On Tocharian origins

Where did the Tocharians originate from? J.P. Mallory's recent talk has been somewhat of an eye-opener for me, as Prof. Mallory brought to my attention two important issues:

1. The lack of a clear connection between Afanasyevo and the Tarim Basin.
2. The existence (in Tocharian) of a rich agricultural IE terminology related to cereals, as well as the domesticated pig, which cannot be easily explained if Tocharians arrived in Xinjiang from the steppes to the north, and, ultimately from eastern Europe.

To begin with, I want to point out an important issue: we cannot assume that the earliest Caucasoids of Xinjiang, including some of the famous early Tarim mummies were Tocharian speaking. There are several arguments why this is so:

1. Tocharian is first attested in the 8th c. AD, that is, about 3 thousand years after the earliest detected Caucasoids in the region
2. There has been a shift in the region from Tocharian and eastern Iranian languages to Turkic over the last thousand years or so. Why assume linguistic continuity in the preceding three thousand?
3. Indeed, there has been linguistic shift throughout other regions of Eurasia in shorter timespans, such as the spread of Slavic across most of eastern Europe, the virtual extinction of Celtic in most of western Europe, the replacement of multiple languages by Arabic in the Near East, and so on. Linguistic continuity does not seem to be an appropriate default position in the absence of direct evidence.
4. The earliest Caucasoids of the Tarim were already substantially mixed with Mongoloids at least in their mtDNA. This reduces our confidence that they spoke an Indo-European language, as there is a pattern of Caucasoid patrilineages combined with Mongoloid mtDNA in present-day non-IE South Siberians
5. Indeed, the current Turkic Uyghurs, who are closer (temporally) to the Tocharians than the early Bronze Age Caucasoids have a rich assortment of Caucasoid Y-chromosome haplogroups, whereas the early Bronze Age ones seem to have belonged uniformly to R1a1. What languages were spoken by the non-R1a1 Caucasoids who arrived in the Tarim prior to the Turkification of the region?

To summarize the first part of the argument: the early population of the Tarim does not have clear steppe connections, it may not have been Indo-European speaking, and even if it were, it did not necessarily speak the same language as the later Tocharians. Moreover, the Tocharian language has a vocabulary without clear steppe associations, but with rich agricultural ones.

In search of the Tocharians

We may discover the origin of the Tocharians by a careful sorting of Y-chromosome lineages in the present-day Uyghur population of Xinjiang that is assumed to have absorbed the pre-Turkic inhabitants of the region:

1. Remove all east Eurasian lineages that are likely to be associated with the Xiongnu, Mongols, or Uyghur
2. Remove all west Eurasian lineages that can be explained from a non-Tocharian source (such as Iranians, or various Silk Road outliers)
3. See if anything is left

A recent paper by Zhong et al. provides rich data on Uyghurs that can be used to carry out this program.

The phylogeographic analysis of these lineages does leave some candidates:

1. Haplogroup D can be excluded as Mongolian/Tibetan
2. Haplogroup E can be excluded as Mediterranean/African
3. Haplogroup C can be excluded as Altaic/South Asian (C5)
4. Haplogroup G2a* (West Asian) does not seem to have an important presence (3 samples)
5. Haplogroup H can be excluded as South Asian
6. Haplogroup I can be excluded as a European outlier (1 sample)
7. Haplogroup J*(xJ2) can be excluded as NE Caucasian/Semitic with small presence (2 samples)
8. Haplogroup NO; haplogroup N has been founded in a Xiongnu context, so it is likely intrusive; O is East Eurasian
9. Haplogroup Q is also associated with Xiongnu nomads from Pengyang

This analysis leaves four candidates: J2-M172, R1a1a-M17, R1b-M343, and L-M20.

We can exclude L-M20 because its overall low frequency in most populations makes it difficult, at present, to make a definitive pronouncement on its origin, except perhaps for its Indian L1 clade which is absent here.

J2, present in both its J2a and J2b subclades here at substantial frequencies has an origin in West Asia, as well as a substantial presence among Indo-Iranian speakers. While it is possible (indeed likely, in my opinion) to have been present among the Tocharians, we cannot exclude the possibility that it represents either a specifically Iranian influence, or even something earlier than both.

R1a1a is present in both the steppe, as well as South Asia and West Asia. Its high frequency among some Indo-Iranian populations also makes it difficult to ascribe a specifically Tocharian origin to it.

This leaves only R1b-M343 as a candidate. Have we found a genuine Tocharian genetic signature?

The West Asian roots of R-M343 (?)

R-M343 and its main R-M269 clade are in a sense exasperating: the combination of their widespread distribution from Africa, the Atlantic, to the depths of Inner Asia, combined with their apparent Y-STR-estimated youth make it nearly impossible to associate them with a specific archaeological or historical phenomenon.

Where could R-M269 have come from? It was not present, as far as we can tell, in early Bronze Age Xinjiang, and neither has it been detected in south Siberians. The steppe/"northern" route seems out.

A southern route, from the Indian subcontinent also seems out, as despite its ubiquity elsewhere in Eurasia, it seems to have (mostly) skipped both India and (to an extent) Pakistan.

An indigenous origin seems highly unparsimonious, as it would require that it trek all the way to the Atlantic, but make hardly an impact in either East Asia or South Asia.

As far as I can tell, the only explanation for the presence of R-M343 in Xinjiang is West Asia, or at least Central Asia west of the Tarim. There it can be found at a high frequency in Armenians, Turks, north Iranians, and Lezgins among others. And, unlike both J2 and R1a1a, R-M343 does not seem to be Indo-Iranian (due to its absence in India).

Gamkrelidze and Ivanov cited W. N. Henning to the effect that the ancestors of the Tocharians could be identified with the Gutians from the Zagros, a people that attacked the Sumerians and founded a dynasty. As usual, I don't presume to know the linguistic evidence for this, but this hypothesis would place the ancestors of the Tocharians in the "right spot": virtually all of their Caucasoid Y-chromosome gene pool could be explained with an origin in north Iran.

A model of Tocharian origins

The model of Tocharian origins I present is simplicity itself:

First, Tocharians are descended from a group of farmers that moved east of the PIE homeland and settled on the Zagros and beyond, south of the Caspian sea.

Second, their trek to the Tarim was a simple west-to-east movement along what would later become the Silk Road, beyond the Taklamakan desert and into the Tarim basin. There they must've mixed with the early pre-IE mixed Caucasoid/Mongoloid population of the early Bronze Age. The desert probably sheltered them, to an extent, from encroachments by the Iranians.

An open question remains: were the Tocharians late fugitives who were pushed out of their ancestral homelands by the emergence of the Iranians and entered the Tarim late? Or were they established there fairly early and were the historical Tocharians are the eastern relics of a once great people that was not Iranized unlike most of the people of Central Asia?

Autosomal evidence

The fine-scale analysis of the Dodecad project on a sample of 10 Uyghurs provides some additional evidence:

The Uyghurs seem to lack the Southwest Asian component that is ubuiquitous in most of West Asia today, and may have, in large part, expanded with the more recent spread of Semitic languages. They are similar, in that respect with South Asians, suggesting that neither the spread of Islam to the east nor the cosmopolitanism of the Silk Road were enough to bring this component to the region. Hence, the plethora of Caucasoid Y-haplogroups in the region cannot be attributed to recent arrivals.

The absence of specific South European components in them also suggests that the opinion of some linguists and archaeologists that would see the Tocharians related to Celts and moving from deep within Europe, or even Western Europe to the Tarim, are unlikely; the south European component is ubuiquitous in Europe, and the Uyghurs, like South Asians, seem to lack it entirely.

Their Caucasoid components are primarily West Asian and North European. Projecting them on the East Eurasian/West Asian/North European PCA plot (left), it is clear that they are more West Asian than North European, a result that is in agreement with their ADMIXTURE results.

Notice also how the North European/West Asian ratio is reversed for the more northern-latitude Uralic/Altaic speakers (Selkups, Dolgans, etc.).

Of course, the results should be interpreted with caution, but they seem perfectly in agreement with the model presented here:

• the Uyghurs are partly Mongoloid both because they may carry the legacy of the ancient mixed population of the Tarim, and also because of their more recent Turkic/Xiongnu associations.
• with respect to their Caucasoid components, they are mainly West Asian (with the West Asian component also being primary in South Asia), but somewhat shifted to the north due to their absorption of mixed Northern Caucasoid/Mongoloid peoples from the steppelands.

Conclusion

The mystery of the Tocharians may be that there is no mystery. The Tocharians are revealed to have been just another West Asian branch of the Indo-European family that, unlike most of its cousins, went east, absorbed Northern Caucasoid, Mongoloid, and South Asian population elements, emerged long enough in history to leave us a written record of their presence, before succumbing to the Xiongnu and the Mongols.

Thankfully, by combining the remnants of their language, and fragments of their DNA in their descendants, we are able to reconstruct the history of this, once forgotten people