The first, second, and third nations

By Razib Khan | July 11, 2012 10:47 pm

By now you’ve probably read about the paper which reports that there seem to have been three waves of humans migrating into the New World prior to the arrival of Europeans. A major aspect of this result is that it does not emerge out of a vacuum, but rather comes close to settling an old question in linguistics. The late Joseph Greenberg generated a series of audacious phylogenies of languages of the world. Greenberg’s attempts received mixed reviews. It seems that there is little controversy about some of his classifications of African languages, but linguists of American native dialects rejected his division of the languages of the New World into three broad families, Eskimo-Aleut, Na-Dene, and Amerind. Eskimo-Aleut is rather self-evident. Na-Dene encompasses a group of languages in northwest North America, along with some significant outliers such as Navajo. Amerind seems to roughly be a grab-bag of everything else. The linguistic trichotomy also lent itself to a narrative of three migrations. L. L. Cavalli-Sforza gave his support to Greenberg’s framework in The History and Geography of Human Genes, and it seems most non-linguists are particularly congenial toward his tendency of ‘lumping.’ In contrast, linguists remain more skeptical ‘splitters,’ at lease those who have a more ethnographic disciplinary bent. Geneticists have not always supported Greenberg’s suppositions. For example, many of the members of the same group which authored this paper implicitly put the kibosh on the attempt to construct a unified linguistic family which spanned the Andaman Islanders and the Papuans.

The method of the paper was relatively straightforward, assuming you are already somewhat familiar with the statistical genetic esoterica which was unveiled a few years ago by this group and others. Basically you take genetic data in the form of hundreds of thousands of SNPs, and you test the patterns of variation in that data across populations against explicit models of demographic history, represented visually by phylogenetic trees. You can see here that the sampling was relatively thick, except for the United States. Chalk this up to politics. I’ve been hearing about this particular problem in relation to this paper for over a year now. Not having asked any of the members of the group directly I obviously am going off hearsay, but the lack of American samples is most definitely not a feature. It’s a bug. In the supplement they also note that they couldn’t get Na-Dene data from another research group. Almost certainly that’s because of bioethical issues and legal contractual constraints.

Despite all this drama, the scientific isn’t too hard to understand. Aside from the nifty statistics one problem is that many of these native groups have European and African admixture, but there are workarounds to that (e.g., just pull out genomic segments which are indigenous, and use those). The outcome is neatly visualized in the figure below:


What you see here is that the “Amerind” group derive from one ancient founding lineage. This comports with other data which seem to imply some sort of bottleneck. The authors couldn’t peg a date from their data, and so they support the proposition of a ~15,000 year date of colonization. This is aligned with archaeology (even if you reject Monte Verde, one can posit a sojourn in Berengia to explain the “gap”). One inference you can make from their caution about timing is that these are not newcomers; if dynamics outlined above were recent they’d probably be able to pick them up (e.g., measuring linkage disequilibrium decay since admixture). And as expected the authors found a north-south gradient in heterozygosity, what you’d see if the population went through serial bottlenecks north to south.

But the non-Amerind populations are more interesting. The Eskimo-Aleut turn out to be a 50:50 mix between an Amerind group, what they term First Americans, and an Asian lineage related to Siberians. This is not totally surprising, but it is good to get solid grounding. Additionally, they report evidence of back-migration to eastern Siberia. Again, not too surprising, but it is a neat confirmation of the reality that the separation between the Old and New World was illusory in some deep ways. The Na-Dene population here, the Chipewyan, are a different case. They are ~90% First American, but also ~10% something else. That something else is also more Asian, but not quite the same as the non-First American component of Eskimo-Aleuts. This population X is quite possibly the Old World source for the Na-Dene. As the authors apologetically note they didn’t have many other Na-Dene samples to explore this question in detail. But, there is one aspect which they explore a great bit in the supplements: some models suggest that the Paleo-Eskimo Saqqaq sample exhibits a mix with First Americans at 15%, and 85% with this Asian population. With these sample sizes the statistical significant doesn’t seem rock solid, but it’s obviously suggestive. If you don’t remember, the Saqqaq sample comes from a man who died ~4,000 years ago. He seems to have resembled modern day Siberians more than Eskimo-Aleut, or indigenous peoples to the south. These data imply that the dominant component of the Saqqaq’s ancestry may indeed have been the minor component in the Na-Dene! A final twist is that the First American ancestry of the Saqqaq and Chipewyan is somewhat different than the First American ancestry of the Eskimo-Aleut. This is important in establishing a distinct ethnogenesis of the various groups, and their relationships to each other (I take their assertion that the First American of the Eskimo-Aleut being more “derived” a hint to likely later ethnogesis of this group via admixture than the Na-Dene).

To me the above implies that the closeness of the Saqqaq to the Siberian groups may be an artifact of the fact that the Eskimo-Aleut are synthetic populations, with ~50% First American, while the Saqqaq were only ~15% First American. Some have expressed curiosity as to how Na-Dene languages spread if the Chipewyan are only ~10% Na-Dene, using a grossly simplistic equivalency between language and ethnicity, but it strikes me that over time admixture could slowly reduce the genetic difference between the Na-Dene and their Amerind neighbors. By analogy, the Hui Muslims of China seem about ~90% Han Chinese genetically, but this can be easily explained by a very moderate amount of intermarriage over the past ~1,000 years, when you consider how small the Hui populations is in relation to the Han. One thing to note is that the Saqqaq may have been the first humans to settle much of the northern fringe of North America, and the Saqqaq man who was sequenced was alive during the very early centuries of this culture. One can easily imagine a rapidly expanding population pushing into uninhabited lands admixing very little with natives who did not exist. In contrast their Na-Dene cousins to the south were pushing into settled territory, and so experienced much greater admixture.

On a big picture note this puts the lie to the idea that before agriculture hunter-gatherer societies were subject purely to differentiation in situ. The Eskimo-Aleut and Na-Dene erupted into a settled landscape, and dispossessed the indigenous groups of their lands. How? The standard explanation for the Thule replacement of the Dorset is that the former were better adapted to northern climes. I know of no explanation for the Na-Dene expansion, but it probably had some rationale, social or technological. The fact relatively “pure” Saqqaq were later supplanted by a genomically hybrid populations also points to the complexity of these migratory narratives. The First Americans “struck back” in this case, shoving aside the pioneers of northern living who had likely originated more recently from the margins of eastern Asia. Of course, the Eskimo-Aleut and related peoples were not First Americans only, rather, it was the old (First American) and new (Asian) ganging up upon the not so old or new (Asian).

Addendum: Geneticists using classical markers and physical anthropologists have supported this three wave model. This genomic approach is simply sharper and more precise.

Citation: Reconstructing Native American population history, doi:10.1038/nature11258

  • Charles Nydorf

    It would be great to use population genetics to help untangle the historical relationships between the extremely diverse set of languages that Greenberg lumped as Amerind.

  • Karl Zimmerman

    Have there been any genetic studies comparing Na-Dene populations with Yeniseians? From what I gather, mainstream linguists are finally accepting there may be a linguistic linkage, as Vajda’s work is considered more persuasive than the work of Merritt Ruhlen and other Greenberg-influenced academics.

    IIRC, genetics has suggested the Selkup are an admixture of a Yenisian substratum and a Uralic group, so genetic links should be found there as well. Certainly Haplogroup Q of of Y-DNA is present in both Yenisians and Selkups, but it is also hugely prevalent throughout the Americas – and actually depressed in the north – which would argue against a linkage.

    One also wonders if this second migration was how Haplogroup X got into North America, which is only found in North America, and more dominant in the northern reaches than the South. From what I have read, the only Asian population known to have it is the Altai, who have the same variant as native North Americans. Oddly, other Turkic populations in central Asia lack X, suggesting it was a holdover from an indigenous population which underwent language shift. This is particularly the case because any expansion into North America must have happened well before the Turkic expansion.

    The Y-DNA haplogroup C3b is also interesting. AFAIK, not a lot of Na-Dene populations have been tested for it, but it has a higher frequency in the interior western Canada, and is fairly prevalent among the Apache (15%). In general it seems like from this study that some Algonquin populations received some admixture from whatever population source birthed the Na Dene, so high frequency in groups like the Cheyenne might be evidence as well.

    Of course, the lack of prevalence of both these rarer Y-DNA and mt-DNA groups further south could in part just be due to the aforementioned genetic bottlenecks. Still, it’s intriguing. Hopefully with the growth of personal genomics in the U.S., more Native American genomes from the USA will be studied soon.

  • Mike Keesey

    I’ve seen it pointed out elsewhere that just because the First Wave was genetically cohesive doesn’t mean it was necessarily linguistically cohesive. It is possible that it was multilingual and that “Amerind” is a polyphyletic assemblage. I personally know next to nothing about the linguistic arguments, but, as I understand it, the linguistic data is generally considered ambiguous on this point.

  • Razib Khan

    #3, agreed. also, wouldn’t 15,000 years arguably result in polyphyletic trees anyway???

  • Karl Zimmerman

    Mike -

    Given from what I have read, the first wave is estimated to have sprung from a very low population base (anywhere from low hundreds to dozens), I don’t know how you’d end retain a linguistically diverse population. In all historical situations where a mixed-language group comes into formation, the population becomes monolingual within a generation, with either the language of most prestige winning wholesale, or (in groups more isolated) an entirely new creole language forming.

    I can believe that 15,000 years is too old for Amerind to be proven linguistically, but ultimately the languages need to have common antecedents, and those are most likely (barring anything like say an introduction of multiple language families by elite dominance by Eurasian/Australian groups who left no genetic record) going to be found within the Americas, or Beringia at worst.

  • anonitin

    Have there been any genetic studies comparing Na-Dene populations with Yeniseians?

    Reich and company are aware of the Yeniseian hypothesis but their analyses didn’t reveal any particularly pronounced connection between Kets and Na-Dene (or any other Native pops). But then again, the n for Kets was very low and (as Wade relates in the NYC) the presently available samples may be unrepresentative due to admixture (I guess with other Siberian pops? I don’t think they seemed especially Russian)

  • Ortu Kan

    @ Karl Zimmerman:

    One also wonders if this second migration was how Haplogroup X got into North America, which is only found in North America, and more dominant in the northern reaches than the South.

    re: mtDNA X in the Americas, you may be interested to see my compilation of records from the literature here.

    From what I have read, the only Asian population known to have it is the Altai, who have the same variant as native North Americans.

    Actually, neither of the two American Indian subclades of X (X2a and the Ojibwa singleton X2g) have so far been detected anywhere in Eurasia. Reidla et al. (2003) established that the Altaian sequences all belong to X2e and suggested recent acquisition from the southern Caucasus. They also remarked on some new finds in Evenks:

    In contrast to the Altaians, the Evenks did not harbor any West Eurasian mtDNA haplogroups other than X. However, neither of the two Evenk X haplotypes showed mutations characteristic of the Native American clade X2a. Instead, one sequence was a member of X2b and the other of X2*.

    So the waters remain murky.

  • Tom T

    The overlay of the geographic distribution of haplogroup X and the known Minoan trading empire is nearly exact, providing strong evidence that the Minoans were the source of this genetic material.–The-Case-of-the-Missing-Copper/9686450?pid=247221

  • most_peculiar

    Very interesting, Tom T, and probably the most parsimonious explanation for such otherwise unaccountable puzzles as the existence of the bull-leaper motif in the wooden sculptures of pre-contact Vancouver and Navajo pottery alike.

    The close contact between Minoan men and Native American women, not surprisingly, appears to have resulted in the mixing of genetic material.

    Yes, most characteristically including the conversion of Y-chromosomes into mitochondria.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar

Login to your Account

E-mail address:
Remember me
Forgot your password?
No problem. Click here to have it e-mailed to you.

Not Registered Yet?

Register now for FREE. Registration only takes a few minutes to complete. Register now »