Beringian in the News.

I started off this post: “I started off this post: ‘The NY Times has another language story [...] and if you’re an aficionado of these things you will have guessed that 1) the story is by the muddled but ever plucky Nicholas Wade…’” And yes, the Times has another story by the muddled but ever plucky Nicholas Wade. This time it’s about “a recent proposal that the ancestors of Native Americans were marooned for some 15,000 years on a now sunken plain before they reached North America”:

This idea, known as the Beringian standstill hypothesis, has been developed by geneticists and archaeologists over the last seven years. It holds that the ancestors of Native Americans did not trek directly across the land bridge that joined Siberia to Alaska until the end of the last ice age, 10,000 years ago. Rather, geneticists say, these ancestors must have lived in isolation for some 15,000 years to accumulate the amount of DNA mutations now seen specifically in Native Americans. [...]

Linguists have until now been unable to contribute to this synthesis of genetic and archaeological data. The first migrations to North America occurred between 15,000 and 10,000 years ago, but most linguists have long believed that language trees cannot be reconstructed back further than 8,500 years. Vocabulary changes so fast that the signal of relationship between two languages is soon swamped by the noise of borrowed words and fortuitous resemblances.

But in 2008, Edward Vajda, a linguist at Western Washington University, said he had documented a relationship between Yeniseian, a group of mostly extinct languages spoken along the Yenisei River in central Siberia, and Na-Dene. [...]

Building on Dr. Vajda’s success, two linguists, Mark A. Sicoli of Georgetown University and Gary Holton of the University of Alaska, have assessed the relationship of the two language families based on shared grammatical features, rather than vocabulary.

In a paper published in the journal PLoS One on Wednesday, they report their surprising finding that Na-Dene is not a descendant of Yeniseian, as would be expected if the Yeniseian speakers in Siberia were the source population of the Na-Dene migration. Rather, they say, both language families are descendants of some lost mother tongue. Their explanation is that this lost language was spoken in Beringia, and that its speakers migrated both east and west. The eastward group reached North America and became the Na-Dene speakers, while the westward group returned to Siberia and settled along the Yenisei River.

Now, I’m suspicious of this on all sorts of grounds, but everything I know about Na-Dene and Yeniseian I’ve learned from commenters here, and I’m hoping some of them will weigh in on the proto-Beringian theory. (Thanks, Eric!)

Comments

  1. The PLOS paper apparently doesn’t necessarily endorse the hypothesis that Na-Dene and Yeniseian are genetically related–it just concludes that, assuming the hypothesis is true, a back-migration from Beringia to Asia (and a migration into North America by another branch) is more probable than a migration from Asia to North America.

  2. Rather, they say, both language families are descendants of some lost mother tongue.

    Of course. Nobody ever thought otherwise. “Jones Wrong, Latin and Greek Not Descendants of Sanskrit! Film at 11.” As to where Proto-Dene-Yeniseian was spoken, there are all kinds of possibilities, and the former Beringian isthmus is certainly one — but so is everywhere in the vast expanse of Siberia from the Yenisei to the Strait. I don’t think we are even sure whether Siberian Yupiks are aborigines or remigrants, though scholarly opinion inclines to the latter because of the greater diversity of the family in Alaska. It’s not as if Eskimos don’t routinely cross the Bering Strait anyway.

    I haven’t read the Sicoli-Holton paper yet, but looking at the cladogram, it purports to show a deeper divide between “coastal” Dene-Yeniseian languages (Eyak, Tlingit, Yeniseian, and Pacific Coast Athabaskan) and “interior” languages (the rest of Athabaskan) than there is between Yeniseian and Na-Dene.

    But of course even if true, that proves nothing. With suitable characters , you’ll find a greater difference between French and the other Romance languages than they have among themselves. But that does not mean that French and non-French are the top-level division of Romance, simply that French has changed faster than the rest. Likewise, it is likely that Athabaskan languages spread from the coast inland, so any changes that affected them on the way there would be isolated from the rest of Dene-Yeniseian. Assume that the Navajo/Apache migrated from inland Alaska, and there you are.

  3. The study uses a bunch of typological characters to cluster the Athabaskan and Yeniseian languages, and finds Yeniseian tucked among the coastal languages, rather than show Yeniseian and Dene as two primary branches.
    My quick observations:
    — The Beringian pause hypothesis is reasonable by itself, based on entirely non-linguistic criteria. That’s neither here nor there.
    — Sicoli and Holton do not entertain any other explanations for their data. For example, their tree is consistent with proto-Dene splitting from Dene-Yeniseian somewhere in Siberia, moving to Alaska and spreading to the interior from there, after which the interior and Apachean groups share some innovations, perhaps because of contact with other language groups.
    — The fact remains that the Athabascan languages (setting Tlingit aside) are very clearly relatable to each other in grammar, lexicon, etc., whereas Yeniseian is remotely related to them. That is clearly inconsistent with their typological tree, but they do not attempt to reconcile the two. The only reference they make to other linguistic methods is to Wichmann et al.’s automatic lexicostatistical technique, which is more consistent with traditional historical method though inferior to it.
    — S&H do not place any other language groups of the area in their typological phylogeny (e.g. Tungusic, Eskimo-Aleut, Nivkh, Salishan, etc.), except Haida. Haida (now considered unrelated to Tlingit-Dené, and typologically distinct from them) shows up as an outlier, but I wonder if the other language groups would be so clearly separated as well.
    — S&H make no attempt to tease out what particular typological features are responsible for separating the two clusters. Without that, it’s hard to make an informed linguistic (as opposed to merely statistical) judgment of their results.

  4. Thanks, my suspicions are confirmed!

  5. My understanding of the Yeniseian-Dene hypothesis is that, while many linguists find it plausible, it’s based primarily on striking correspondences in the structure of verb-forms, but that sound-correspondences haven’t been securely established (and consequently some prominent American language specialists such as Lyle Campell have remained skeptical). I don’t have the background to really dig deep into the S&H paper, but I wonder how they could produce these results on the basis of what are apparently skimpy data.

  6. Trond Engen says:

    Striking correspondences in the structure of verb-forms beats sound correspondences any day. At least when the morphological correspondences go beyond typology and into plausible phonology. Cognates are secondary, supporting evidence.

  7. Here’s more. I got the data files from the paper’s supplementary information. There are 116 characters in total, 81 phonological and 35 morphological. If you run SplitsTree on the morphological characters alone, there’s a very clear separation of the Yeniseian languages and all the rest (with Haida clustering with Eyak and Sarsi. A NeighborNet graph based on phonological characters alone is close enough to Sicoli and Holton’s diagram.
    In other words, this all comes down to comparing phonological inventories. pretty weak. I spent about a whole hour on this. Why didn’t the authors? Feh.

    If I spent another hour or two on this, I could tell you which phonological features set the coastals + Yeniseian apart from the rest.

  8. Trond Engen says:

    They say they threw out 26 of them for being of no diagnostic value, i.e. present in none or all languages, so the real number is 90. I don’t know the resulting distribution.

    I did note that the morphological features they list would be of very different importance. Since they’ve been working under the assumption that Dene-Yeniseian is correct, rather than seeking independent evidence, a much more interesting result might have been obtained by comparing which languages have which (combinations of) slots, and cognate elements, in the verbal paradigm.

  9. “Striking correspondences in the structure of verb-forms beats sound correspondences any day.”

    That may be true for the purpose of supporting the hypothesis that the Yeniseian and Dene languages are genetically related, but for purposes of determining degrees of relatedness, or the likelihood of the back-migration hypothesis, as opposed to the out of Asia hypothesis, I question how morphological correspondences can be used effectively. I really don’t know what I’m talking about here, but I think I’m saying the same thing as Y, and maybe Trond, too. At any rate, the NYTimes article seems like a gross distortion.

  10. Trond Engen says:

    I really don’t know what I’m talking about here

    Neither do I! But Y does. Yes, I think we’re more or less on the same page. But I do think that a complex morphology like the Dene-Yeniseian could be used to build a linguistic family tree. I don’t know how to do it, though.

    Linguistic arguments alone, however elaborate, will hardly solve any homeland question, but they can tell what connections to make, and in which order, on the way back in time. This particular result put Yeniseian within a sub-branch of DY, i.e. departing after Tlingit, Eyak and Pacific Athabascan (!) split from Continental Athabascan. Since that’s a geographical split today, it would have to have been retained as such since before the branching off of Yeniseian, I assume they find the Beringian scenario most parsimonious, but also two waves of DY speakers from Asia might explain it. (If we take the result at face value. Which we don’t.)

  11. marie-lucie says:

    Has anyone read the Language Log thread about this? I read it a few days ago, was trying to find it again, but could not, in spite of it being so recent. Perhaps it is hidden under a misleading title? I wanted to look at some of the comments again, especially the one by Sally Thomason who is a well-known historical linguist, notably about the Salishan family and the problem of mixed languages (pidgins, etc).

    Anyway, I find Wade’s article to be a little muddled, but perhaps not as much as expected. Of course the linguistic work being reported on was done by genuine linguists who have both worked on Athabaskan languages (Holton has been in Alaska for quite a few years, and Sicoli was on staff there for a couple of years), not by other kinds of scientists basically ignorant of linguistics.

    I have followed Vajda’s work (from a distance since I am not an Athapaskanist) ever since he presented evidence for a Ket (Yeniseian) – Na-Dene at a conference some years ago, using morphological data (the similar verbal “templates” in both families), and I also attended the 2008 conference in Alaska were he presented further evidence, especially lexical-phonological comparisons (words and their sound correspondences) and proto-language reconstructions. (It should be said that Vajda was not the first person to link the two families: a fair amount of work had been done by Russian linguists, dating from the time when there were several related languages in the Yenisei valley, so Vajda had the benefit of their descriptive and comparative work, and some of those linguists had already noticed resemblances with Na-Dene). At the conference, a number of specialists in Athabaskan and historical linguistics had been invited to attend, and Vajda was highly praised by them, even by people like Michael Krauss who had until then very skeptical of the hypothesis. There was agreement that more work was needed to really flesh out the hypothesis, but that it could be considered very plausible. Of course, not everyone is going to agree, but not that many linguists are specialists in Athabaskan languages, which are (to my mind) fiendishly difficult as well as morphologically quite different from other families in the Americas. I would not take Lyle Campbell’s opinion very seriously: he is a competent enough historicist when dealing with languages which have already been hypothesized to be related, but his reputation is also that of a negativist always ready to shoot down other people’s work.

    Not being an Athabaskanist, I can’t comment on the details of the hypothesis, but the one thing that bothered me was the time depth: the idea that the split between Yeniseian and Dene must have occurred 14,000 or so years abo, even though there were all those similarities between the two families. So the hypothesis that a population ancestral to both peoples spent a long period in an intermediary location, probably “Beringia”, sounds plausible to me. However, I don’t think one can assume that this population split up into two groups as soon as the ice started retreating: even assuming a fast rate of climate warming and sea level raise, it would have taken some time before the landscape changed drastically and all the people had to move away from the now submerged plain, some going West and other East. Also, although the majority of those languages are now found in North America, there is evidence (especially in toponyms) that Yeniseian languages were once spoken over a much wider area of Siberia, so a lot of the evidence which would be needed to fully substantiate a Dene-Yeniseian genetic relationship is probably lost forever.

    As for Haida, which Sapir had classified together with Tlingit and Athapaskan. Krauss wrote something like “As soon as I started to look at Haida, I realized that the language was completely different from Na-Dene. Perhaps Holton & Sicoli included it in their database to try to determine who was right about the language. It is probably relevant that the traditional histories of the Haida tell of their ancestors arriving on the islands and finding there an existing people that they refer to as “the Old Haida”, which might have belonged to the same refugee population as in Beringia, since the islands too are supposed to have been free of ice while most ot the hemisphere was glaciated.

    I cited Sally Thomason above: she is very leery of invoking typological arguments for determining genetic relationships, since she points out that there are many cases of historical changes drastically altering typological features. I think that she may be exaggerating the problem, which may depend on what are called “typological features”: phonological features are almost useless in this respect, since phonology can change quite radically and prevent intelligibility while vocabulary and grammar remain largely intact. Other general features such as presence or absence of gender, case, articles, and many others can indeed change, especially under situations of bilingualism. The most likely marker of relationship would seem to be shared complex morphology (such as the verbal system of Latin, Greek and Sanskrit), but such complexity can end up greatly simplified). All in all, there is no single element that is absolutvely decisive. One has to consider everything!

  12. m-l, the time depth is a really good point. Dene-Yeniseian looks something like an 8000-y. old family: sort of discernable, but difficult. Adding 6000 years to that would make the relationship quite indiscernible. It would also place the proto DY at the same time as Clovis culture, which would be even more surprising.

    So I would modify what I said before: the statistical/typological method is not convincing, but the general idea of Beringia as the dispersal point for Dene-Yeniseian is not plausible either.

  13. Has anyone read the Language Log thread about this?

    I didn’t know there was one! In fact, as I was putting the post together I was thinking grumpily “I wish Language Log would cover this stuff.” I read it every day and I’d think I would have seen it.

  14. “The most likely marker of relationship would seem to be shared complex morphology (such as the verbal system of Latin, Greek and Sanskrit), but such complexity can end up greatly simplified).”

    When I looked at this in 2008, the shared complex morphology was really quite striking–according to Vajda, there is a string of four or five verbal prefixes in both Ket and the Dene languages which indicate various attributes of the verb, not just person and number, but other attributes such as the shape of the object, and they occur in the more or less same sequence on both continents. I’m not competent to judge this myself, but it really did strike me as more than coincidence.

    “I read it every day . . .” Who doesn’t? I couldn’t find a discussion of this on LL, either.

    Thanks for your, as usual, informative comment, M-L!

  15. marie-lucie says:

    Where could I have read that comment by Sally Thomason? I will have to ask her in person.

    Bill, thank you. The shared verbal template was the topic of Vajda’s early presentation which I attended, I don’t remember when but some time in the 1990′s, and I was very impressed. Michael Krauss who also attended was rather dismissive, but at the 2008 conference, by which time Vajda had added more evidence, especially lexical-phonological, he admitted publicly that he had been very doubtful but was now convinced of the link.

    Re Beringia: I realize that I made a mistake about the dates: 14,000 years BP is supposed to be (the start) of deglaciation. I have just read relevant articles on Wikipedia. The one titled “Beringia” is very informative. Among other things it has an “interactive” map of the area showing the geography of Beringia from about 20,000 years ago to the present in about 10 increments that show the changing landscape, as well as other maps. It is obvious that the deglaciation process took several thousand years, as did the widening of the strait to its current size, allowing for plenty of time for a population in Beringia to have travelled back and forth easily, especially in the winter, and also at first in small boats, until the crossing became wider and more difficult. According to the information in the chapter, Siberia was not iced up while North America was, because there was little snowfall in Siberia where the climate was very dry. Consequently, an expanding population in shrinking Beringia would have been much more likely to turn westwards (back to Siberia) than eastwards where they would be blocked by the ice. This would have a bearing on when and how the Dene entered North America, closer to 6,000 years than 14,000 years ago, making the language relationship more explainable.

  16. marie-lucie says:

    p.s. There is updated information on Wikipedia : Dene-Yeniseian languages.

  17. Actually, there are two claims in this paper. One is that the coastal languages (Pacific Coast Athabascan + Eyak + Tlingit) cluster together; the other is that Yeniseian fits in that cluster. I am skeptical that either of these has anything having to do with shared innovations.

  18. Or else, as I noted above, it represents a shared innovation in all coastal/Apachean languages.

  19. “Rather, geneticists say, these ancestors must have lived in isolation for some 15,000 years to accumulate the amount of DNA mutations now seen specifically in Native Americans. [...]”

    If so, why in Beringia? why not in the Americas? why are people so determined to hold to the Clovis-first hypothesis despite earlier dates from Pedra Furada and Meadowcroft Rockshelter?

  20. Marja, I suspect it has something to do with the timing of the glaciations. The assumption seems to be the areas further east (North America) were still under ice. Beringia and the areas westward were free of ice.

    Clovis first doesn’t come into this. That has to do with “Amerindian” settlement, long before Na-Dene or Dene-Yeneseian was a thing in NA.

  21. J. W. Brewer says:

    It is remarkably convenient (even if true . . .) to be able to say that “well, the place we think they were located for thousands and thousands of years is now under the ocean, so that’s why we don’t have any archeological evidence confirming the claim.”

  22. “Clovis first doesn’t come into this. That has to do with “Amerindian” settlement, long before Na-Dene or Dene-Yeneseian was a thing in NA.”

    Pretty sure it does:

    “Rather, geneticists say, these ancestors must have lived in isolation for some 15,000 years to accumulate the amount of DNA mutations now seen specifically in Native Americans.”

  23. why are people so determined to hold to the Clovis-first hypothesis despite earlier dates from Pedra Furada and Meadowcroft Rockshelter?

    The NYT yesterday ran an article titled Discoveries Challenge Beliefs on Humans’ Arrival in the Americas. No mention of matters linguistic though some of its links may.

    A thought struck me while reading the piece: Maybe people from Asia migrated to the Americas twice, once say 25,000 years ago, not thriving and and dying off some time later, and another migration or series of migrations at times that fit with the Clovis theory. Perhaps something like the pre-Inuit Dorset culture writ large.

Trackbacks

  1. […] Hat touches upon the linguistic controversies surrounding migrations between Siberia and North […]

Speak Your Mind

*