Razib Khan One-stop-shopping for all of my content

March 11, 2018

Turks are Anatolian under the hood, somewhat more Greek than Armenian

Filed under: Armenians,Greeks,Human Population Genetics,Turks — Razib Khan @ 11:40 pm

My post, Are Turks Armenians Under The Hood?, attracted a little bit of controversy. The main criticism, which was a valid one, is that I did not sample Anatolian Greeks. A reader passed on three Anatolian Greek samples. I also added a Cypriot data set. To my mild surprise, the Anatolian Greeks and Cypriots cluster together, at the end of the Greece cline toward West Asians. Therefore, for further analysis, I pooled the three Greeks with the Cypriots.

Additionally, there are two Balkan Turk samples. Even on the PCA it’s pretty clear that they’re genetically very different from the other Turks (one of them is from what has become Bulgaria), though the shift toward East Asians indicates that Turkification is very rarely a matter purely of religious conversion to Islam and assimilation of the Turkish language (obviously it initially is for many people, but these people then intermarry with those with some East Asian ancestry).

One of the major problems is that the Armenian sample and the Anatolian Greek/Cypriot sample are genetically very close. This is obvious in the Fst distance. This is also totally reasonable since both populations occupy Anatolia, and historically there would have been a lot of gene flow between the two groups through isolation-by-distance dynamics.

The Turk position closer to East Asians is due to their East Asian admixture.

You can see it in the admixture plot too. As we all know there is definitely some northern admixture in the mainland Greeks. I haven’t bothered to check with the Mycenaean paper, but I assume that some of this is due to the migration of Slavs after much of the Balkans was abandoned after the reign of Maurice.

Of course, I ran Treemix too. Again, the closeness of the Anatolian Greeks/Cypriots and the Armenians is an issue in making a definitive conclusion.

In terms of drift the Turks seem about as far from Anatolian Greeks as Armenians. There’s the gene flow you’d expect, there are two from East Asians to Turks. I think that’s due to the East Asian source being somewhat heterogeneous, and the Dai outgroup not modeling the source populations perfectly.

Finally, there’s the f3 statistics. They basically show what I’m saying above: Armenians and Anatolian Greeks are both good model sources for Turks. The likely truth is that there is gene flow from all across Anatolia into these Turkish samples.

Group X1 X2 f3 z
Turkey anatolian_cypriot Dai -0.0029 -36.3940
Turkey Armenians Dai -0.0026 -34.2083
Turkey Greece3 Dai -0.0025 -32.7389
Turkey Georgian Dai -0.0026 -30.8836
Turkey Greece2 Dai -0.0024 -29.5462
Turkey GreekCentral Dai -0.0025 -23.6454
Turkey Greece1 Dai -0.0026 -23.0283
Turkey GreekThessaly Dai -0.0024 -20.1595
Turkey Armenians Lithuanians -0.0005 -11.5356
Turkey Lithuanians Dai -0.0012 -10.7473
Turkey Georgian Lithuanians -0.0004 -7.9691
GreekThessaly Armenians Lithuanians -0.0006 -6.6390
GreekThessaly anatolian_cypriot Lithuanians -0.0006 -6.3748
GreekThessaly Greece3 Lithuanians -0.0005 -4.6347
Greece2 anatolian_cypriot Lithuanians -0.0008 -15.7114
Greece2 Armenians Lithuanians -0.0008 -14.4083
Greece2 Greece3 Lithuanians -0.0005 -10.4508
Greece2 Georgian Lithuanians -0.0005 -8.2727
Greece1 anatolian_cypriot Lithuanians -0.0006 -6.5563
Greece1 Armenians Lithuanians -0.0005 -6.3712
Greece1 Greece3 Lithuanians -0.0004 -4.7896
Greece1 Georgian Lithuanians -0.0003 -3.1104
balkan_turk Greece1 Dai -0.0024 -7.1425
balkan_turk GreekThessaly Dai -0.0021 -6.1764
balkan_turk GreekCentral Dai -0.0021 -5.7848
balkan_turk Greece2 Dai -0.0019 -5.6794
balkan_turk anatolian_cypriot Dai -0.0019 -5.5944
balkan_turk Armenians Lithuanians -0.0014 -5.3207
balkan_turk Greece3 Dai -0.0017 -5.0815
balkan_turk Lithuanians Dai -0.0017 -5.0190

December 30, 2010

Are Turks acculturated Armenians?

To the left you see a zoom in of a PCA which Dienekes produced for a post, Structure in West Asian Indo-European groups. The focus of the post is the peculiar genetic relationship of Kurds, an Iranian-speaking people, with Iranians proper, as well as Armenians (Indo-European) and Turks (not Indo-European). As you can see in some ways the Kurds seem to be the outgroup population, and the correspondence between linguistic and genetic affinity is difficult to interpret. For those of you interested in historical population genetics this shouldn’t be that surprising. West Asia is characterized by of endogamy, language shift, and a great deal of sub and supra-national communal identity (in fact, national identity is often perceived to be weak here). A paper from the mid-2000s already suggested that western and eastern Iran were genetically very distinctive, perhaps due to the simple fact of geography: central Iran is extremely arid and relatively unpopulated in relation to the peripheries.

But this post isn’t about Kurds, rather, observe the very close relationship between Turks and Armenians on the PCA. The _D denotes Dodecad samples, those which Dienekes himself as collected. This affinity could easily be predicted by the basic parameters of physical geography. Armenians and Anatolian Turks were neighbors for nearly 1,000 years. Below is a map which shows the expanse of the ancient kingdom of Armenia:

Historic Armenia was centered around lake Van in what is today eastern Turkey. The modern Republic of Armenia is very much a rump, and an artifact of the historic expansion of the Russian Empire in the Caucasus at the expense of the Ottomans and Persians. Were it not for the Armenian genocide there may today have been more Armenians resident in Turkey than in the modern nation-state of Armenia,* just as there are more Azeri Turks in Iran than in Azerbaijan. Many areas once occupied by Armenians are now occupied by Kurds and Turks. But a bigger question is the ethnogenesis of the Anatolian Turkish population over the past 1,000 years.

Dienekes has already shed light on this topic earlier, adding the Greek and Cypriot populations to the mix as well as Turks and Armenians. The disjunction between Kurds and the Armenian-Turk clade suggests to us that Turks did not emerge out of the milieu of Iranian tribes in the uplands of southeast Anatolia and western Persia. Like the Armenians the Kurds are an antique population, claiming descent from the Medes, and referred to as Isaurians during the Roman and Byzantine period.

Below is a reformatted K = 15 run of ADMIXTURE with Eurasian population. I’ve removed the labels for the ancestral components, but included in populations which have a high fraction of a given ancestral component. The geographical labels are for obscure populations. I’ve underlined the four populations of interest:

First, let’s get out of the way the fact that Turkish samples have non-trivial, though minor, northeast Asian ancestry. The Yakut themselves are a Turkic group situated to the north of Mongolia. The more southerly and central Asian affinities the nomadic ancestors of the Anatolia Turks may have picked up in their sojourns over the centuries between their original homeland in east-central Siberia and Mongolia and West Asia. The rest of ancestry is rather typical of northern West Asian groups. In particular, Armenians! Here is the ancestral breakdown for the four groups I want to focus on using Dienekes’ labels:


Population Greek Cypriots Turks Armenians
West Asian 37.6 54.1 47.2 56.3
Central-South Asian 5.3 8.6 18.2 18.4
North European 25.1 5.6 12 12.3
South European 27.4 20.8 9.4 8.4
Arabian 3.4 8 4.3 3.4
Altaic 0.3 0 2.6 0.1
East Asian 0.3 0.2 2.2 0
Central Siberian 0.1 0.2 1.4 0.2
Chukchi 0 0 1.1 0.2
South Indian 0 0.1 0.8 0.3
Nganasan 0.1 0 0.4 0.2
Koryak 0.1 0 0.2 0.1
East African 0 0.4 0.1 0
West African 0 0 0.1 0
Northwest African 0.3 1.9 0.1 0

And now the correlations between the populations by ancestral components:



Greek Cypriots Turks Armenians
Greek * 0.863 0.823 0.813
Cypriots * * 0.941 0.946
Turks * * * 0.997
Armenians * * * *

Let’s remove the East Eurasian and African components, and recalculate the proportions by taking what remains as the denominator:


Population Greek Cypriots Turks Armenians
West Asian 38.1 55.7 51.8 57.0
Central-South Asian 5.4 8.9 20.0 18.6
North European 25.4 5.8 13.2 12.4
South European 27.7 21.4 10.3 8.5
Arabian 3.4 8.2 4.7 3.4

And the recomputed  correlations:



Greek Cypriots Turks Armenians
Greek * 0.747 0.640 0.647
Cypriots * * 0.901 0.908
Turks * * * 0.999
Armenians * * * *

With all the ~0 ancestral components which were common across these four populations removed the correlations have gone down. Except in the case of the Armenian-Turk pair, because I’ve removed the ancestries which differentiate them.

So what’s a plausible interpretation? A straightforward one would be that the Muslim Turk population of Anatolia has a strong bias toward having been assimilated Armenians, rather than Greeks. The cultural plasticity of Armenians in late antiquity and the early medieval period was clear: individuals of ethnic Armenian to origin rose the pinnacles of the status hierarchy of the Orthodox Christian Greek Byzantine Empire. The Macedonian dynasty of the Byzantines under which the civilization reached its mature peak were descended from Armenians who had resettled in Macedonia. Just as plausible to me is that eastern Anatolia as a whole exhibited little genetic difference between Greeks and Armenians, and the former were wholly assimilated or migrated, while the Armenians remained. One way to test this thesis would be type the descendants of Greeks who left eastern Anatolia during the population exchange between Greece and Turkey in the 1920s. But the difference between Greeks and Cypriots also points us to another possibility: perhaps the Greeks of Greece proper (as opposed to Anatolia) were much more strongly impacted by the arrival of Slavs? One need not necessarily rely solely on the Scalveni migrations either, water tends to be a major dampener to conventional isolation-by-distance gene flow, so the Greek mainland may always have been subject to more influence from the lands to the north.

Whatever the details of ethnogenesis may be, it will be interesting to see how things shake out as we increase sample sizes and get better population coverage. These results may be due to regional selection bias. One might expect that the descendants of Rumelian Turks be more “European” than Anatolian Turks. But, these data do seem to suggest on face value that Armenians are the population which Anatolian Turks have the most genetic affinity with.

* My main hesitation would be that Armenians are a very mobile population, and their numbers within a modern Turkey may have declined simply through emigration, just as those of Christian Arabs have over the 20th century.

Powered by WordPress