Growing weary of archaeogenetics

[WeepingElf]

As you may know, I have been quite interested in archaeogenetics, especially as a tool for tracing past human migrations and thus the spread of language groups. But now I have grown quite weary of it. One reason is that have been almost constantly banging my head against walls, be it paywalls that prevented me from getting seemingly interesting papers into my hands, or walls of unintelligible terminology and concepts.

But often, despite genetics being allegedly a "hard science", different research teams (anyone who has checked out a significant number of archaeogenetics papers will have noticed that they usually have dozens of authors listed) often come to different, mutually exclusive conclusions about the people from the same time and place. Why is that so? The problem, I think, is a combination of the non-existence of pure genetic stocks in human history and the small sample sizes. Old genomes are hard to come by - you need exceptionally well-preserved bones - and when a research team has analyzed a dozen old genomes, that's quite something. With such small samples, the results may be easily thrown off the scent by chance!

And then, there are interesting ancient cultures who practiced cremation, thereby literally burning big holes into the archaeogenetic record - nobody can extract ancient DNA from ashes, and nobody ever will. Cultures such as the Hittites, who would be especially interesting (to me) whom they were related to. Such holes are frustrating if you want to know what really was going on by means of archaeogenetics.

Also, the results from archaeogenetics can be just as easily abused for racist purposes as the "callipers science" of the early 20th century. I don't know what kind of discussions this new science has inspired in the Telegram (or whatever) groups of the extreme right as I do not follow those groups, but I can easily imagine that those people construe the archaeogenetic results as a "proof" that the "Aryan race" really existed!

And finally, of course, genes don't speak languages. Sure, most people get their native language from the same people they have their DNA from, but language shifts are common enough to foil linguistic conclusions from archaeogenetic results. It is often said that Haak et al. "proved" in their famous 2015 article that Indo-European was spread by Yamnaya settlers around 3000 BC; but what they really showed was just that there was a massive migration from the Pontic steppe into much of Europe at that time, and the "proof" was indirect and not perfectly certain. The migration they found gives a probably earliest possible date for the spread of IE because such a migration would have marginalized whatever languages were spoken in Europe before, and it lies in the bracket of earlier estimations on the time of the spread attained before by linguistic means.

Thus, I think that I would probably fare better if I took much less heed of this mumbo-jumbo than I have done in the past few years. My interest lies chiefly in the languages of prehistoric Europe, and genes don't speak them!

[Moose-tache]

There's no pseudoscience like confident pseudoscience.

But this stuff is barely ten years old, at least as it pertains to sweeping generalizations about migration, language, culture, etc. I assume things will get better once we have more data, more researchers, and (perhaps most importantly) more data and researchers not from a narrow band of Western countries.

[WeepingElf]

I wouldn't say that archaeogenetics is pseudoscience as, for instance, astrology or ufology are, but I have seen contradictory claims about the same archaeological cultures from different research teams, and contradictory interpretations from the same data by different authors. You know, the claims of the sort "We have found Y-DNA haplogroup R1b in three Yamnaya kurgans, so the Yamnaya people probably were overwhelmingly R1b" - what we can say here is that there were three Yamnaya chiefs with the Y-DNA haplogroup R1b; which says little about the Yamanya upper class in general and nothing about Yamnaya commoners. Things probably will get better when, as you say, we'll have more data and more researchers from different backgrounds, but right now it is all in flux and uncertain, and building on such foundations is perilous. And after all, genes don't speak languages.

[Skookum]

I find this is a problem when it comes to arguments about the settlement of the Americas as well, where it seems extremely difficult (or maybe impossible) to correlate the lack of genetic diversity with the overwhelming linguistic diversity.

[vegfarandi]

I've been interested to follow the non-archeo-genetics of 23 and Me, Ancestry.com and the like. As an Icelander, I expected to get a precise result when I did it because I happen to know the Icelandic genome is among the best studied in the world thanks to DeCode genetics being headquartered in Iceland and it using the small population as a sample for a while. But my results didn't have that data apparently and pinned me as something like 59% Norway, 25% Sweden/Denmark, 13% Ireland, 3% England & Northwestern Europe. Well… that's certainly true, historically, but that precise mixture should yield fairly confident results that I'm in fact Icelandic, not to mention the number of fairly unique Icelandic markers that exist as well that I'm sure I have some of. So I think ultimately, the field of practical genetics is still quite new and sorting through the data is probably an unbelievably massive undertaking needs to occur on a global scale. And in order to yield results where we start to see revolutions in applications for history and the like, it will require careful correlation with personal family histories which is even hairier.

EDIT: I just checked and it does now say that I'm part of the "Iceland DNA Community" so as they've gathered data, these things have updated a bit.

[Moose-tache]

I find this is a problem when it comes to arguments about the settlement of the Americas as well, where it seems extremely difficult (or maybe impossible) to correlate the lack of genetic diversity with the overwhelming linguistic diversity.

Why? Even in the lowest genetic diversity scenario, most Native Americans began to diverge from one another twelve thousand years ago, at least twice the length of time we would need to explain the observed linguistic diversity.

[Skookum]

I find this is a problem when it comes to arguments about the settlement of the Americas as well, where it seems extremely difficult (or maybe impossible) to correlate the lack of genetic diversity with the overwhelming linguistic diversity.

Why? Even in the lowest genetic diversity scenario, most Native Americans began to diverge from one another twelve thousand years ago, at least twice the length of time we would need to explain the observed linguistic diversity.

How do you know that? Sincere question and not trying to be sarcastic here, but that sounds to me like something that is basically unknowable.

[Moose-tache]

I find this is a problem when it comes to arguments about the settlement of the Americas as well, where it seems extremely difficult (or maybe impossible) to correlate the lack of genetic diversity with the overwhelming linguistic diversity.

Why? Even in the lowest genetic diversity scenario, most Native Americans began to diverge from one another twelve thousand years ago, at least twice the length of time we would need to explain the observed linguistic diversity.

How do you know that? Sincere question and not trying to be sarcastic here, but that sounds to me like something that is basically unknowable.

Well, sure, it relies on some inductive reasoning that is inherently shaky. But the logic goes like this:
1a) Very few proto-languages have been reconstructed that are more than five thousand years old.
1b) Very few language families can even be confidently proposed, without reconstruction, that are more than ten thousand years old.
2a) The languages of South and most of North America show no reconstructable proto language.
2b) The languages of South and most of North America can be lumped into families, but no super family.

Conclusion 1a-2a tells us that the peopling of the Americas took place at least five thousand years ago. Conclusion 1b-2b tells us it happened at least ten thousand years ago. So maybe not "at least twice as long," but certainly twelve thousand years is enough time to account for the diversity.

Obviously, I am not presenting this as a hard argument that must be correct. But if, say, Europe was populated just once twelve thousand years ago, we would expect to see exactly what we have in the Americas: some vaguely defined families, but no reconstructable proto-language. Beyond that point, there is nothing we can say. in other words, a place peopled ten thousand years ago would look the same as a place peopled fifty thousand years ago. Once the reconstructable proto-language is out of sight, no chronological linguistic evidence will be available to us (e.g. would anyone guess that Australia is the product of fifty thousand years of diversification if they only had linguistic evidence to go on? Sure there have been subsequent migrations that hide some of the diversity, but that also applies to the Americas).

[Richard W]

2a) The languages of South and most of North America show no reconstructable proto language.
2b) The languages of South and most of North America can be lumped into families, but no super family.

I think (2a) and (2b) are axioms, or at least not linguistically based facts. When attempts to relate families are shot down on the basis that common words are 'Pan-Americanisms', that smells to me of dismissing a relationship between Latin and Sanskrit because Latin shares features with Brythonic. Reconstruction has to work with what is tractable - some probable relatives may have to wait for a parent to be reconstructed.

(e.g. would anyone guess that Australia is the product of fifty thousand years of diversification if they only had linguistic evidence to go on? Sure there have been subsequent migrations that hide some of the diversity, but that also applies to the Americas).

I doubt that internally Australian is the product of fifty thousand years of diversification, though it has probably had that long to diverge from other parts of the world.

[Moose-tache]

2a) The languages of South and most of North America show no reconstructable proto language.
2b) The languages of South and most of North America can be lumped into families, but no super family.

I think (2a) and (2b) are axioms, or at least not linguistically based facts. When attempts to relate families are shot down on the basis that common words are 'Pan-Americanisms', that smells to me of dismissing a relationship between Latin and Sanskrit because Latin shares features with Brythonic. Reconstruction has to work with what is tractable - some probable relatives may have to wait for a parent to be reconstructed.

Uh... I'm confused. If we can create a Native American super family, wouldn't that make the linguistic diversity of the Americas less striking, rather than more? The original argument by Skookum was that a conservative view of New World migrations, with its few migration events and subsequent lack of genetic diversity, would not be enough to explain the current linguistic diversity. But if there is one language family for the whole lot of them, then if anything their linguistic diversity would be shockingly low! for the length of time they've been living there.

For the record, I'm not making any claims about how related NA languages are, or should be. I am merely saying that, as a very rough rule of thumb, the amount of linguistic diversity in the New World is not incompatible with a scenario in which the population of most of the New World is derived from a single migration twelve thousand years ago. Did it happen that way? I don't know. I'm just saying it doesn't jump out as odd to have that much linguistic diversity paired with that much genetic diversity, because after twelve thousand years we wouldn't expect to see easily reconstructed families no matter how they arrived.

As for that stuff about Australia, I think you misunderstood me again. It was merely an illustration to show that linguistic diversity doesn't necessarily stack indefinitely. You can't expect a place to tell you exactly how long people have been there based on their linguistic diversity, unless that migration took place less than... I don't know, five to ten thousand years, maybe?

[Skookum]

I guess I just meant that I'm not sure how reliable the dating of protolanguages/linguistic diversification is, unless we're talking about protolanguages that have strong evidence of being spoken by a particular archaeological culture (like maybe Yamnaya and PIE). But you could really make the same argument about "genetic diversity", that our understanding of the historical linguistics of the Americas is still so poor that it doesn't really sense to speculate on the number of language stocks present as a way of quantifying diversity. Unfortunately I don't think anyone is really doing that research anymore... (I don't want to derail this thread too much into a discussion of indigenous languages of the Americas, although I have been considering making that thread.)

I don't think a conservative timeline is necessarily incompatible with the apparently genetic diversity observed, just that its odd that the genetic and linguistic facts don't line up the way you'd expect. And I do find it interesting to speculate about these topics, even if I doubt how precise/scientific anyone can be!

[Richard W]

But you could really make the same argument about "genetic diversity", that our understanding of the historical linguistics of the Americas is still so poor that it doesn't really sense to speculate on the number of language stocks present as a way of quantifying diversity. Unfortunately I don't think anyone is really doing that research anymore...

Claims 2a and 2b seem to be invalidated by the strong discouragement of attempts. On the other hand, your thesis that there has been long enough for great diversity to occur feels plausible - unless one can argue that there is an inherited Eurasiatic feel to modern Eurasiatic languages.

[WeepingElf]

I have a gut feeling that the Americas are linguistically much more diverse than northern Eurasia. In northern Eurasia, we have about a dozen language families many of which show some similarities in their morphology, which is why such entities as "Nostratic", Eurasiatic" or "Mitian" have been proposed, which look more substantial than Greenberg's "Amerind" (which, atop on lacking linguistic evidence to support it, rests on the "Clovis First" theory which since then has been falsified).

[So Haleza Grise]

(e.g. would anyone guess that Australia is the product of fifty thousand years of diversification if they only had linguistic evidence to go on? Sure there have been subsequent migrations that hide some of the diversity, but that also applies to the Americas).

I doubt that internally Australian is the product of fifty thousand years of diversification, though it has probably had that long to diverge from other parts of the world.

There's been more work in recent times on Proto-Australian: It's highly likely that Pama-Nyungan can be brought closer to at least some of the "northern" languages, and that they can be brought more closely to each other. It's very hard to make guesses about the time depth of Australian families because it's not that easy to identify lots of cognates. I personally doubt the language map that we see is the result of ~ 50-60,000 years of diversification. Human settlement may be that old, but the languages themselves must be more recent, is my feeling.

A big mystery to my mind is why P-N managed to spread over so much of the continent while having a time depth that seems relatively shallow. Some of the interesting suggestion are repopulation after a big climate shift, or the export of a particular kinship system. Maybe what happened with PN has happened many times over history - one family sweeps across the whole continent starting from the north and then diversifies.

[Travis B.]

(e.g. would anyone guess that Australia is the product of fifty thousand years of diversification if they only had linguistic evidence to go on? Sure there have been subsequent migrations that hide some of the diversity, but that also applies to the Americas).

I doubt that internally Australian is the product of fifty thousand years of diversification, though it has probably had that long to diverge from other parts of the world.

There's been more work in recent times on Proto-Australian: It's highly likely that Pama-Nyungan can be brought closer to at least some of the "northern" languages, and that they can be brought more closely to each other. It's very hard to make guesses about the time depth of Australian families because it's not that easy to identify lots of cognates. I personally doubt the language map that we see is the result of ~ 50-60,000 years of diversification. Human settlement may be that old, but the languages themselves must be more recent, is my feeling.

A big mystery to my mind is why P-N managed to spread over so much of the continent while having a time depth that seems relatively shallow. Some of the interesting suggestion are repopulation after a big climate shift, or the export of a particular kinship system. Maybe what happened with PN has happened many times over history - one family sweeps across the whole continent starting from the north and then diversifies.

Remember that the protolanguages of many of the major European languages groups are not that old at all - they all date to the first centuries AD, or in the case of Common Slavic, even a bit later than that. The languages before then in many cases have been nearly completely erased. We are only able to reconstruct PIE accurately at a significant time depth because there are a number of relatively well-attested early IE languages from separate branches. But in the case of Australian languages we do not have this luxury, so what was spoken before the spread of P-N is simply lost, at what we know about older Australian languages is limited to our ability to reconstruct them from the languages that managed to be attested in modern times.

[WeepingElf]

Ah, Australian languages. I dimly remember reading somewhere - years ago - that a genetic study had found evidence of an immigration from India 4,500 years ago, which apparently coincides with some cultural innovations in Australia and the introduction of the dingo. That may help answering the question why Australian consonant inventories are so similar to Dravidian ones, though otherwise, Dravidian and Australian languages are about as different as they could be. Also, I don't know what the current state of research in Australian archaeogenetics is; the study I am referring to was AFAIK based on recent DNA, and it would not be the first time such a result was overthrown by data from actual ancient DNA. At any rate, though, if Australian is a family, it is much older than Dravidian and not closely related to it because otherwise, linguists would have figured it out by now.

[hwhatting]

I dimly remember reading somewhere - years ago - that a genetic study had found evidence of an immigration from India 4,500 years ago, which apparently coincides with some cultural innovations in Australia and the introduction of the dingo. That may help answering the question why Australian consonant inventories are so similar to Dravidian ones, though otherwise, Dravidian and Australian languages are about as different as they could be. Also, I don't know what the current state of research in Australian archaeogenetics is; the study I am referring to was AFAIK based on recent DNA, and it would not be the first time such a result was overthrown by data from actual ancient DNA. At any rate, though, if Australian is a family, it is much older than Dravidian and not closely related to it because otherwise, linguists would have figured it out by now.

It is quite possible that Dravidian would be neither here nor there. It seems that Dravidian immigrated to India 3000 - 4000 years ago (s. §1.11 in this paper), through Southern Iran. If the Indian immigration to Australia is confirmed, and if we take the higher bound of the 3000 - 4000 year range, the immigrants to Australia might be people displaced by Dravidian Immigration, and the supposed similarity in the consonant inventories could be due to a substrate influence of the pre-Dravidian population on Dravidian. But all this is highly speculative.

[WeepingElf]

Interesting. It is indeed anyone's guess what language(s) the Indus Valley civilization spoke; while they have left some writing, it is undeciphered, and some scholars even have given up, claiming that those pesky symbols aren't writing at all. While many scholars think that the language in question was most likely Dravidian, we can't be sure. Some scholars maintain that Dravidian is related to the Elamite language of Bronze Age Southwestern Iran, whatever that may be worth - Elamite does not look much like Dravidian, and the majority of relevant scholars remain skeptical. I know close to nothing about the relevant facts and thus abstain from this discussion.

[zompist]

It is quite possible that Dravidian would be neither here nor there. It seems that Dravidian immigrated to India 3000 - 4000 years ago (s. §1.11 in this paper), through Southern Iran.

That's a new one for me. 3000 years ago, by his own account, means that the Dravidians came to India after the Arya, and somehow leapfrogged past them to end up in the south. Not impossible, but surely two invasions (or tricklings) is a bit much, and why would the later invasion go the farthest?

(The usual story is that the Dravidians occupied most of India-- which is why a Dravidian language, Brahui, is still found west of the Indus-- and were displaced by the Arya.)

His section on Meluhha is interesting but not very convincing; it's worth noting that some scholars of the ancient Middle East don't even accept that Meluhha is the Indus Valley. It's even less certain that the name is actually the self-designation of the people involved, and the relationship to mleccha seems like a stretch. As I'm sure Witzel is aware, Skt. ch is a palatal stop, not a velar fricative, so we're dealing with just two similar consonants. Might as well relate them to PIE *melid-- the honey people.

[WeepingElf]

I concur with you on all points, Mark. Michael Witzel is an accomplished Indologist and not the usual kind of self-published crackpot you find on the Web so often, but even the most accomplished scholars may have bad ideas within their specialty (speculation is part of the trade in scholarship; a scholar who never speculates probably won't achieve much). It is quite obvious that the Dravidians were in India before the Aryans, and it seems likely that the Indus Valley civilization spoke Dravidian (perhaps Proto-Dravidian proper) and Brahui is a residue thereof.