests. What would adjudicate it is a preregistered reanalysis with g-loadings and instrument mix fixed in advance — proposed below.

The measurement question

The 1994 statement’s point 5 says IQ tests are not culturally biased against native-born, English-speaking Americans and predict outcomes equally accurately across groups.¹⁶ Wicherts and Dolan show that the standard checks behind such claims test the wrong thing: equal factor loadings are not enough, intercept equality must be tested separately, and reanalysis of a Dutch test battery found bias against minorities that looked small under loadings-only checks and was severe once intercepts were tested; stereotype threat shows up in the data as exactly such intercept violations.²⁸²⁹ Both can be true at once, because they are different claims: predictive fairness (the test predicts the same outcomes for everyone) and measurement invariance (the test measures the same latent thing in everyone). “Tests are biased” and “tests are unbiased” in this literature routinely answer different questions, and any verdict that does not say which claim it is scoring is equivocating.

Heritability and poverty

Turkheimer and colleagues’ 2003 twin study found that the heritability of IQ in seven-year-olds varies with class: in impoverished families, roughly 60% of the variance was shared environment and genes nearly none; in affluent families, the reverse.³⁰ If environment dominates exactly where deprivation lives, then group means can be environmentally depressed while within-group heritability stays high. The limits are real: later meta-analyses find the interaction robust mainly in the United States, and the sample is children — and Cofnas’s fadeout argument is precisely that childhood is the low-heritability window. The two can both be true. Same-question discipline is what keeps them in one conversation.

The genomic frontier

The strongest published genomic test is Kevin Bird’s 2021 paper: using African and European samples from the 1000 Genomes Project, educational-attainment polygenic scores, and tests for divergent selection, it finds no signal of divergent selection once less-biased within-family effect sizes are used — the expected mean genetic difference “substantially smaller than postulated by hereditarians,” consistent with drift.³¹ Its limits are stated by its author and its critics alike: statistical power, a circularity worry about running selection tests on the very variants the scores are built from, and residual population-structure confounding, which Bird acknowledges. The published critiques with the most bite are not peer-reviewed, and this page marks them as such. And the portability problem Cofnas concedes on the podcast caps what any current polygenic-score test can show in either direction. This is the frontier: the only method that could settle the question directly, currently unable to.

What experts actually think

The best check on both “the science is settled” and “the science has discredited it” is Rindermann, Becker and Coyle’s 2020 survey of 102 intelligence researchers: on US Black–white differences, the mean attribution was 49% genetic and 51% environmental; 40% leaned environmental, 43% genetic, 17% even; 83% said at least partially genetic.³² Neither side’s headline survives those numbers. And one more finding turns the survey inward: endorsement of genetic influence correlated with the experts’ own political orientation at about r = .50. Expert opinion on this question tracks politics — which is evidence for Cofnas’s own sociology-of-science worry, applied symmetrically, to his side as much as to his opponents’. The survey was fielded in 2013–14; it is more than a decade stale.

A distinct middle position

Kathryn Paige Harden’s The Genetic Lottery (Princeton, 2021) is not a side in this fight and should not be conscripted into one: within-population genetic variation matters substantially for educational and economic outcomes; egalitarians should incorporate genetics rather than deny it; and she explicitly frames race-science readings as pseudoscience.³³ She is answering a different question — what follows politically from genetic influence on individuals within a population — and her answer is compatible with any resolution of the group-gap question. Citing her for or against the between-group claim is a category error, in whichever direction it is done.

Two citation-hygiene findings

This page’s standard is that every citation is checked against its source. Holding the podcast’s two most load-bearing citations to the same standard produces two findings. Neither is a gotcha; both change what the citations can carry.

The Minnesota study is two studies. The adoption finding Cofnas cites — black children adopted into advantaged white homes scoring well above the population average in childhood (an early-adopted mean of 110), the means falling and the group ordering persisting and widening by adolescence — is real. But it belongs to the Minnesota Transracial Adoption Study (Scarr and Weinberg 1976; adolescent follow-up Weinberg, Scarr and Waldman 1992), not to “the Minnesota study of twins reared apart,” which is Bouchard’s MISTRA, a different project with no adoption design.³⁴³⁵ The misnaming matters because the twins study cannot lend the adoption finding its authority — and because the real study is one dataset with three live readings. Scarr and Weinberg read the childhood gains as malleability. The hereditarian reading takes the adolescent follow-up, where the gaps widened, as the signature. And a 2017 reanalysis argues that attrition, age-at-adoption and Flynn-effect artifacts undermine both readings — noting, among other things, that the graph Rushton and Jensen themselves cite shows heritability flat from ages 6 to 20, which sits awkwardly with the fadeout gloss.³⁶ The limit every side acknowledges: no adoption design can fully separate genes from the society the children grow up in — Scarr herself warned this — and the decisive version (random placement, controlled pre-adoption environments, no differential attrition) is ethically impossible. The dataset is underdetermined, and it is the only one there is.

The Nature paper’s own caveat. Akbari et al. 2026 applies a new time-series method to 15,836 ancient West Eurasians and finds hundreds of alleles under strong directional selection over the past ten millennia, including shifts in allele combinations that, in today’s populations, predict cognitive performance and educational attainment.¹⁹ Two scope facts decide what it can be cited for. It measures change over time within West Eurasia — it does not compare present-day groups, and does not claim to. And its own abstract warns:

These effects were measured in industrialized societies, and it remains unclear how these relate to phenotypes that were adaptive in the past.¹⁹

Reich himself, in a 2026 podcast interview, reports that the overwhelming share of allele-frequency change in the study was drift and other factors, not directional selection. So the paper does refute “human evolution stopped 200,000 years ago” — a position no source on the other side of this map actually defends — and it does establish that recent selection on cognition-associated variants occurred within one population. What it cannot do is the work the podcast’s framing gestures toward: it is not evidence about the causes of present-day between-group differences. Cited accurately, it makes the debate more interesting. Cited past its scope, it becomes a trophy.

Where the debate actually stands

Three questions are routinely fused into one. Separated, each has a different state.

The descriptive layer is not the debate. A US Black–white test-score gap of roughly one standard deviation, high adult within-group heritability of IQ, and the predictive validity of IQ tests are affirmed by the 1994 “Mainstream Science on Intelligence” statement, by the APA’s 1996 task-force report, and by the counter-side’s own 2012 review alike.¹⁶¹⁷¹⁸ The 1994 statement — fifty-two signatories, forty-eight explicit refusals, assembled in the wake of The Bell Curve³⁷ — has been miscited by both sides ever since: it did not endorse a genetic component (its point 22 is explicitly agnostic: “There is no definitive answer to why IQ bell curves differ across racial-ethnic groups”), and it did endorse the gap’s existence and the tests’ predictive fairness. Anyone who claims the gap’s existence is contested by mainstream science is wrong. Anyone who claims its cause is settled is wrong.

The causal layer is logically settled and empirically open. Since Schraiber and Edge’s 2024 theorem, no deductive path runs from within-group heritability to between-group cause; the hereditarian case is therefore entirely a convergence-of-indirect-evidence case, and it is met by a direct-evidence scorecard (Nisbett) and a genomic null result (Bird) that is real but underpowered and portability-limited — limits Cofnas himself concedes in the podcast’s own terms. The decisive study is specifiable and has never been run: genome-wide association studies of cognitive phenotypes with large African-ancestry samples and within-family effect sizes, applied to admixed populations, with preregistered predictions. A robust directional-selection signal would support the hereditarian reading; a robust null under adequate power would support Bird’s. Neither exists at decisive power, and both the podcast’s “bamboozle” dismissal and the strongest “not supported in the least” phrasing outrun the available designs.

The free-inquiry layer is independent of both. Cofnas’s actual peer-reviewed contribution is the meta-argument: that suppressing research into possible genetic contributions is epistemically and socially harmful whatever the empirical answer turns out to be, and that this field’s history shows moral commitments driving acceptance.⁴¹⁰ The counter-arguments are normative, not empirical — the 2020 open letter’s own text argued that free inquiry “should be guided by norms of accuracy and expertise.”⁵ No genomic result settles that question, in either direction; and fusing it with the empirical claim is the category error his paper warns about — committed by critics who treated the paper as an empirical defense, and by anyone who would cite it as evidence that the gap is genetic.

The live questions

The debate’s cruxes, stated in plain language, with their current state:

What would settle it

1. Adequately powered cross-population genomics. Within-family effect sizes, large African-ancestry samples, admixed populations, preregistered predictions. This is the only design that answers the causal question directly.
2. Invariance-first gap estimation. Cross-group comparisons only on instruments passing intercept-level invariance, reported separately from predictive validity — forcing both sides onto the same question.
3. Adversarial convergence adjudication. A preregistered reanalysis of the standardization-sample narrowing, with g-loadings and instrument mix fixed in advance, run by both camps together.
4. Triangulation for adoption. The Minnesota dataset’s limits are not fixable by reanalysis; only convergence with the genomic designs in item 1 moves it.
5. For free inquiry: argument, not data. A demonstrated case where suppression produced better epistemic outcomes than inquiry, or where the predicted harms of suppression failed to materialize, would move it. Nothing else will.
6. For expert opinion: a new survey. Re-fielded, preregistered, and reported with political-composition controls rather than as a bare headcount.

This page records the strongest case and the strongest challenge currently on file, and it ends there, because that is where the evidence ends. The descriptive facts are agreed. The logic is settled. The genomics is not done. And what may be researched is a separate question from what is true — which is the one thing everyone in this debate keeps proving by forgetting it.

References

1. “The Controversial Race Scientist Who Exposed Jason Arday – Nathan Cofnas,” Triggernometry, August 2026, ~1h55m, youtube.com/watch?v=zX1q-ZOUAQY. Quotations are from YouTube’s automatic English captions; timestamps and the full provenance trail are in the underlying research file. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

2. Nathan Cofnas, nathancofnas.com/papers — publications and affiliations as listed by the author, fetched 27 August 2026. ↩

3. Nathan Cofnas, “Judaism as a Group Evolutionary Strategy: A Critical Analysis of Kevin MacDonald’s Theory,” Human Nature 29, no. 2 (2018): 134–156, doi.org/10.1007/s12110-018-9310-x. Follow-ups: “Is Kevin MacDonald’s Theory of Judaism ‘Plausible’? A Response to Dutton (2018),” Evolutionary Psychological Science 5, no. 1 (2019): 143–150; “Still No Evidence for a Jewish Group Evolutionary Strategy,” Evolutionary Psychological Science 9, no. 2 (2023): 236–259. ↩

4. Nathan Cofnas, “Research on group differences in intelligence: A defense of free inquiry,” Philosophical Psychology 33, no. 1 (2020): 125–147, published online 23 December 2019, doi.org/10.1080/09515089.2019.1697803. ↩ ↩²

5. Justin Weinberg, “Scholars Object to Publication of Paper Defending Race Science,” Daily Nous, 20 January 2020, dailynous.com. ↩ ↩²

6. The reply: Larsen, De Cruz, Kaplan, Fuentes, Marks, Pigliucci, Alfano, Smith, and Schroeder, commentary on “Research on group differences in intelligence,” Philosophical Psychology 33, no. 7 (2020): 893–898, doi.org/10.1080/09515089.2020.1805199. ↩

7. Justin Weinberg, “Controversy at Philosophical Psychology Leads to Editor’s Resignation,” Daily Nous, 24 June 2020, dailynous.com. ↩

8. Mitchell Herschbach, “Editor’s views,” Philosophical Psychology 33, no. 7 (2020): 899–901, doi.org/10.1080/09515089.2020.1805198. Cofnas’s characterization: “Philosophy Is Being Hijacked by Woke Twitter Mobs,” Quillette, 21 October 2020. ↩

9. Lisa Bortolotti and Katrina Sifferd, “Journal News: Philosophical Psychology and Neuroethics (guest post),” Daily Nous, 15 October 2021, dailynous.com. ↩

10. Nathan Cofnas, “Science Is Not Always ‘Self-Correcting’: Fact–Value Conflation and the Study of Intelligence,” Foundations of Science 21, no. 3 (2016): 477–492, doi.org/10.1007/s10699-015-9421-3. ↩ ↩²

11. Justin Weinberg, “Emmanuel College Dissociates from Cofnas (updated),” Daily Nous, 22 April 2024, dailynous.com — note the triple-s typo in the original URL slug. ↩

12. University of Cambridge, “Statement on Disciplinary Investigation (Dr Cofnas),” cam.ac.uk. ↩

13. Arthur R. Jensen, “How much can we boost IQ and scholastic achievement?”, Harvard Educational Review 39, no. 1 (1969): 1–123, doi.org/10.17763/haer.39.1.l3u15956627424k7. ↩

14. J. Philippe Rushton and Arthur R. Jensen, “Thirty years of research on race differences in cognitive ability,” Psychology, Public Policy, and Law 11, no. 2 (2005): 235–294, doi.org/10.1037/1076-8971.11.2.235; the rejoinder in the same issue is “Wanted: More Race Realism, Less Moralistic Fallacy.” ↩ ↩²

15. Richard Lynn, Race Differences in Intelligence: An Evolutionary Analysis (Washington Summit Publishers, 2006), bibliographic record. ↩

16. Linda S. Gottfredson, “Mainstream Science on Intelligence,” Wall Street Journal, 13 December 1994, p. A18; reprinted with 52 signatories in Intelligence 24, no. 1 (1997): 13–23, doi.org/10.1016/S0160-2896(97)90011-8. ↩ ↩² ↩³ ↩⁴

17. Ulric Neisser et al., “Intelligence: Knowns and unknowns,” American Psychologist 51, no. 2 (1996): 77–101, doi.org/10.1037/0003-066X.51.2.77. ↩ ↩²

18. Richard E. Nisbett, Joshua Aronson, Clancy Blair, William Dickens, James Flynn, Diane F. Halpern, and Eric Turkheimer, “Intelligence: New findings and theoretical developments,” American Psychologist 67, no. 2 (2012): 130–159, doi.org/10.1037/a0026699. ↩ ↩² ↩³ ↩⁴

19. Ali Akbari et al., including David Reich, “Ancient DNA reveals pervasive directional selection across West Eurasia,” Nature (published online 15 April 2026), doi.org/10.1038/s41586-026-10358-1. ↩ ↩² ↩³

20. Richard C. Lewontin, “Race and intelligence,” Bulletin of the Atomic Scientists 26, no. 3 (1970): 2–8, doi.org/10.1080/00963402.1970.11457774. ↩

21. Ned Block, “How heritability misleads about race,” Cognition 56, no. 2 (1995): 99–128, doi.org/10.1016/0010-0277(95)00678-R. ↩

22. Neven Sesardic, Making Sense of Heritability (Cambridge University Press, 2005), doi.org/10.1017/CBO9780511487378. ↩

23. Joshua G. Schraiber and Michael D. Edge, “Heritability within groups is uninformative about differences among groups: Cases from behavioral, evolutionary, and statistical genetics,” Proceedings of the National Academy of Sciences 121, no. 12 (2024), doi.org/10.1073/pnas.2319496121. ↩

24. Richard E. Nisbett, “Heredity, environment, and race differences in IQ: A commentary on Rushton and Jensen (2005),” Psychology, Public Policy, and Law 11, no. 2 (2005): 302–310, doi.org/10.1037/1076-8971.11.2.302. ↩

25. William T. Dickens and James R. Flynn, “Heritability estimates versus large environmental effects: The IQ paradox resolved,” Psychological Review 108, no. 2 (2001): 346–369, doi.org/10.1037/0033-295X.108.2.346. ↩

26. William T. Dickens and James R. Flynn, “Black Americans reduce the racial IQ gap: Evidence from standardization samples,” Psychological Science 17, no. 10 (2006): 913–920, doi.org/10.1111/j.1467-9280.2006.01802.x. ↩

27. J. Philippe Rushton and Arthur R. Jensen, “The rise and fall of the Flynn Effect as a reason to expect a narrowing of the Black–White IQ gap,” Intelligence 38, no. 2 (2010): 213–219, doi.org/10.1016/j.intell.2009.12.002. ↩

28. Jelte M. Wicherts and Conor V. Dolan, “Measurement invariance in confirmatory factor analysis: An illustration using IQ test performance of minorities,” Educational Measurement: Issues and Practice 29, no. 3 (2010): 39–47, doi.org/10.1111/j.1745-3992.2010.00182.x. ↩

29. Jelte M. Wicherts, Conor V. Dolan, and David J. Hessen, “Stereotype threat and group differences in test performance: A question of measurement invariance,” Journal of Personality and Social Psychology 89, no. 5 (2005): 696–716, doi.org/10.1037/0022-3514.89.5.696. ↩

30. Eric Turkheimer, Andreana Haley, Mary Waldron, Brian D’Onofrio, and Irving I. Gottesman, “Socioeconomic status modifies heritability of IQ in young children,” Psychological Science 14, no. 6 (2003): 623–628, doi.org/10.1046/j.0956-7976.2003.psci_1475.x. ↩

31. Kevin A. Bird, “No support for the hereditarian hypothesis of the Black–White achievement gap using polygenic scores and tests for divergent selection,” American Journal of Physical Anthropology 175, no. 2 (2021): 465–476, doi.org/10.1002/ajpa.24216. ↩

32. Heiner Rindermann, David Becker, and Thomas R. Coyle, “Survey of expert opinion on intelligence: Intelligence research, experts’ background, controversial issues, and the media,” Intelligence 78 (2020): 101406, doi.org/10.1016/j.intell.2019.101406. ↩

33. Kathryn Paige Harden, The Genetic Lottery: Why DNA Matters for Social Equality (Princeton University Press, 2021), press.princeton.edu. ↩

34. Sandra Scarr and Richard A. Weinberg, “IQ test performance of black children adopted by white families,” American Psychologist 31, no. 10 (1976): 726–739, doi.org/10.1037/0003-066X.31.10.726. ↩

35. Richard A. Weinberg, Sandra Scarr, and Irwin D. Waldman, “The Minnesota Transracial Adoption Study: A follow-up of IQ test performance at adolescence,” Intelligence 16, no. 1 (1992): 117–135, doi.org/10.1016/0160-2896(92)90028-P. ↩

36. Drew Thomas, “Racial IQ Differences among Transracial Adoptees: Fact or Artifact?”, Journal of Intelligence 5, no. 1 (2017): 1, doi.org/10.3390/jintelligence5010001. ↩

37. Richard J. Herrnstein and Charles Murray, The Bell Curve: Intelligence and Class Structure in American Life (New York: Free Press, 1994), bibliographic record. ↩