by David Reich
To understand why it is no longer an option for geneticists to lock arms with anthropologists and imply that any differences among human populations are so modest that they can be ignored, go no further than the “genome bloggers.” Since the genome revolution began, the Internet has been alive with discussion of the papers written about human variation, and some genome bloggers have even become skilled analysts of publicly available data. Compared to most academics, the politics of genome bloggers tend to the right—Razib Khan17 and Dienekes Pontikos18 post on findings of average differences across populations in traits including physical appearance and athletic ability. The Eurogenes blog spills over with sometimes as many as one thousand comments in response to postings on the charged topic of which ancient peoples spread Indo-European languages,19 a highly sensitive issue since as discussed in part II, narratives about the expansion of Indo-European speakers have been used as a basis for building national myths,20 and sometimes have been abused as happened in Nazi Germany.21 The genome bloggers’ political beliefs are fueled partly by the view that when it comes to discussion about biological differences across populations, the academics are not honoring the spirit of scientific truth-seeking. The genome bloggers take pleasure in pointing out contradictions between the politically correct messages academics often give about the indistinguishability of traits across populations and their papers showing that this is not the way the science is heading.
What real differences do we know about? We cannot deny the existence of substantial average genetic differences across populations, not just in traits such as skin color, but also in bodily dimensions, the ability to efficiently digest starch or milk sugar, the ability to breathe easily at high altitudes, and susceptibility to particular diseases. These differences are just the beginning. I expect that the reason we don’t know about a much larger number of differences among human populations is that studies with adequate statistical power to detect them have not yet been carried out. For the great majority of traits, there is, as Lewontin said, much more variation within populations than across populations. This means that individuals with extreme high or low values of the great majority of traits can occur in any population. But it does not preclude the existence of subtler, average differences in traits across populations.
The indefensibility of the orthodoxy is obvious at almost every turn. In 2016, I attended a lecture on race and genetics by the biologist Joseph L. Graves Jr. at the Peabody Museum of Archaeology and Ethnography at Harvard. At one point, Graves compared the approximately five mutations known to have large effects on skin pigmentation and that are obviously different in frequency across populations to the more than ten thousand genes known to be active in human brains. He argued that in contrast to pigmentation genes, the patterns at genes particularly active in the brain would surely average out over so many locations, with some mutations nudging cognitive and behavioral traits in one direction and some pushing in the other direction. But this argument doesn’t work, because in fact, if natural selection has exerted different pressures on two populations since they separated, traits influenced by many mutations are just as capable of achieving large average differences across populations as traits influenced by few mutations. And indeed, it is already known that traits shaped by many mutations (as is probably the case for behavior and cognition) are at least as important targets of natural selection as traits like skin color that are driven by a small number of mutations.22 The best example we currently have of a trait governed by many mutations is height. Studies in hundreds of thousands of people have shown that height is determined by thousands of variable positions across the genome. A 2012 analysis led by Joel Hirschhorn showed that natural selection on these is responsible for the shorter average height in southern Europeans compared to northern Europeans.23 Height isn’t the only example. Jonathan Pritchard led a study showing that in the last approximately two thousand years there has been selection for genetic variations that affect many other traits in Britain, including an increase in average infant head size and an increase in average female hip size (possibly to accommodate the increased higher average infant head size during childbirth).24
It is tempting to argue that genetic influence on bodily dimensions is one thing, but that cognitive and behavioral traits are another. But this line has already been crossed. Often when a person participates in a genetic study of a disease, he or she fills out a form providing information on height, weight, and number of years of education. By compiling the information on the number of years of education for over four hundred thousand people of European ancestry whose genomes have been surveyed in the course of various disease studies, Daniel Benjamin and colleagues identified seventy-four genetic variations each of which has overwhelming evidence of being more common in people with more years of education than in people with fewer years even after controlling for such possibly confounding factors as heterogeneity in the study population.25 Benjamin and colleagues also showed that the power of genetics to predict number of years of education is far from trivial, even though social influences surely have a greater average influence on this behavior than genetics. They showed that in the European ancestry population in which they carried out their study, it should be possible to build a genetic predictor in which the probability of completing twelve years of education is 96 percent for the twentieth of people with the highest prediction compared to 37 percent for the lowest.26
How do these genetic variations influence educational attainment? The obvious guess is that they have a direct effect on academic abilities, but that is probably wrong. A study of more than one hundred thousand Icelanders showed that the variations also increase the age at which a woman has her first child, and that this is a more powerful effect than the one on the number of years of education. It is possible that these variations exert their effect indirectly, by nudging people to defer having children, which makes it easier for them to complete their education.27 This shows that when we discover biological differences governing behavior, they may not be working in the way we naively assume.
Average differences across populations in the frequencies of the mutations that affect educational attainment have not yet been identified. But a sobering finding is that older people in Iceland are systematically different from younger people in having a higher genetically predicted number of years of education.28 Augustine Kong, the lead author of the Icelandic study, showed that this reflects natural selection over the last century against people with more predicted education, likely because of selection for people who began having children at a younger age. Given that the genetic underpinnings of the number of years of education a person achieves have measurably changed within a century in a single population under the pressure of natural selection, it seems highly likely that the trait differs across populations too.
No one knows how the genetic variations that influence educational attainment in people of European ancestry affect behavior in people of non-European ancestries, or in differently structured social systems. That said, it seems likely that if these mutations have an effect on behavior in one population they will do so in others, too, even if the effects differ by social context. And educational attainment as a trait is likely to be only the tip of an iceberg of behavioral traits affected by genetics. The Benjamin study has already been joined by others finding genetic predictors of behavioral traits,29 including one of more than seventy thousand people that found mutations in more than twenty genes that were significantly predictive of performance on intelligence tests.30
For those who wish to argue against the possibility of biological differences across populations that are substantial enough to make a difference in people’s abilities or propensities, the most natural refuge might be to make the case that even if such differences exist, they will be small. The argument would be that even if there are average differences across human populations in genetically determined traits affecting cognition or behavior, so little time has passed since the separation of po
pulations that the quantitative differences across populations are likely to be trivially small, harkening back to Lewontin’s argument that the average genetic difference between populations is much less than the average difference between individuals. But this argument doesn’t hold up either. The average time separation between pairs of human populations since they diverged from common ancestral populations, which is up to around fifty thousand years for some pairs of non-African populations, and up to two hundred thousand years or more for some pairs of sub-Saharan African populations, is far from negligible on the time scale of human evolution. If selection on height and infant head circumference can occur within a couple of thousand years,31 it seems a bad bet to argue that there cannot be similar average differences in cognitive or behavioral traits. Even if we do not yet know what the differences are, we should prepare our science and our society to be able to deal with the reality of differences instead of sticking our heads in the sand and pretending that differences cannot be discovered. The approach of staying mum, of implying to the public and to colleagues that substantial differences in traits across populations are unlikely to exist, is a strategy that we scientists can no longer afford, and that in fact is positively harmful. If as scientists we willfully abstain from laying out a rational framework for discussing human differences, we will leave a vacuum that will be filled by pseudoscience, an outcome that is far worse than anything we could achieve by talking openly.
The Genome Revolution’s Insight
On the question of whether traditional social categories of race correspond to meaningful biological categories, the genome revolution has already provided us with new insights that go far beyond the information that was available to the first population geneticists and anthropologists who grappled with the issue. In this way, the data provided by the genome revolution are potentially liberating, providing an opportunity for intellectual progress beyond the current stale framing of the debate.
As recently as 2012, it still seemed reasonable to interpret human genetic data as pointing to immutable categories such as “East Asians,” “Caucasians,” “West Africans,” “Native Americans,” and “Australasians,” with each group having been separated and unmixed for tens of thousands of years. The 2002 study led by Marc Feldman produced clusters that corresponded relatively well to these categories, and the model seemed to be doing a good job of describing variation in many parts of the world (with some exceptions).32 In other papers, Feldman and his colleagues proposed a model for how this kind of structure could arise among human populations. Their proposal was that modern humans expanding out of Africa and the Near East after around fifty thousand years ago left descendant populations along the way, which in turn budded off their own descendant populations, with the present-day inhabitants of each region being descended directly from the modern humans who first arrived.33 Their “serial founder” model was more sophisticated than that imagined by biological race theorists in the seventeenth to twentieth centuries, but shared with it the prediction that after being established, human populations hardly mixed with each other.
But ancient DNA discoveries have rendered the serial founder model untenable. We now know that the present-day structure of populations does not reflect the one that existed many thousands of years ago.34 Instead, the current populations of the world are mixtures of highly divergent populations that no longer exist in unmixed form—for example, the Ancient North Eurasians, who contributed a large amount of the ancestry of present-day Europeans as well as of Native Americans,35 and multiple ancient populations of the Near East, each as differentiated from the other as Europeans and East Asians are differentiated from each other today.36 Most of today’s populations are not exclusive descendants of the populations that lived in the same locations ten thousand years ago.
The findings that the nature of human population structure is not what we assumed should serve as a warning to those who think they know that the true nature of human population differences will correspond to racial stereotypes. Just as we had an inaccurate picture of early human origins before the ancient DNA revolution unleashed an avalanche of surprises, so we should distrust the instincts that we have about biological differences. We do not yet have sufficient sample sizes to carry out compelling studies of most cognitive and behavioral traits, but the technology is now available, and once high-quality studies are performed—which they will be somewhere in the world whether we like it or not—any genetic associations they find will be undeniable. We will need to deal with these studies and react responsibly to them when they are published, but we can already be sure that we will be surprised by some of the outcomes.
Unfortunately, today there is a new breed of writers and scholars who argue not only that there are average genetic differences, but that they can guess what they are based on traditional racial stereotypes.
The person who has most recently made a prominent argument that there is a genetic basis to stereotypes about differences across human populations is the New York Times journalist Nicholas Wade, who in 2014 published A Troublesome Inheritance: Genes, Race and Human History.37 The abiding theme of Wade’s reporting is the propensity of academics to band together to enforce orthodoxies and to be shown up by a band of rebels speaking the truth (he has written on scientific fraud, described the Human Genome Project as a monolith wastefully spending the public’s money, and attacked the value of genome-wide association studies for finding common genetic variations contributing to risk for diseases). Wade’s Troublesome Inheritance ran with the theme again, suggesting that a politically correct alliance of anthropologists and geneticists has banded together to suppress the truth that there are significant differences among human populations and that those differences correspond to classic stereotypes. One part of the argument has something to it—Wade correctly highlights the problem of an academic community trying to enforce an implausible orthodoxy. Yet the “truth” that he puts forward in opposition, the idea that not only are there substantial differences, but that they likely correspond to traditional racial stereotypes, has no merit. Wade’s book combines compelling content with parts that are entirely speculative, presenting everything with the same authority and in the same voice, so that naive readers who accept the parts of it that are well argued are tempted to accept the rest. Worse, when compared to Wade’s previous writing, in which the rebels speaking the truth were scholars of creativity and accomplishment, he does not identify any serious scholarship in genetics supporting his speculations.38 And yet by celebrating those who have opposed the flawed orthodoxy, he implies wrongly that their alternative theories must be right.
As an example of the speculations to which Wade gives pride of place, one of his chapters focuses on a 2006 essay by Gregory Cochran, Jason Hardy, and Henry Harpending suggesting that the high average intelligence quotient (IQ) of Ashkenazi Jews (more than one standard deviation above the world average), and their disproportionate share of Nobel Prizes (about one hundred times the world average), might reflect natural selection due to a millennium-long history in which Jewish populations practiced moneylending, a profession that required writing and calculation.39 They also pointed to the high rate in Ashkenazi Jews of Tay-Sachs disease and Gaucher disease, which are due to mutations that affect storage of fat in brain cells, and which they hypothesized rose in frequency under the pressure of selection for genetic variations contributing to intelligence (they argued that these mutations might be beneficial when they occur in one copy rather than the two needed to cause disease). This argument is contradicted by the evidence that these diseases almost certainly owe their origin to random bad luck—the fact that during the medieval population bottleneck that affected Ashkenazi Jews, the small number of individuals who had many descendants happened to carry these mutations40—yet Wade highlights the work on the basis that it might be right. Harpending has a track record of speculating without evidence on the causes of behavioral differences among populations. In a talk he gave at a 2009 conference on “P
reserving Western Civilization,” he asserted that people of sub-Saharan African ancestry have no propensity to work when they don’t have to—“I’ve never seen anyone with a hobby in Africa,” he said—because, he thought, sub-Saharan Africans have not gone through the type of natural selection for hard work in the last thousands of years that some Eurasians had.41
Wade also highlighted A Farewell to Alms, a book by the economist Gregory Clark suggesting that the reason the Industrial Revolution took off in Britain before it did elsewhere was the relatively high birth rate among wealthy people in Britain for the preceding five centuries compared to less wealthy people. Clark argued that this higher birth rate spread through the population the traits needed for a capitalist surge, including individualism, patience, and willingness to work long hours.42 Clark admits that he cannot distinguish between the transmission of genes and the transmission of culture across the generations, but Wade nevertheless takes his argument as evidence that genetics might have played a role.
I have spent some space discussing errors in Wade’s book because I feel it is important to explain that just because many academics have been engaged in trying to maintain an implausible orthodoxy, it does not mean that every unorthodox “heretic” is right. And yet Wade suggests precisely this. He writes, “Each of the major civilizations has developed the institutions appropriate for its circumstances and survival. But these institutions, though heavily imbued with cultural traditions, rest on a bedrock of genetically shaped human behavior. And when a civilization produces a distinctive set of institutions that endures for many generations, that is the sign of a supporting suite of variations in the genes that influence human social behavior.”43 In a written version of a nod and a wink, Wade is suggesting that popular racist ideas about the differences that exist among populations have something to them.