On the MCM6 Milk Consumption Switch & Lactase Persistence; & the Nonexistence of Any Such Switches in IQ Genetics "Race Science"
This is too close to the “let’s get out the skull-measuring calipers” discourse. I shouldn’t have to do this. Any genetic variants that made us genuinely smarter at dealing with situations without incurring a strong energy-use penalty would have rapidly made a “hard sweep” through the East African Plains Ape population long ago. We see this in cultures with domesticated potential-dairy animals. Strong selection pressures—2% to 5% per generation—has created large groups of humans who have DNA variants that flip switches so they can comfortably drink milk as adults; but there are no counterpart genetics-embedded switch-flips we have found that make people higher-IQ…
We know exactly how adults keep digesting milk: enhancer tweaks at MCM6 turn up LCT. Decades of promises about IQ genetics have delivered correlations, not causes. IQ inheritance remains a statistical fog—limited heritability, many environmental and nurture cofounders, really tiny identifiable mechanisms. Hype, excuses, and vanishing variance explained.
Let us start with milk:
We understand the genetic basis of lactose persistence—LP, the ability of adults and not just the very young to digest the sugar lactose, and hence milk, comfortably—in the East Afrfican Plains Ape. The phenotype is typically defined by sustained intestinal lactase-phlorizin hydrolase (LPH; gene LCT) activity beyond weaning.
Genetically, LP primarily associates with variants in the regulatory-enhancer DNA region upstream of LCT within the adjacent gene MCM6. Multiple, geographically structured alleles—including the canonical rs4988235 DNA single-nucleotide polymorphism (SNP) characteristic of northern and western Europeans, as well as distinct variants common in East African and Middle Eastern pastoralist populations—demonstrate convergent evolution during the Holocene. These variants exhibit some of the strongest signals of recent positive selection in the human genome, with estimated selection coefficients often between ~0.02 and ~0.10, depending on ecological context and modeling assumptions.
Ancient DNA reveals a striking lag between the arrival of domesticated dairy-producing animals in human settlements and the spread of LP alleles, implying gene–culture coevolution with temporally variable fitness advantages, mediated by pathogen burden, seasonality, famine risk, and life-history trade-offs.
At the level of molecular biology and regulatory architecture at the LCT/MCM6 locus, we understand LP thus: In the typical mammal, lactase-phlorizin hydrolase (LPH; gene LCT) is expressed in the brush-border membrane of small-intestinal enterocytes. In most mammals, LCT transcription decreases after weaning via developmental epigenetic programming (notably changes in histone marks and DNA methylation in LCT regulatory regions), resulting in lactose malabsorption. But LP is the maintenance of high LCT expression into adulthood.
Note that genotype-to-phenotype mapping is high effect but context-dependent. The gastro-intestinal tract is a complicated system. Adults with the genes for LP can still get sick from drinking milk. Adults without the genes for LP can—sometimes, under some circumstances—tolerate drinking milk without getting sick.
The LCT gene resides on chromosome 2q21, adjacent to MCM6. Adult expression of LCT to produce LP hinges on enhancers embedded within MCM6 introns. The canonical European variant is rs4988235. 13910 DNA bases upstream of the start of the LCT gene, what is normally cytosine base has been replaced by a thymine base. This enhances transcription factor binding to increase expression of the LCT gene in adults.
Epigenetically, LP alleles correlate with reduced methylation in enhancer regions in adults, consistent with active chromatin states. Importantly, cis-regulatory control explains the large allele-specific effect sizes without coding changes in LCT. The underlying gene making LPH never changes. What changes is the degree to which MCM6 allows it to keep making proteins after weaning.
In addition, in many Europeans, there is also rs182549: 22018 bases before the start of LCT, a guanine has been replaced by an adenine. This single-nucleotide polymorphism (SNP) is less functionally effective an enabling LP, but rs182549 and rs4988235 tend to occur together.
In East African and Middle Eastern populations, there are other sources of LP. We have:
14010 bases upstream a guanine is replaced by a cytosine (rs145946881)
13915 bases upstream a thymine is replaced by a guanine (rs41380347)
13907 bases upstream a cytosine is replaced by a guanine (rs41525747)
13913 bases upstream a thymine is replaced by a cytosine (rs429699).
At the level of population-genetic signatures and evolutionary dynamics: This is convergent evolution in the East African Plains Ape: once you acquire domesticated animals that can potentially be used for dairy, your chances of survival as an adult if you can comfortably drink milk go up. The amount of going-up is substantial across the three hundred generations of human evolution since the beginnings of non-dog animal domestication. Hence at least some degree of LP has evolved and been selected for at least five different times in human populations.
Estimates of the positive selection pressure for those with the European rs4988235 variant often fall within the range of an increase of between 2% to 5% of an individual’s chance of passing on his or her genes, both because milk provides calories and because milk is sometimes much safer when the water carries pathogens. The East African pastoralist variants show similarly strong or stronger signals, with rapid rises over the past 3–5 millennia consistent with cultural adoption of pastoralism.
At the level of the archaeological and demographic context inferred from aDNA and ethnographic data on the spread of domesticated dairy-producing animals: It looks as though, in European populations, variant rs4988235 mutated and evolved and then swept through the population, hard and fast, millennia after domestication, for it looks as though there was a substantial time lag between the taming of potential dairy animals and LP: first technologies of fermentation and cheese-making were used to purge the lactose from the dairy food, and only later did rs4988235 sweep through populations late in the Bronze and early in the Iron Age. In Africa, the expansion of LP aligns with pastoralist expansion across the Great Rift Valley and into the Sahel.
Even today, LP exhibits a very strong geographically and ancestrally derived pattern across the population of the East African Plains Ape:
70% or so in northern and western Europeans and their diasporas (e.g., rs4988235).
20%-50% in certain Middle Eastern and South Asian groups; significant heterogeneity reflects varied dairying traditions and admixture.
Multiple African pastoralist societies show high LP frequencies via non-European alleles; for example, the 14010 base upstream replacement of guanine by cytosine (rs145946881) attains high frequencies among certain East African groups (e.g., Maasai, Beja), while other pastoralists possess different enhancer alleles.
5% or in East Asians and many Indigenous American groups, consistent with late or limited availability of potential dairy animals, and also reliance on technology: lactose processing via fermentation rather than fresh milk consumption.
But that was not the point of this post. The point of this post is that we understand the genetic influences on LP, but we do not understand the genetic influences on IQ.
And yet Charles Murray claims that we do.
Now comes to my feed Eric Turkheimer:
Eric Turkheimer: Your Genes Are Simply Not Enough to Explain How Smart You Are <https://www.theatlantic.com/science/2025/10/genetics-intelligence-charles-murray/684544/>: ‘Seven years ago, I took a bet from one of the most controversial figures in the scientific world. Charles Murray, the political scientist who—along with the late psychologist Richard Herrnstein—wrote The Bell Curve in 1994, wagered that one of his core ideas about genetics and intelligence would be proved true by 2025. He emailed me some time after I’d helped stoke an online furor about his insistent defense of The Bell Curve’s main points, which he’d recently reiterated on a popular podcast and which I, along with two other psychologists and intelligence researchers, had denounced in Vox. I took the bet because I was confident I would win….
The bet’s premise was simple enough. Murray quoted himself on the podcast, arguing that “we will understand IQ genetically. I think most of the picture will have been filled in by 2025—there will still be blanks, but we’ll know basically what’s going on.” And he proposed that, in seven years, he’d sit through a lecture I gave on the topic: “Who Was More Right?” It is now 2025, and I am here to declare that I was more right…
But Charles Murray disagrees:
Murray… insisted that he is still right. “Of course I think I won the bet, and I will lay out my reasons for thinking that,” he wrote in an email. He told me he plans to do so in a few months, when he has more time…. Murray said that any argument I’d lay out here [in the Atlantic Monthly] would be insufficient to make this case. “We’re not talking about something that lends itself to 1,500 words, or 3,000 for that matter, written for a lay audience, but a 20-page (at least) technical presentation,” he wrote. (He did not respond to an invitation to elaborate on this, or respond to the argument made in this piece….
The terms of the bet… took an unanswerable question about what might be discovered on some indefinite timescale and made it real. The world would be a different place if scientists understood intelligence at the genetic level. But right now, like wealth and health, IQ remains a node in the uncontrolled matrix of human development, causing some things and being caused by others, as genes and environment interact in the background…
From my perspective, Murray’s claim that this is not “something that lends itself to 1,500 words, or 3,000 for that matter, written for a lay audience” is pure bullshit.
Lactose persistence—the ability of adults to drink milk without getting sick—in European and European-diaspora populations is overwhelmingly the result of the replacement of a cytosine base by a thymine base 13910 DNA bases upstream of the start of the LCT gene. As a result, the MCM6 gene does not flip the switch that turns off the LCT gene after weaning, and prevents it from producing the intestinal lactase-phlorizin hydrolase enzyme needed to comfortably digest lactose.
That took 80 words. Suppose that IQ differences in the East Afericn Plains Ape population were to be strongly genetically determined by twenty different SNP mutations. With 80 words to describe each one and its effects, we are well under 3000 words. The Atlantic Monthly cannot publish the technical bioneurochemistry, but it can summarize what the technical bioneurochemistry means quite well.
Before he settled on the “I really won, and I will explain, but not now” explanation, Murray had a different line. He agreed that he had lost his bet with Turkheimer, because:
Charles Murray: <https://twitter.com/charlesmurray/status/1874111749299056996>: ‘I failed to take into account that people trying to develop such explanations would be denied access to the databases that would permit such explanations. Silly me…
To which Sasha Gusev has a full rebuttal:
Sasha Gusev: <https://x.com/SashaGusevPosts/status/1977927672371069023>: ‘Murray being so wrong in his prediction should make us question his world model, but it’s also worth commenting on his response. Murray has, for some time now, been workshopping the excuse that progress on IQ genetics was blocked by researchers being denied the access to the relevant databases. This is patently untrue! First, one of the largest genetic analyses to date of *any* trait is of educational attainment, a phenotype Murray himself has used as a proxy for intelligence. Surely a study of 3 million should have been enough to satisfy Murray’s prediction.
Second, a massive (n=~500k) rare variant study of cognitive function (including a short IQ test) was published in 2023 in one of the most prestigious journals in the field. That study identified a mere eight genes, and the overall variance explained was just 0.15% (that’s right). Finally, a recent pre-print imputed higher quality IQ into the full >450k UK Biobank population. Putting *that* data together with the largest available studies of IQ produced a common variant genetic score that explained... 2.6% of the trait.
In short, the genetic mechanisms of IQ have been extraordinarily well studied both from the rare and common variant perspective and now explain... essentially none of it. Turkheimer clearly wins the bet, and makes a compelling case that the whole premise was flawed:
So what’s Charles Murray up to now? He’s been laundering fake analyses that he himself doesn’t understand from anonymous race bloggers who can’t be bothered to write them up (FWIW “substantive reply” did not come). Not with a bang but with a whimper…
“Question his world-model…”—now that is excessively polite.
The basic genetic basis of lactase persistence can be told in semi-plain English: within European and European-diaspora populations, the SNP at −13910 C>T near LCT boosts adult lactase digestion via the LPH enzyme, with parallel variants across pastoralist ecologies. Ancient DNA shows a lag, then rapid sweeps; selection coefficients are quantifiable.
Europe’s rs4988235 SNP DNA variant is a textbook “hard sweep;” East Africa’s LP variants are a “soft sweep “mosaic—both traceable, testable, and meaningful. These tiny enhancer edits built real fitness in tough environments where milk was calories and clean hydration.
Meanwhile, the grand project to decode IQ at the genetic level stalled: big samples, tiny effects, higher heritability on paper than in DNA, and mechanisms missing in action. Common- and rare-variant studies explain a sliver of variance, collapse under sibling controls, and fail to deliver biological pathways.
The contrast is the point.
Gene–culture stories can be precise and cumulative when biology cooperates; with intelligence, precision has not arrived. That gap makes a mockery of premature certainty.