Gene Expression

What it means to be a Turk

In the 19th & 20th centuries with the emergence of nationalism and its various scholarly subsidiaries in archaeology, philology and ethnography quite a few pan-ethnic movements rooted in language were born. Pan-Slavism, the Greater German idea (Grossdeutsch) and Pan-Arabism come to mind. As evident in their names these ideas shadowed relationships of language, but they often veered into racialist territory. In The History and Geography of Human Genes L. L. Cavalli-Sforza reported a substantial overlap between phylogenies generated from classical autosomal markers, and those of linguistic family trees. But obviously there were deviations on the margins, sometimes substantial ones.

Pan-Turkism is an idea which came to the fore after the collapse of the Ottoman Empire, though its roots go back to the 19th century. The role of a Turkish nationality was essential in the creation of the modern Turkish state by Kemal Ataturk. Though the Ottoman Empire had a Turkish speaking rule class at its heart, it was not fundamentally an ethnic polity. The Ottoman Empire was far less Turkish than the Hapsburg Empire was German. The Ottoman bureaucracy and military was open to many ethnicities, though often conditional on conversion to Islam. Albanians and Slavs played an important role in the Ottoman military, while the Janissaries were famously recruited from Christian subject peoples.

Ataturk aimed to replace this ethnically cosmopolitan but religiously Islamic Ottoman identity with a Turkish secular one. To a great extent he was successful, though not fully. Because of the vogue for racial theories in the early 20th century the new Turkish government naturally did fund research which purported to illustrate the distinctions between the Turkish peoples who had settled Anatolia and Southeastern Europe after the year 1000, and the native Greek, Slavic and Armenian populations. There is of course a natural problem with this: the basic origin of the Turkic peoples in Western Mongolia and the trans-Siberian steppe is well known, and Turkic speaking peoples still reside in that region, and physically they do not look much like Anatolian Turks at all. In fact there is a clear cline of Mongolian to European appearance from Central Asia to Anatolia. Of course common sense is often too easy to brush aside, and Pan-Turkish theories still seem to presuppose ideas of common ancestry.

This is where genetics come in. There have been several recent papers on attempting to adduce the relationship of Anatolian Turks to peoples of the Balkans and Central Asia, but generally they have utilized uniparental markers, mtDNA and Y. Alu insertion polymorphisms and an assessment of the genetic contribution of Central Asia to Anatolia with respect to the Balkans uses 10 alu elements to estimate admixture between these putative parental populations:

In the evolutionary history of modern humans, Anatolia acted as a bridge between the Caucasus, the Near East, and Europe. Because of its geographical location, Anatolia was subject to migrations from multiple different regions throughout time. The last, well-known migration was the movement of Turkic speaking, nomadic groups from Central Asia. They invaded Anatolia and then the language of the region was gradually replaced by the Turkic language. In the present study, insertion frequencies of 10 Alu loci…have been determined in the Anatolian population. Together with the data compiled from other databases, the similarity of the Anatolian population to that of the Balkans and Central Asia has been visualized by multidimensional scaling method. Analysis suggested that, genetically, Anatolia is more closely related with the Balkan populations than to the Central Asian populations. Central Asian contribution to Anatolia with respect to the Balkans was quantified with an admixture analysis. Furthermore, the association between the Central Asian contribution and the language replacement episode was examined by comparative analysis of the Central Asian contribution to Anatolia, Azerbaijan (another Turkic speaking country) and their neighbors. In the present study, the Central Asian contribution to Anatolia was estimated as 13%. This was the lowest value among the populations analyzed. This observation may be explained by Anatolia having the lowest migrant/resident ratio at the time of migrations.

Figure 2 illustrates the conclusion starkly:


As noted in the abstract it is important to remember that Anatolia has among the longest histories of settled agriculturalists in the world. Population estimates suggest nearly 12 million residents during the late Roman Empire. Though I am skeptical that the population was nearly so high during the early medieval period, even if it was 1 million that would be substantial. There is an asymmetry between the two source populations as farmers tend to greatly outnumber nomads.

Also remember that the positions of the Central Asian groups are likely closer to the Anatolian Turks than would be from Turkic populations closest to the ancestral homeland. The Turkish expansion occurred late in history, after the fall of the Roman Empire, but before the rise of Islam. Groups like the Huns and Avars who ravaged Central Europe during Late Antiquity were likely Turks, or had Turkic speaking peoples as part of their hordes. The famous Khazar Jews were also Turks. Turks took refuge among the Magyars after the expansion of the Mongol hordes. It is the last event which obscures the Mongolian origin of the Turks. The Mongols were a minor tribe on the northeastern fringes of what is today Mongolia before the rise of Genghis Khan. Western Mongolia was dominated by Turkic groups, and this was the likely point of departure for many of the earlier expansions. The Uyghurs, East Turks south of the original homeland are themselves highly admixed with a “Western” element which was indigenous to the region prior to the Turkish migrations.

The question then is why some regions became Turkic speaking, as Anatolia and northwest Iran did, and others, such as Iran, the Volga region and the Balkans, did not. There is no point in offering a one-size-fits-all explanation. But in the vast regions of Western Eurasia where the Turks did settle, and were genetically assimilated by the far more numerous sedentary populations, there are some instructive differences. In the Persian world the Turks arrived during the early Islamic period as mercenaries and slave soldiers. They converted to Islam and identified with the world of Islam, and accepted from the Persians their high culture. From the year 1000 Turkic elites ruled Iran up until 1900 except for the Mongol interlude. This left an impact, as 25% of Iranians are Turkic speaking, and Azeribaijan is predominantly Turkic. But they came as cultural supplicants. Iran retains its ancient name. Turkey does not.

Not so when the Seljuks and Ottomans pushed into Anatolia, and later the Balkans. Here the dominant elite language was identified not just with a conquered population, but with an identity, the Greek Orthodox Christian one, which was a rival to the Turks’ own self-conception as Muslim warriors, ghazis. The Ottomans encountering the Greeks of Anatolia already had the accoutrements of high culture, and demanded conversion to their religion for easy advancement into their power structure. Obviously being a Muslim did not entail speaking Turkish, as it is Ottomans elites cultivated poetry in Farsi, the language of Persia. But in Anatolia and the Balkans to become Muslim did in the end signal that one had turned “Turk.” Russia serves as a reverse example. During the 17th and 18th centuries large swaths of the Ural and Volga region were conquered from Turkish Muslim tribes. In the course of history a substantial number of these populations were Christianized, and Christianization was compulsory for the admission of Turkish elites into the boyar class who served the Czars. Though there are still Turkish Orthodox ethnicities in Russia, it seems likely that most of them were Russified, just as the Finnic peoples to their north were over time.

Of course specific historical events cleaned up the map a bit. The Mongol eruption likely transformed Central Asia from a predominantly Persian, minority Turkic, region to the inverse, as it is today. The exchange of Christians and Muslims during the wars between Turkey and Greece in the early 20th century also might have made things simpler, as there are suggestions of extant populations of Turkish speaking Christians (so likely Turkicized Greeks who retained their religion) being transferred to Greece, and non-Turkic speaking Muslims leaving the Balkans and settling in Istanbul and Anatolia.

In the 20th century politics has had a large effect on our conceptions of the history of populations. Before World War II every twist and turn in linguistic affinity was spun into a folk migration. After World War II it was “pots not peoples” who migrated. Both of these views are basically ideological fantasy, the reality is both to differing extents by region. The reliance on theory is enabled by the fact that empirically there weren’t ways to test particular hypotheses. This was clearest with the Etruscans, where the genetic data now show substantial exogenous origin, overturning a generation of archaeological theory which posits cultural continuity from the indigenous substrate to the early Iron Age Etruscan civilization. The Turks are a different case. Here “pots not peoples” is predominantly accepted. The Turkish spoken in Kashgar and Istanbul is reputedly intelligible, and the Uyghur nationalist movement to this day is based out of Turkey. But the affinity here is more of language and culture, not descent. Or it is? A conservative estimate is that 10% of modern Anatolia’s population is from Central Asia. Knowing that Anatolia has traditionally been one of the more populated regions of the old world due to its ancient settlement patterns, a 10% admixture within the last 10,000 years is actually substantial. If for example the population of Anatolia was 5 million in the year 1000 (likely an underestimate, as this was the peak of the Byzantine Empire), then hundreds of thousands of Turks might have entered Anatolia in the next few hundred years. Even with considerations of skewed fitness (some men having many descendants) the Turkish migration was not trivial. But neither was it preponderant.

Update: Where the samples came from in Turkey:



  1. #1 coldequation
    June 4, 2009

    These attempts to come up with hard racial boundaries that map to old linguistic/cultural/religious ones are kind of sad. You see the same thing on the other side, where atheist WN types imagine a hard racial boundary that just happens to map to the old boundaries of Christendom, even though they hate Christianity.

    Wikipedia has a list of the mothers of Ottoman Sultans:

    Note how many of them were not of Turkish descent. I don’t enough about Turkish history to say whether this reflects male preferences or political alliances.

  2. #2 Sogrom
    June 4, 2009

    In the present study, the Central Asian contribution to Anatolia was estimated as 13%. This was the lowest value among the populations analyzed.

    Which populations are they exactly?

    The Turkish spoken in Kashgar and Istanbul is reputedly intelligible

    Where did you read that bit? That is a wild exaggeration. I am a Turk myself. I’ve never heard Uyghur spoken, but several times heard Turkmen from the successive Turkmen caretakers of my grandmother while they were speaking on the telephone to their relatives in Turkmenistan. I must confess I couldn’t understand most of the words. The ones I understood were mostly the Arabic or Persian-descended words and the most common originally Turkic words. As to the caretakers themselves, a period of several months had to pass for each of them to be able to understand and speak Turkish in a degree sufficient enough to do their job efficiently. I give Turkmen caretakers as examples, because Turkmen language is much closer to Turkish than Uyghur. So I can assume it would be even harder for Uyghur caretakers. Unfortunately I can’t test my this assumption, as Uyghur language is written in Latin alphabet. But my reading attempts of Uzbek, a Turkic language from the same sub-branch with Uyghur, are largely confirming my this assumption. It is even harder for me to understand than Turkmen is, and again most easily recognizable words are foreign-descended words and the most common Turkic-origined words.

  3. #3 John Emerson
    June 4, 2009

    Something I’ve mentioned here but have never gotten very far with myself, mostly because I don’t have the tools: the Eurasian steppe is much more fluid and frequently stirred than more or less anywhere else in the world. As early as about 500 AD (Turkish empire) it was assuredly possibly for a single individual to travel from Beijing to Constantinople; it’s likely but not proven that some individuals did. (As I’ve said, it’s only about 5,000 miles from Beijing to Paris, a distance managable even on foot in a few years. The difficulty was always resistance by political entities, not distance or difficult terrain — the desert and mountain areas were routinely crossed or bypassed.)

    In addition, steppe political and social groups were thoroughly non-exclusive, practicing adoption, fostering, hostage-taking, kidnapping, slaving, diplomatic marriage, military alliance, the absorption of defeated enemy populations, etc. Groups also split and mrgd freely.

    Contrast Iceland, which is genetically unique (per C-S) strictly because of the island effect plus decimation bottlenecks — the original population is precisely known (a rare case) — all Norse and Irish. Or for a more common case, take any mountain valley in Switzerland, the Balkans or the Caucasus, or even a peasant area in the flatlands where families are tied to their plot of land for generations even though there are no natural borders.

    The earliest Scythians and Tocharians found were Europoid / Mongoloid mixes, IIRC, with the mixes varying according to time and place.

    In short, the steppe gene pool was defined by processes quite different than the processes forming sedentary gene pools, so contemporary C. Asian Turkish groups are also only very weakly representative of “the original Turks”. (Contrast Basques, for a vivid case).

    Anecdotally, my Chinese Hui friend and my Kyrgyz student both were East Asian in appearance, but informed m that members of their families were more Caucasian in appearance. They both seemed to slightly favor Caucasian appearance.

  4. #4 razib
    June 4, 2009

    Which populations are they exactly?

    they were talking about in reference to the other turkic populations. in any case, i’ve also put a map up where they got the anatolian samples above.

    Where did you read that bit?

    Sons of the Conquerors: The Rise of the Turkic World. i wonder if there is diglossia that the author was not aware of?

  5. #5 Sogrom
    June 4, 2009

    I made a typo when I wrote “Uyghur language is written in Latin alphabet.” It would be “is not written in Latin alphabet.” Sorry for the mistake.

  6. #6 Koray
    June 4, 2009

    I am an Anatolian Turk and as far as I know I can’t understand anything but Anatolian and Azerbaijani Turkish.

  7. #7 dearieme
    June 4, 2009

    “The Ottoman bureaucracy and military was open to many ethnicities, though often conditional on conversion to Islam” – and on getting castrated?

  8. #8 Sogrom
    June 4, 2009

    they were talking about in reference to the other turkic populations.

    The full passage reads:

    “Furthermore, the association between the Central Asian contribution and the language replacement episode was examined by comparative analysis of the Central Asian contribution to Anatolia, Azerbaijan (another Turkic speaking country) and their neighbors. In the present study, the Central Asian contribution to Anatolia was estimated as 13%.”

    Turkey and Azerbaijan are explicitly mentioned, but there is also the mention of “their neighbors.” Which neighbors are they talking about? There is no other dominantly Turkic neighbor in the region. Can they be majority non-Turkic countries like Syria, Iraq, Iran, Armenia, Georgia and the Russian Caucasus? How about Iranian Azeris?

    Sons of the Conquerors: The Rise of the Turkic World. i wonder if there is diglossia that the author was not aware of?

    Depends on what he means by intelligibility. You may shop at a bazaar in Kashgar with Istanbul Turkish with relatively little difficulty, but other than shopping and similar simple speech events, I guess you will have a hard time in communicating with people. BTW, Uyghurs and other Central Asian Turkic ethnicity members in Turkey all use Turkish (I mean Turkish of Turkey) when communicating with the locals, otherwise we cannot understand what they say most of the time.

  9. #9 Sogrom
    June 4, 2009

    “The Ottoman bureaucracy and military was open to many ethnicities, though often conditional on conversion to Islam” – and on getting castrated?

    The only castration applied by the Ottomans was the castration of the black Harem officials (recruited from the African slave markets). No other people, including the white Harem officials, were castrated. So it was apparently a racist application, as the Harem was the only place black slaves were normally recruited for.

  10. #10 Kaveh
    June 4, 2009

    the basic origin of the Turkic peoples in Western Mongolia and the trans-Siberian steppe is well known, and Turkic speaking peoples still reside in that region

    Is it possible the people of Western Mongolia and the trans-Siberian steppe *today* have different ancestry from the people who lived there when the migration of Turkic peoples into western Asia occurred? I mean, why shouldn’t there have been migration *into* those regions that would have changed the genetic makeup of the populations, just like what happened all over Central and Western Asia? Also, what about Chinese-Mongol intermarriages? I understand it was not uncommon for Chinese farmers close to the border to “go nomad”, and for nomads to settle down, so that would imply a lot of genetic mixing between Mongol and Chinese populations especially during the last 1000 years.

    Also, I’m surprised they didn’t include part of Iran as one of the bordering regions they compared Anatolia to, unless ‘Azerbaijan’ includes the Azeri parts of Iran.

    and on getting castrated?

    Only rarely.

  11. #11 razib
    June 4, 2009

    Is it possible the people of Western Mongolia and the trans-Siberian steppe *today* have different ancestry from the people who lived there when the migration of Turkic peoples into western Asia occurred?

    sure. but in general i think we can be sure that the original turks were a mongoloid people in appearance. they are described as such by both chinese and western (by western i include european, persian and islamic) sources. to a first approximation any mongolian origin peoples are basically the same when compared to the populations of southeast europe & anatolia.

  12. #12 Sogrom
    June 6, 2009

    I have just accessed the full paper. Just as I guessed, those neighbors they mention are the non-Turkic countries in the Middle East and Caucasus that were sampled (Syria, Georgia, Armenia and the Northern Caucasus). As it is stated in the abstract, they are all genetically more Central Asian than Anatolia according to the admixture analysis.

    From the paper:

    “Results indicated that the Central Asian contribution was 13% for the Turkish population. Furthermore, it was observed that the Central Asian contributions in the neighboring regions were higher than that of Anatolia and ranged between 20% and 33%. The Central Asian contribution was highest in Syria and Azerbaijan. Moreover, the estimates were approximately the same for Armenia, Georgia, and the Northern Caucasus.”

    Below is a diagram from the paper that shows the admixture estimates for each population (indicated under the title “Median”):

    TABLE 4. Summary statistics for the posterior distributions of Central Asian admixture estimates (p1) in admixed populations:

    Admixed population _ Median _ HPDa
    Anatolia _____________ 0.127 ___ 0.000–0.474
    Azerbaijan ___________ 0.315 ___ 0.000–0.843
    Armenia _____________ 0.201 ___ 0.000–0.707
    Georgia _____________ 0.217 ___ 0.000–0.596
    Northern Caucasus ___ 0.221 ___ 0.000–0.771
    Syria ________________ 0.326 ___ 0.000–0.858

    As stated in the above quotation, the highest Central Asian contribution is seen in Syria (~33) and Azerbaijan (~32) in the diagram, while the lowest contibution is seen in Anatolia (~13). Again in line with the quotation, Armenia, Georgia and the Northern Caucasus display approximately the same ratio of Central Asian contribution (~20).

    Now what to make of all this? First, we should take into account the distance of each sampled area to Central Asia. Second, we should also take into account that these results are in respect to the Balkans. And lastly, Central Asian contribution doesn’t necessarily mean Turkic contibution; all contributions since Palaeolithic times are included in these results.

    So how should we read the results? Anatolian Turkey, with a Central Asian admixture lower than its neighbors to the east and higher than its neighbors to the west, more or less genetically mirrors its geographic position. The same can be stated for Armenia, Georgia and the Northern Caucasus. But when we come to Syria and Azerbaijan, there is a clear rise in the Central Asain admixture. Whatever the reason of this rise is, we can assume a decreasing south to north cline for CA admixture in the region, in addition to the east to west cline. To test this assumption, admixture analysis of populations of Iran, Iraq, Lebanon and the other Arab countries of the region is required.

  13. #13 Sogrom
    June 7, 2009

    I just noticed that I forgot to add percent signs besides the approximate admixture rations 33, 32, 13 and 20. I should have written them as 33%, 32%, 13% and 20% respectively, as they are the approximate percentage ratios of Central Asian admixture for each population.

  14. #14 Korshad
    June 7, 2009

    There is alot of confusions and misconceptions about Turkic peoples and it basically stems down to a general lack of knowledge about us.

    Your post is a good introduction but I’d like to add some more.

    1. The Eurasian steppe has always been a mixing pot, people of various appearances (Mongoloid, Caucasoid). The Turks have never had “one” look, if you look at the Chinese records they describe Turks as being different from themselves looks wise, they have been described as having different colored eyes and hair. To Europeans Turks looked Oriental, to the Chinese Turks looked European. What I’m getting at is, there is no “Turkic gene”, just like no other nation has its own gene. Therefore its hard to determine Turkicnes through genes. We don’t even know the genetics of the immigrants who came to Anatolia, Iran, Azerbaycan and its hard to work this out as it was before the Mongol expansion.

    2. Turks of Azerbaycan and Turkey don’t look very different to people of Turkmenistan and even parts of Ozbekistan and Uygur regions. Infact many Turkmens look closer to their Oghuz Turk brethren then their Kazakh Turkic neighbors. To understand this one must look at regions history, most of the Mongol tribes who were Turkicized were in the modern Kazakhstan region.

    3. There is a high degree of mutual intelligibility between the Oghuz and Karluk/Chaghtai Turkic groups. The Turkmen of southern Turkmenistan can be understood especially in parts of Anatolia and Azerbaycan. Uygur-Ozbek is quite close to Oghuz Turkish, however, Kazakh and Kirgiz have notable differences.

    4. Nearly all the main Oghuz tribes have been recorded as having a presence in the Turkey/Azerbaycan region and founded countless Beyliks, Atabeks and states like the Aq-qoyunlu – Qara-qoyunlu the later which were nomadic Central Asian style nomadic Turkic states which didn’t have a permanant capital city. The influx of Turks must have been considerable as it wasn’t even Turks who came up with the name “Turkiye”, it derives from the Latin, Turchia = Land of the Turks and was coined in the 13th century, before the Ottoman Empire existed. The Turks called the region, “Roum”, in other words Roman lands.

  15. #15 Sogrom
    June 8, 2009

    Korshad, did you read the paper? According to it, not just Kazakhs and Kyrgyzes, but also Uyghurs and Uzbeks are genetically distant from Turks of Anatolia and Azerbaijan. You may object, as I think rightfully, that Turkmens, who are assumed to be the source population for the western Turkic migrations, weren’t included in the analysis. In the absence of them, we have no choice but to make our interpretations based on what we have.

    I already wrote that according to this paper Anatolian Turks are more Central Asian than their western neighbors. But if AT have some Central Asian Turkic origin genetically, we should also look at the eastern neighbors of Turkey. That is what the authors of the paper do, and the result is: Not just the Turkic neighbors like Azeris, but also the non-Turkic ones like Syrians, Armenians, Georgians and various Northern Caucasian ethnicities are genetically more Central Asian than AT.

    These results can be interpreted in various ways:

    1- Turkic and Mongolian invasions affected the indigenous peoples of the Middle East and Caucasus in an east to west direction. So it is very normal that AT, as the westernmost people in the Middle East, have the lowest rate of Central Asian admixture. Before the invasions, these regions had very low or no Central Asian admixture.

    2- The current Central Asian admixture rates are not the result of Turkic or Mongolian invasions; the genetic impact of these invasions is negligible. They are just a result of being geographically close to Central Asia, as the two regions have been in contact for many thousand and possibly ten thousand years.

    3- Anything between the option 1 and option 2. So, for instance, it can be assumed that a portion of the Central Asian admixture of AT and Azeris is the result of Turkic invasions.

  16. #16 Korshad
    June 9, 2009

    I said Turkmens are closest and next are the Ozbeks-Uygurs, the Turkmenistan and Khorrosan region were the major areas Turkic migrations to Azerbaycan (North and South) and Anatolia occurred. But its even more complicated as its likely the migrants may have settled and mixed with locals everywhere they stopped off.

    There is also the problem of what is Central Asian genes and if it was the same as when the major Turkic migrations occurred centuries ago. The Mongols had a considerable impact.

    I don’t think Turkicness can be measured or categorized by genetics. The tribal nature of the early Turks and the region they inhabited would have meant they were already genetically diverse. The tribes encouraged marrying outside the tribe and Eurasia has had so many different peoples cross over it.

    I think its easier to look at the records of tribes, clans and populations which especially in the Ottoman era were well documented. Unfortunately before that we can only make assumptions as there are not as many detailed records.