Of sensible sematics

One of my met peeves is the confusion which some ethno-linguistic terms can cause. For example, the fact that there were Iranian language speakers on the plains of Ukraine ~2,000 years ago naturally indicates to people that Scythian nomads issued out of Iran northwards. Similarly, the existence of Indo-Aryan Mitanni in what is today Syria also suggests to people that there was a migration of Indians which traversed much of West Asia in a drive toward the Mediterranean from the Indus. Of course our perception of the center of gravity of these ethno-linguistic groups today is a function of historical contingency. If we didn't know much more about Antique and Medieval European history we might posit that the Celtic Galatians of ancient Anatolia were originally from Ireland, based on the contemporary distribution of Celtic languages!

This issue is now cropping up South Asian archaeogenetics. In my opinion the paper Reconstructing Indian population history is probably the most important contribution to the field in a generation. The authors explain technically why a "South Asian" ancestral component falls out of ancestry inference algorithms at the heart of ADMIXTURE, STRUCTURE, or frappe. In short, when you have a population which is a hybrid, but where the hybridization event is very distant in the past, recombination breaks up the signatures of that event (a decay of the linkage disequilibrium between two putative ancestral populations). Additionally, in the Indian case there doesn't seem to be a "pure" population of one of the two ancestral groups, what they termed "Ancestral South Indians" (ASI). The closest reference they found were Onge Andaman Islanders, whose last common ancestors with ASI was on the order of tens of thousands of years in the past. They do have excellent proxies for the other population, "Ancestral North Indians" (ANI). Compared to ASI all West Eurasians can be used as reasonable proxies for ANI.

But are the terms ANI and ASI really satisfying? It looks like all Indo-Aryan, Dravidian, and Austro-Asiatic groups are at least in part combinations of ANI and ASI. It also seems that the Iranian speaking Pathans of Pakistan are part of the the ANI-ASI cline. The Burusho, Brahui, Kalash, and Baloch are more marginal cases. In Reconstructing Indian population history the authors discarded these groups from their analyses because they did not seem to be modeled so easily by a two-way admixture (note that they admit that a two-way admiture is oversimplified for other South Asian populations, but on a coarse scale it suffices). Some genome bloggers have attempted to extrapolate ANI/ASI ratios though, and in general find that these groups have less ASI than the Pathans.

This is particularly intriguing for the Brahui, who are themselves a Dravidian population. Prior to the publication of Reconstructing Indian population history I had assumed that Dravidian dialects were the ur-languages of the subcontinent, at least back to the Paleolithic. But the findings in the paper made me reconsider. It was clear from exchanges with one of the authors that the admixture event had occurred rather distantly in the past. It was not difficult, for example, to come up with estimates for Uygurs as being the products of an admixture event ~2,000 years in the past. So it seems possible that the very low boundary of the admixture event was ~3-4,000 years in the past. Mostly likely it was deeper in time than that. The fact that the even South Indian tribal populations were ~40% ANI is also suggestive of an event that had time and opportunity to distribute itself relatively evenly across the Indian subcontinent. The widespread distribution of ANI ancestry amongst these groups is suggestive to me of the likelihood that the ANI may not have been Indo-Aryans. It follows then that the ANI may have been the Dravidians, and that Dravidian language are therefore not ur-languages o South Asia, since ANI was likely intrusive.

More recently the results generated by genome bloggers have convinced me that we seriously need to posit that the Dravidians post-date the ANI! This would mean that the languages of the ASI and the ANI are lost to us, overlain by later arrivals. Because of this I think perhaps it is time to give more informative labels to the ASI and ANI. For the ASI I think the answer is easy: Paleolithic South Eurasians (PSE). It seems clear that some element of PSE-like ancestry can be found across Southeast Asia at low levels. The PSE themselves hare a closer affinity to East Eurasian and Oceanian populations than West Eurasian ones, even if they are at a remove. And of all the world's populations those of the Andaman Islands seem to be closest to the PSE, and that last common ancestral population across these two groups existed on the order of ~20,000 years in the past. This indicates that the ASI do have Paleolithic roots in southern Eurasia (whether they are a descendants of the first arrivals from Africa is a different issue).

The ANI are more difficult. They are not too exotic when you come down to it. At this point for various reasons I lean toward Dienekes' proposition that the ANI are closer in affinity to trans-Caucasian and West Asian populations than any other West Eurasian groups. Their distinctiveness from these other groups has to do with their intrusion into South Asia. And, they were arguably the first mass arrivals to South Asia from the west since the Out of Africa event. I believe that they arrived as farmers, so I think it is more informative to term them Neolithic West Eurasians (NWE). More awkwardly, Initial West Eurasians in South Asia (IWEiSA) would also be accurate.

I believe it is important to create some accurate non-generic terms for these groups because it isn't useful for us to continue to impute ethno-linguistic identities to these abstract ancestral components. The ASI/PSE were not Andaman Islanders, anymore that Japanese and Siberians are the same because they share ancestors ~20-30,000 years before the present. And we don't know if the ANI/NWE were Dravidians or Indo-Aryans, or something altogether different. My own assessment is that they probably predated the Dravidians and Indo-Aryans, but my confidence is low in that claim, and I believe they are more likely to have been Dravidians than Indo-Aryans.

Overall, the main issue is that we may need to beyond the vague sketch outlined in the perfunctory first chapter of Indian historical surveys where we read about a succession of peoples, Austro-Asiatic Mundas, then Dravidians, then Indo-Aryans. These three groups may all be relatively latecomers, preceded by the groups which we have been terming ASI and ANI.


  1. If both Dravidian and Austro-Asiatic came from outside peninsular India, the biggest question that is yet to be answered is what the original ASI folk in mainland India spoke. The same question holds good for the original ANI language (assuming Dravidians aren't the original ANI), although Burushaski might potentially be the lost language.

    • I would speculate that Nahali, Kusunda, and potentially Vedda may be remnants of the ANI languages. There's a substantial substrate remaining in modern South Asian languages, including place names, names for local crops/fauna, etc. Broadly they would have spoken something close to Elamite -- as it is their script is similar to what they used.

      ASI languages are lost completely. But we don't have a good sense of what indigenous hunter gatherer groups spoke prior to the advent of agriculture in many parts of the world.

      • How does Masica's "Language X fit into the scenario. What relation does it have to language(s) of the Indus, if any?

  2. The other thing I'd add is that Dravidians may have entered South Asia prior to the Aryans. At Pirak in Baluchistan you have a culture c. ~1800-1700 bc with terracotta horses and camels -- the first depictions as far as I know of those animals in India. The assumption has been that this reflected some Aryan influence, but that seems far too early.

    Also, in Chanhu-daro, a post-Harappan settlement in the Indus Valley, Allchin 1995 writes:

    "The absence of Harappan seals, the symbols of the Harappan style and, presumably, of Harappan political and economic power, and their displacement by these foreign seal types, which too were doubtless the symbols of a group with a different identity, must indicate that a new power was dominant at Chanu-daro and by inference in the middle and lower Indus region."

    Again, this is post-Harappan collapse, and pre-Aryan. *Someone* moved in.

  3. Do we have any proof of any kind of AMH living in any part of Europe, West Asia, Central Asia, or South Asia between ~90000ybp and ~45000ybp?
    If not, how do we explain that gap.

    It appears that AMH exited Africa ~100000 ybp, had a very temporary stay in West Asia (mixed a bit but could not compete with the Neanderthal), and moved to regions where the Neanderthal did not exist. This could explain the 45000year gap.

    • There are hominin tools in South India prior to and after the Toba ash layer, ca. 75,000 ybp that seem to show continuity with each other. There is a credible argument that both the before and after layers should be attributed to AMHs. This is probably the most credible potential AMH find outside the Levant more than 60,000 ybp.

      There have been speculatioons in the academic literature that Toba contributed to the population genetic divide between SE Asia and S Asia that is more or less at the Burma border which is arguably more stark than one would expect given the modest biogeographic barriers present there. The gist of that speculation goes that intermediate populations in the border area were wiped out by Toba and repopulated from further West (more or less Pakistan and Iran) and further East (basically further East within SE Asia) respectively. Hence, the population genetic divide there for AMHs is steeper than it was for AMHs in the Middle Paleolithic.

      The hominin species that should be associated with new Arabian tool finds in that time period is not well established. Both Neanderthal and AMH are possible affinities and at that point the differences in the tools used by each were subtle. Paleoclimate data suggests that the best evidence for AMH in that region in that time period should be underneath the Persian Gulf right now, which would have been a natural refugium.

      Evidence of AMH in the Levant is arguably up to 75,000 ybp, rather than 90,000. The gap in the AMH presence in the Levant is sometimes described as one lasting about 25,000 years from 75,000 ybp to 50,000 ybp, and usually explained as either a localized extinction or a retreat to some (unspecified) refugium. The Persian Gulf, the Caucasus Mountains, Anatolia, Persia, South Asia, the Nile Valley, or East Africa have been variously advanced as possible refugiums without much convincing proof one way or the other.

      There are apparently no AMH in Europe much pre-45,000 ybp.

      There is one site in China that arguably contains AMH, but there is tens of thousands of years gap before the next putatively AMH example (or for that matter trace of any hominin) appears in the record. This absence of evidence could be due to an absence of digs in the region, or could be due to actually absence of AMHs.

      Papuan and Aboroginal Australians might be older than 45,000 ybp, but probably not by more than a dozen millenia. Flores seems to show no AMH until about 18,000 ybp.

  4. Do we have any proof of any kind of AMH living in any part of Europe, West Asia, Central Asia, or South Asia between ~90000ybp and ~45000ybp?
    If not, how do we explain that gap.

    some of the australian dates do predate 45 K. they're debated, but i think we need to shift toward more probable now that AMH origins are being pushed back.

  5. though i agree that there is a "gap" fwiw or some sort.

  6. Nirjhar mukhopadhyay

    Dear Rajib i think this reasearch will give you some help: http://genomebiology.com/2010/11/11/R113Dear Rajib i think this reasearch will give you some help: http://genomebiology.com/2010/11/11/R113

  7. Nirjhar mukhopadhyay

    Hmmm. I didnt write that twice.

  8. Nirjhar mukhopadhyay

    And what do you came up with?And what do you came up with?