What this covers
Sean Carroll interviews David Reich, a geneticist at Harvard Medical School, on how ancient DNA has rewritten the narrative of human population history. Rather than stable ancestral groups that gave rise to present-day peoples, the genetic record reveals constant flux: populations migrating, intermixing, and leaving traces in living populations long after vanishing without descendants. Reich traces how reconstructions based on living populations are routinely overturned once DNA from specific archaeological contexts is actually sequenced, a pattern that has now appeared across Europe, Africa, India, Central Asia, East Asia, and the Pacific.
The discussion covers concrete cases: the Yamnaya expansion across the steppe five thousand years ago, which replaced diverse local cultures and contributed substantially to northern European ancestry; the discovery of the Denisovans, a previously unknown human group whose existence emerged from genetic analysis of a single finger bone; and the repeated finding of ghost populations—ancestral groups statistically necessary to explain present-day mixture but unobserved until their DNA is recovered from the right place and time. Reich examines sex-biased migration patterns using X chromosome data, the caste system's role as a frozen picture of ancient Indian population structure, and the relationship between genetic patterns and the spread of Indo-European languages. He also addresses the evidence for interbreeding between modern humans and Neanderthals—a discovery that contradicted prior genetic orthodoxy—and discusses whether the Biological Species Concept remains scientifically useful when past human groups demonstrably interbred and produced viable offspring.
Reich argues that ancient DNA reveals human population history to be one of constant migration, mixture, and now-vanished 'ghost populations,' so present-day diversity systematically obscures and misrepresents the deeper past.
- Reconstructions from living populations are repeatedly contradicted when DNA from known ancient times and places is actually sequenced
- Non-Africans carry Neanderthal DNA, proving interbreeding that the prior orthodoxy denied
- Almost no group is without major mixture in its history; migration and admixture are integral to human history at the scale of millennia
When ancient populations are directly sequenced, genetic reconstructions contradict prior predictions based on living populations and archaeological theory.
- Whenever DNA is collected from an archaeological context never before interrogated and a prior guess is made about how those people relate to later and present-day populations, the guess is almost always wrong and often profoundly wrong, a pattern now seen across Europe, multiple parts of Africa, India, Pakistan, Central Asia, East Asia, Japan, the Americas, and the South Pacific.
“when one goes and collects DNA from archaeological context that have never been interrogated before with this technology and makes a guess beforehand about how these people would be related to people living afterward and people living today, almost always that guess is wrong, and often profoundly wrong”
- When you sequence DNA from ancient archaeological sites and test how it relates to other sites and to people today, the reconstructions are consistently and often profoundly different from those based on living populations, because people have moved around so much that present-day diversity obscures what was there before.
“when one goes and collects DNA from archaeological context that have never been interrogated before with this technology and makes a guess beforehand about how these people would be related to people living afterward and people living today, almost always that guess is wrong, and often profoundly wrong”
- Genetic data from the 1980s onward suggested modern humans shared no DNA plausibly derived from Neanderthals, since mitochondrial and Y-chromosome lineages all coalesced within the last couple hundred thousand years; population geneticists argued that even slight interbreeding during a range expansion should produce nuclear swamping by the resident group, and its absence implied very little interbreeding or some incompatibility.
“when I went to grad school, that was the orthodoxy, that was the situation, we thought there was no evidence amongst modern humans of divergent ancestry that might be consistent with substantial interbreeding with a group like Neanderthals”
- Analysis of Neanderthal genome data showed a clear, consistent signal of interbreeding: essentially all populations outside sub-Saharan Africa carry a small amount of Neanderthal DNA, the signal was just as strong in East Asians as in Europeans (surprisingly, since no Neanderthals lived in East Asia), and multiple independent statistical techniques all pointed to a history of interbreeding.
“when we looked at the data, we saw such a signal, and we were actually quite incredulous at that signal... They were all pointing very clearly and consistently at a history of interbreeding”
Ancient DNA technology has advanced exponentially in extraction efficiency and now enables genome-scale reconstruction from degraded archaeological remains.
- Degraded ancient DNA is naturally fragmented into pieces rarely more than 60-70 bases, with typical sequenced reads of 30-50 bases, but because a degraded extract still contains billions of molecules, brute-force sequencing of hundreds of millions of independent fragments can be puzzled together into genome-scale data.
“there are billions and billions of molecules even in a degraded DNA extract and by brute-forcing it or using various tricks, we can get many hundreds of millions of independent DNA fragments and puzzle together, genome scales worth of data”
- The ancient DNA revolution of the last decade is technically driven: DNA is now extracted with roughly eight or nine orders of magnitude more efficiency (in cost terms) than 13 years ago, enabled by short-read sequencing in the late 2000s and by abandoning PCR, which sacrificed 30-40 of the typically only 40-50 base fragments and thus destroyed nearly all the target.
“we're probably getting DNA out with maybe eight or nine orders of magnitude more efficiency than we were 13 years ago”
- DNA does not preserve more than about a million years and mostly not more than hundreds of thousands of years, so ancient DNA cannot illuminate deep-time mammalian radiations, but it powerfully interrogates the last tens to hundreds of thousands of years during which modern humans descend from a common ancestral population.
“ancient DNA, DNA doesn't really preserve more than a million years, and mostly doesn't preserve more than hundreds of thousands of years”
- The number of sequenced ancient genomes went from zero in 2009 to five in 2010, around 40 by 2014, and well over 7,000 today.
“in 2009, it was zero, and in 2010, it was five, and maybe in 2014, it was maybe 40 or so, and now it's well over 7,000”
Humans 70,000 years ago were vastly more genetically diverse and differentiated into at least five distinct groups that were subsequently subsumed or displaced.
- Genetic differentiation within a US-census 'white'/Western-Eurasian category a few hundred years ago resembled today's, but roll back 8,000 years and groups within that same European region were as different from each other as Europeans and East Asians are now; if census categories were assigned in the deep past they would break along completely different fault lines, so the past is not well-described by the present.
“roll back 8000 years and there would be many groups in that region is different from each other as European and East Asians are... if you were to assign census categories in the past, they would have broken down along completely different fault lines from the way you break them down today”
- Only 70,000 years ago there were at least five groups living around the world, known through skeletal and DNA evidence, each more different from one another than the most diverse human groups today, but within a few tens of thousands of years these were subsumed, mixed, or displaced by the expansion of modern humans, leaving us alone on the planet.
“only 70,000 years ago, there were at least five groups living around the world that we know about today... that were each as different from each other, much more different from each other than any groups living today... But within a few tens of thousands of years, we were alone on the planet”
- FST measures how genetic variation is partitioned between versus within populations; today it is roughly 0.1, meaning the genetic difference between two random individuals within a population is about six times greater than the average difference across populations, but this cross-population fraction would have been higher 70,000-100,000 years ago.
“That's known as FST, it's a measure of how the variation is partitioned... Today, that number is maybe 0.1... the amount of genetic difference between two individuals or two genomes within a population is six times greater than the average across two populations”
- Western chimpanzees alone have about as much genetic diversity as all humans, and central, eastern, and Nigerian chimpanzees taken together are more diverse than modern humans, suggesting claims about human youth and homogeneity are sometimes exaggerated, though humans 70,000 years ago may have rivaled chimpanzee diversity.
“today, chimpanzees, western chimpanzees, if I remember right, have about as much diversity as humans do, but there are also some other chimpanzees groups... and together all of those chimpanzees are more diverse than modern humans are”
Population histories are far more complex than present-day census categories or ethno-linguistic distributions suggest, with constant churning and mixture.
- Almost all pre-Columbian Native American ancestry derives from a single group called the First Americans who spread across the Bering land bridge around 15,000 years ago, but genetic hints—especially groups in Brazil/Amazonia more closely related to Southeast Asians, Australians, and New Guineans—suggest the earliest spread was substructured, implying the 'First Americans' name is misleading, though Reich notes these are only four-sigma effects of uncertain significance.
“almost all the ancestry of Native Americans prior to today that doesn't come from European and African migrants in the last 500 years comes from a single ancestral group that we call First Americans, that likely spread into the Americas maybe around 15,000 years ago”
- The Indian caste system includes thousands of endogamous local groups that, despite living among one another, have not mixed for thousands of years, producing in their DNA a frozen picture of population structure from a couple thousand years ago; modern Indians are well-approximated as mixtures of Ancestral North Indians (closer to Central Asians, Middle Easterners, Europeans) and Ancestral South Indians, themselves more anciently mixed from Iranian-related, steppe-pastoralist, and ancient Southeast-Asian-related ghost groups.
“many of these groups really have not mixed with very much with other groups, they've lived amongst... for thousands of years... through social norms, made it almost impossible for members of their groups to mix with neighbors”
- Looking at genetic data makes clear that there are very few, maybe no, groups in the world that lack major mixture in their history; like the cyclical comfort-then-pogrom pattern of European Jewish history, on a long time scale human populations are constantly churning, so the common belief that one descends from people who were always there without migration or mixture is generally false.
“looking at genetic data makes it absolutely clear that there are very few of maybe no groups in the world that don't have major mixture as part of their history... migration and mixture are very central and integral to us when we think at the temporal scale of many thousands or tens of thousands of years”
- Steppe-pastoralist ancestry pulsed into India between 4,000 and 3,500 years ago, is never more than about 20% (sometimes near 0%), and is relatively higher in traditionally higher-status caste groups and especially in custodians of Indo-European texts like Brahmins and Bhumihars, providing strong genetic evidence that Yamnaya-derived steppe ancestry spread Indo-European languages into India and militating against a cross-Iranian-plateau origin.
“it's relatively much higher in frequency in people who have traditionally such higher social status in the Indian caste system, and it's especially high frequency adjusting for other factors in people who are from the traditionally custodians of the Indo-European texts, like the Brahmins and Bhumihars”
Major population expansions across Eurasia and the Pacific, including the Yamnaya and Austronesian seafarers, reshaped ancestry without centralized state organization.
- Austronesian-speaking, Taiwan-derived seafarers settled the previously uninhabited far Pacific islands about 3,000 years ago arriving as almost entirely East-Asian/Taiwanese ancestry, but roughly 500 years later a massive, sex-biased wave of primarily male migrants from New Guinea (Papuans) nearly completely displaced them, so present-day Vanuatu retains only about 10% of the initial Taiwanese-related ancestry—the opposite of the intuitive expectation that the seafaring technologists would be the expanding population.
“there's this initial expansion 3000 years ago or so that gets to Vanuatu with almost entirely Taiwanese-related ancestry. Today it's maybe only 10%, and then maybe 500 years later, there's a massive wave from New Guinea... It's a very sex-biased process where primarily male people from New Guinea mixed with these seagoing ocean people”
- The Yamnaya expanded across the steppes north of the Black and Caspian seas about 5,000 years ago using two new inventions—the wheel and the domesticated horse—to graze larger herds in previously uninhabitable steppe, replacing diverse local ceramic cultures with a monoculture from central Europe to Mongolia; genetic data show this had a massive demographic impact, with more than half the ancestry of many northern European populations today descending from this expansion.
“we now know from the genetic data that this was a massive impact, so in Northern Europe, more than half the ancestry of many northern parts of Europe today descends from this expansion”
- Despite their vast geographic reach, the Yamnaya were almost certainly highly decentralized with no evidence of centers of power, central settlements, large congregations, or systematic warfare; their effectiveness likely came from exploiting unused grazing niches and possibly cattle/resource raiding and male-centered expansion practices reflected in Indo-European mythology, not from organized state-level invasion.
“these people were highly decentralized, there's no evidence of... centers of power or central settlements, there's no evidence of large congregations of individuals, there's no evidence of systematic war-like events”
Ancient DNA research requires integration of archaeological expertise and domain knowledge beyond current technical specialist-dominated practices.
- Europe yields the most ancient DNA not because it is special but because the technology developed in European labs, European archaeologists assembled and curated skeletal material for over a century, and resources exist there; thus Europe functions as a laboratory showing what the technology can do, now yielding a nearly gap-less genomic record over the last 10,000 years and a gappier one back to 50,000 years.
“Europe's not more important than other places of the world, but it's a place that you can use almost as a kind of laboratory to understand what's possible with applying this type of technology to learn about the past”
- Reich's physics background helps his work because his lab is a hybrid that both generates data and develops data-driven methodology to analyze it in new ways, which requires sophisticated quantitative thinking.
“a lot of my work relies on trying to be sophisticated about quantitative thinking about the data that we produce”
- Ancient DNA is currently in a prospecting, 'Wild West' phase dominated by technicians who can generate and analyze the data, but the people who know the archaeological questions best are not doing the primary research; like genome-wide association studies that moved from geneticists to epidemiologists assembling hundreds of thousands of subjects, the field's future is transferring the technology to domain experts who can pose the best questions.
“the future is one where the people who really know about these topics will be able to learn enough about these fields, to use the technology to address the questions that they know are most interesting”
- From a single British tomb dating to about a century after farmers arrived nearly 6,000 years ago, the team reconstructed a 27-person family entirely descended patrilineally from one patriarch who reproduced with four women, but kinship was not purely biological—there are multiple cases where the patriarch's descendants adopted sons their wives had previously had with other men, revealing socially-defined rules of fatherhood and kinship.
“the tomb was one where it was entirely patrilineal, everybody buried in the tomb, it descends from the same patriarch who reproduced with four women... we actually see multiple cases of adoption. So, this man and his descendants adopted other sons that their wives previously had with other men and incorporated them into the pedigree”
- The lineage giving rise to the great majority of people today was clearly in Africa from about 400,000-500,000 years ago, the earliest anatomically modern skeletons date to 200,000-300,000 years ago in different parts of Africa, and there was an explosion of modern humans out of Africa and the Near East between 50,000 and 100,000 years ago, reaching Australia and New Guinea quickly and the Americas after 15,000 years ago.
“the earliest skeletal remains of people whose skeletons look anatomically modern are 200,000 to 300,000 years ago in different parts of Africa. And then there is an explosion of anatomically modern humans out of Africa and the Near East between 50,000 and 100,000 years ago”
- Mitochondrial DNA is about 16,000 letters long, only a tiny fraction of the genome, and records solely the entirely maternal lineage, so it is a single statistical instantiation of the evolutionary process and provides limited information, whereas the whole nuclear genome records tens or hundreds of thousands of genealogical ancestors and yields precise relatedness information.
“it contains the record of our mother's mother's mother's mother's mother and entirely maternal lineage, which is of course fascinating. Important, but it's limited in terms of its statistical information it provides about the past, 'cause it's only a single statistical instantiation of the evolutionary process”
- The most recent common female (mitochondrial) ancestor is roughly 150,000-180,000 years ago, but a Y-chromosome type found at substantial frequency in Cameroon is about twice as deep, so the oldest common male ancestor predates the oldest common female ancestor—contrary to the intuition that men spread their lineages faster.
“The oldest common male ancestor is about twice as deep as that of the most recent common female ancestor... the common female ancestor is roughly 150 to 180,000 years ago... a Y chromosome type that was twice as deep in African-Americans”
- In 2010 the team sequenced high-quality DNA from a finger-bone tip in a Siberian cave that proved to belong to a previously unknown human group, almost as different from Neanderthals as Neanderthals are from modern humans and separated by perhaps 400,000 years, which they named Denisovans rather than coining a new species name.
“it turned out not to be from any known human group, it was from a group that was almost as different from Neanderthals as Neanderthals were from human. It was separated by many hundreds of thousands of years, maybe 400,000 years from anything we had seen before”
Neanderthals and modern humans interbred successfully despite substantial genetic differentiation and unclear causes of Neanderthal disappearance.
- When Neanderthals and modern humans co-existed in any region it was limited to at most a couple thousand years before Neanderthals disappeared, but many co-existing modern human groups also disappeared around 39,000 years ago and were replaced by a relatively homogeneous archaeological culture, so it is not obvious Neanderthals were simply out-competed; whether displacement, conflict, incompatible psychology, or absorption through mixture caused their disappearance remains unknown.
“it's not obvious that the Neanderthals were out-competed by modern humans... after about 39,000 years ago, many of the early modern human archaeological cultures that are evident in Europe before that time, as well as Neanderthals, disappear and are replaced by a relatively homogeneous archaeological care”
- Neanderthals appeared a few hundred thousand years ago and persisted in Europe until about 40,000 years ago; they were as large or larger than modern humans, had allometrically scaled brain capacities as large or larger, and made complex tools requiring rich learning and skill for hundreds of thousands of years.
“Their brain capacities were as large or larger than those of modern humans, even though their skulls were differently distributed”
- Under Ernst Mayr's biological species concept, species are groups that do not in practice interbreed and form successful mixed populations, but Neanderthals, Denisovans, and modern humans did interbreed and produced successful mixed groups (such as all non-Africans), so the team declined to designate Denisovans a new species, treating the question as philosophical rather than scientifically useful.
“the biological species concept, originally suggested by the biologist Ernst Mayr who argued that species are groups of organisms that do not in practice interbreed with each other, and what we had seen in our genetic data is that these groups were interbreeding with each other, and we're able to form successful mixed groups”
Ghost populations inferred from mixture models in present-day peoples are later directly observed in ancient DNA, validating the statistical approach.
- A ghost population is a group that emerges from statistical mixture models as a necessary ancestral source for present-day populations but no longer exists in unmixed form; for example present-day Europeans fit well as a mixture of three ancestral populations (none of which exist today, Sardinians being the closest proxy), and repeatedly such predicted populations are later directly observed once DNA is obtained from the right time and place.
“That's what we call a ghost population, a population we reconstruct statistically from groups living today, but we don't have samples from today, and what we find again and again with DNA is we predict these populations and then we see them once we obtain DNA from the right time in place”
- Ethnographic studies show languages tend to spread through movements of people, so the wide distribution of Indo-European languages from India to the Atlantic, and of Austronesian from Taiwan across the Pacific, can be tested against genetic data; the genetics provides a 'tracer dye'—distinctive frequency distributions of a source ghost population—that traces the likely movements, though language conversion without migration does occur (e.g., Hungary).
“we know from ethnographic studies that languages tend to spread through movements of people, and so large scale movements of people... usually it's through large scale movements of people”
X chromosome patterns reveal sex-biased demographic processes, with women moving locally and long-distance expansions driven primarily by male movement.
- Because two-thirds of X chromosomes reside in women versus half-and-half for the rest of the nuclear genome, the X is effectively a 'female-colored' chromosome carrying far more information than the single-instantiation Y and mitochondrial DNA; by solving the system of equations comparing the two-thirds-one-third X ratio to the half-half autosomal ratio, one can extrapolate the male versus female contribution to past population mixture events.
“because X chromosome is a kind of female-colored chromosome, because two-thirds of the X chromosomes running around in the world today are in women compared to only one-third in men... by looking at the X chromosome and solving the system of equations... one can extrapolate out what the male and female contribution to different events would have been”
- Mitochondrial differentiation tends to be lower at short geographic scales but higher at large scales, while Y chromosomes show the opposite, implying that in many (often patrilocal) communities women move modestly between neighboring communities causing local-scale mixing, whereas long-range population movements over thousands of years tend to be propelled to a larger extent by male movement.
“in a lot of communities... women are the ones who move between communities... long, long range mitochondrial movements seem to be... human movements seem often to be propelled to a larger extent by male movement”