The Question That Clears a Room

The degree I didn’t have

The civilian police turned me down because I didn’t have a degree. Not because of anything I’d done, and not because of anything I couldn’t do, but because a certificate I didn’t hold was being used as a stand-in for a whole bundle of qualities that nobody in the recruitment office could be bothered to measure in me directly. It was a proxy, and proxies are wonderfully efficient right up until you happen to be the person they get wrong. The RAF Police took me at 26, a decade older than most of the lads in my intake, and I spent the next seven-odd years quietly proving that the proxy had been, shall we say, a little hasty.

I mention it because it’s the closest I’ve come to being on the wrong end of a statistic, and because it’s the lens I can’t help bringing to the question I want to look at today. Years of police work and senior security management taught me that institutions deal with uncomfortable questions in one of two ways. They either measure the wrong thing very confidently, or they refuse to measure anything at all and hope the question gets bored and wanders off. Neither approach has ever caught a thief, and I’m not convinced either has ever found a truth.

So here’s the question, and I’ll warn you now that it’s the one most likely to empty a dinner party faster than a fire alarm. Are there average differences in measured cognitive ability between groups of people of different ancestry, and if there are, how much of that is down to genes and how much to everything else? I’m a retired copper writing from a cave in rural Spain, not a geneticist, and I have no horse in this race other than a stubborn dislike of being told which questions are allowed. What follows is me trying to do what I’d have done with any contested case file… read the evidence on both sides, note where it’s solid and where it’s been padded, and resist the very human urge to decide the verdict before the witnesses have spoken.

The most discouraged sentence in psychology

In 2021 a team of researchers did something rather brave, which was to ask American psychology professors, anonymously, what they actually believed. They interviewed 41 academics to find the ideas people were nervous about, then surveyed 470 psychology professors at 133 universities about ten “taboo conclusions” their own peers had nominated. Of those ten, the one the professors were most inclined to discourage colleagues from even studying was this: that genetic differences explain a non-trivial share, 10% or more, of the variance in race differences on intelligence tests. Not a claim that genes explain everything, mind. Ten per cent. That was the sentence that made the room go quiet.

What I found more interesting than the ranking was what sat underneath it. For every one of the ten conclusions, some professors were 100% certain it was true, and others were 100% certain it was false, which tells you this isn’t a settled field with a few cranks on the fringe; it’s a genuine fault line. The professors most confident that a taboo conclusion was true also reported self-censoring the most, and tenured staff reported as much fear of consequences as those still waiting for job security, including fear of being sacked. Just over half, 52.3%, agreed that scholars should be completely free to pursue research without fear of institutional punishment for their conclusions. Which means the other half didn’t, at least not completely.

As an old policeman, that self-censorship pattern sets off every alarm I’ve got. If the people who believe one thing keep their mouths shut and the people who believe the opposite speak freely, then what looks like a scientific consensus may partly be a consensus of who felt safe talking. The survey’s own authors flagged exactly this risk. It doesn’t mean the quiet ones are right. It means you can’t tell from the noise alone, and any investigator who has ever taken a statement in a room full of people who all knew the boss was listening will recognise the problem.

The heat hasn’t cooled since. This year the Journal of Controversial Ideas, a publication set up by philosophers Jeff McMahan, Francesca Minerva and Peter Singer partly so that authors can publish under a pseudonym, ran an issue on the subject. Its editorial describes how 45 philosophers signed a letter opposing the hiring of Nathan Cofnas at Ghent, calling research on group IQ differences pseudoscience and “rubbish”. One contributor, Tezuka, argued that banning the research could itself harm minorities, while the physicist Alan Sokal asked the sharper question… whether the dismissal rests on evidence or simply on sociology. That, for me, is the right question to carry into the evidence. Not “is this comfortable?” but “is this true, and how would we know?”

What the file actually says

Let’s start with the bits that aren’t seriously in dispute, because there are more of them than either side’s loudest voices like to admit. In the United States, average scores on general cognitive tests have shown a gap between Black and White Americans for as long as anyone has been testing, historically around 15 points, or one standard deviation. It shows up early in childhood; it is reduced but not erased when researchers control for family income and parental education, and it predicts group-level differences in schooling and earnings in the same direction that individual scores predict individual outcomes. Gaps between other groups exist too, smaller, larger, or pointing the other way, depending on who you compare. Pretending none of this exists isn’t kindness; it’s just bad record-keeping.

The second solid bit is that, within a population, intelligence is substantially heritable. Twin and adoption studies routinely put adult heritability somewhere between 50% and 80%, and even the researchers most associated with the environmental side of the argument accept it. William Dickens, of whom I’ll speak shortly, put it bluntly when he said the evidence that intelligence is partly inherited within races is rock solid. The molecular work keeps adding detail. Just this month, researchers at Amsterdam UMC and the Wellcome Sanger Institute published in Nature Genetics the largest genetic study of intelligence yet. By statistically estimating scores for the roughly 40% of UK Biobank volunteers who never sat the test, they lifted their sample from about 270,000 people to more than 455,000, found around 40% more of the tiny common DNA variations linked to test scores, and identified 26 genes where rare variants were linked to lower scores, up from five.

Here’s the detail that tends to fall off the back of the lorry when that study gets shared on social media. Every one of those 455,000 people was of European ancestry. It is a study of why individuals within one broad population differ from each other, and it says precisely nothing, by design, about why averages might differ between populations. Fourteen of those 26 genes were already known to cause neurodevelopmental conditions, which is valuable medical knowledge and about as far from a racial league table as you can get. The researchers themselves framed it as helping to understand neurodevelopment and social inequality. If you’re reading it as ammunition for a group-differences argument, you’ve walked into the wrong courtroom.

And that is the hinge on which the whole debate swings. Everybody serious agrees the gaps in test scores exist. Everybody serious agrees genes influence individual differences. The genuinely open question, the one in that taboo survey, is whether genes explain any meaningful part of the gap between groups. Hereditarians, from Arthur Jensen onwards, argue that convergent lines of evidence make some genetic contribution likely. Environmentalists argue that history, poverty, discrimination, schooling, health and the tests themselves can account for it. To answer that you need a different kind of evidence altogether, and it’s when we go looking for it that things get properly interesting.

Where the case gets padded

The oldest rebuttal is still the best one, and it involves seeds. Take a sack of genetically varied seed, split it in two, and plant one half in rich soil and the other in poor. Within each tray, differences in height will be largely genetic, because every plant got the same treatment. Between the trays, the difference is entirely down to the soil. High heritability within groups tells you nothing, on its own, about the cause of a gap between them. Hereditarian writers know this perfectly well and argue they’re relying on other evidence instead, which is fair, but you would be amazed how often the within-group figure gets waved around online as if it settled the matter. It doesn’t. It can’t.

Then there’s the Flynn effect, which still feels like a magic trick every time I think about it. Across the twentieth century, raw scores on IQ tests rose so steeply that the average person in many countries today would score far above their grandparents’ average on the old norms, a shift bigger than the Black-White gap itself, in a timescale far too short for genes to have changed. Whatever caused it, nutrition, schooling, smaller families, a world that rewards abstract thinking, it proves that environments can move whole populations by more than the gap we’re arguing about. Hereditarians reply that those gains may not be on the same underlying ability the gap is measured on, and that’s a legitimate technical point. But it does make “the environment can’t possibly explain 15 points” a very hard sentence to say with a straight face.

The gap itself hasn’t stood still either. In 2006 William Dickens and James Flynn analysed 30 years of national standardisation data and found that Black Americans had gained about 5.4 points on White Americans between 1972 and 2002, roughly a third of the traditional gap. On one set of children’s tests the averages had come in to 92.1 and 103. Philippe Rushton and Arthur Jensen accused them of picking favourable tests, and Charles Murray published data suggesting the narrowing had stalled by the late 1980s. So the honest summary is “it narrowed, then the experts argued about how much and for how long”, which is less satisfying than either side’s slogan but a lot closer to the truth.

This next bit lands close to home for me, because I spent part of my childhood as a service kid in Germany. After the Second World War, a researcher called Klaus Eyferth tested German children fathered by Black and White occupation soldiers, all raised by white German mothers in the same post-war society, and found no meaningful difference in their average scores. Flynn cited it as one reason he thought the gap was more environmental than genetic. It has its critics, who point out that the soldiers were screened by army tests and the children were mostly tested young. I don’t think it proves anything on its own. But I’d have given a lot, in my RAF Police days, for a natural experiment that clean in any of my cases.

The newest weapon in the hereditarian armoury is the polygenic score, which adds up thousands of tiny genetic effects into a single number per person. Some researchers have pointed out that average scores differ between ancestry groups in the same direction as average test results. The problem is the ruler. These scores are built almost entirely from people of European descent, and they travel badly. In one of the biggest education studies, the score explained 10.6% of the variation in educational attainment among European Americans but only 1.6% among African Americans, and a study of 245 traits found prediction drops off even between European populations, falling further the more genetically distant a group is from the one the score was trained on. Comparing group averages with a ruler calibrated on only one of the groups isn’t evidence. It’s a measurement artefact wearing a lab coat.

And underneath all of it sits the slippery word “race” itself. The anthropologist John Relethford described it as “a culturally constructed label that crudely and imprecisely describes real variation”, which is about the most even-handed sentence anyone has managed on the subject. Ancestry is real and measurable. The boxes we tick on a census form are a much rougher thing, shaped by history, law and who was doing the counting. When you hear a confident claim about “races”, it’s always worth asking which boxes, drawn by whom, and why there.

None of this means the environmentalists get a free pass. The gap’s persistence after controlling for income is genuinely awkward for the simplest poverty explanations, and “it’s all discrimination” is a hypothesis too, one that also needs testing rather than reciting. Where I come out, having read the file, is this. A genetic contribution to the gap between groups has not been demonstrated. Nor has it been ruled out to everyone’s satisfaction. The direct evidence that would settle it doesn’t yet exist, and the indirect evidence is a good deal weaker than its most enthusiastic promoters claim.

A security failure, before it’s anything else

In January, The New York Times published an investigation by Mike McIntire that I read twice, once as a writer and once with my old security manager’s hat firmly back on. It concerned the Adolescent Brain Cognitive Development Study, a federally funded American project that began in 2015 and put more than 11,000 children through regular brain scans and clinical tests and took their DNA. Families were promised the data would be closely guarded. The promotional material even featured a cartoon of a Black child saying it felt good to know scientists were keeping their information safe. The scientists did not keep it safe.

According to the Times, a group of fringe researchers who were not eligible for access got round the National Institutes of Health’s safeguards anyway, and used the children’s data to produce at least 16 papers ranking ethnic groups by IQ and suggesting Black people earn less because they’re less intelligent. Mainstream geneticists rejected the work as biased and unscientific. One social media post built on the data was viewed around 670,000 times. The NIH had approved more than 92,000 access requests since 2007 and admitted it couldn’t properly monitor every download, and the Times also found the same data had been improperly obtained in 2024 by an unidentified researcher in China.

I spent years writing memos that required signed acknowledgement, precisely because I’d learned that “we told people the rules” is not the same as “we enforced the rules”. This is that lesson writ large, with children’s genomes instead of a stock room. Whatever you believe about group differences, a promise made to parents was broken, and that is a breach of trust before it is a scientific argument. The defenders of open data have a point that restricting which conclusions researchers are allowed to reach is corrosive to science. But nobody’s academic freedom extends to helping themselves to data they weren’t entitled to.

And here’s the vicious little loop I can’t stop seeing. Every time something like this happens, it hands the people who want the whole subject closed down a perfect exhibit for the prosecution. And every time the subject is closed down, careful mainstream researchers walk away from it, leaving the field to exactly the sort of people who don’t fill in the access forms. Suppression feeds misuse, misuse justifies suppression, and round it goes. I watched the same dynamic in security work more than once… ban a thing outright, and you don’t get rid of it, you just lose sight of who’s doing it.

The person standing in front of you

Which brings me back to that missing degree. A group average is a description of a crowd. It tells you nothing whatsoever about the particular individual standing in front of you, because the spread of ability within any group is vastly wider than the difference between group averages. Pick any two populations you like and the overlap is enormous. Every competent copper learns this early, usually the hard way, because the moment you start policing the profile instead of the person, you start arresting the wrong people and missing the right ones. A proxy is a lazy officer’s best friend and an honest officer’s worst habit.

So why does the between-group question matter at all, if it can’t tell you anything about individuals? Because policy is made about crowds. And here both camps have an argument that deserves a fair hearing, even if you end up disagreeing with it. Those on the hereditarian side argue that if you assume every group has identical average ability, then every gap in outcomes must be evidence of discrimination, which leads to policies that misdiagnose the problem, blame the wrong institutions and fail the very people they’re meant to help. Tezuka’s argument in the Journal of Controversial Ideas runs along these lines… that refusing to look can harm minorities rather than protect them.

Those on the environmental side argue the opposite danger is far older and has a body count to prove it. Ideas about innate group inferiority have been used to justify segregation, forced sterilisation and immigration bans, and the moment people believe a gap is “natural”, the political will to fix schools, housing and healthcare quietly evaporates. Why fund what you’ve decided can’t be changed? Given that the Flynn effect and the narrowing of the gap both show scores responding to circumstance, they argue, fatalism isn’t just cruel; it’s empirically wrong.

My own view, for what it’s worth from a man who lives in a hillside cave, is that the practical conclusions barely change whichever way the science eventually falls. Good nutrition, decent schools, safe streets and fair treatment raise the floor for everyone, and nobody has ever produced evidence that they stop working on certain children. Judging people as individuals rather than as representatives of their category is right whether group averages differ for reasons of genes, history or both. Where the answer does matter is in diagnosis, in working out why a gap persists so we stop throwing money at the wrong cause. And you can’t diagnose anything if half the doctors are too frightened to examine the patient.

Notes from the cave

Out here in rural Spain, life has a way of stripping arguments down to their working parts. The well either has water in it or it doesn’t, and no amount of shouting at it changes the level. I find myself wishing we could treat this subject the same way, as a question with an answer we don’t yet fully have, rather than a loyalty test where your opinion is a badge you wear to show which tribe you belong to. The people who are certain genes explain the gap have run well ahead of their evidence. The people who are certain the very question is evil have decided the verdict before the witnesses have been called. Both, in their own way, have stopped investigating.

What bothers me most isn’t either answer. It’s the silence in the middle, the professors who believe something and won’t say it, the professors who doubt something and won’t say that either, and the cranks who are only too happy to fill the gap with charts built on data they nicked from children. I joined the RAF Police because I wanted to find things out properly, and I left the civilian security world known as Mr Memo because I believed that what isn’t written down can’t be held to account. Whatever the truth turns out to be here, it won’t be found by people who are afraid to write it down.

I don’t know how much of that gap, if any, lives in our genes. I’m not sure anyone does yet, whatever they tell you on the internet. What I do know is that the civilian police once looked at a statistic about people like me and decided they already knew the answer, and they were wrong. Perhaps that’s the only lesson I’m really qualified to offer. Averages are real, but they’ve never once walked through a door. People do, one at a time, and they deserve to be met that way… even while the rest of us keep arguing about the crowd.