The Xiongnu wrote nothing that survives. What can be said about their language rests on two bodies of evidence, both of them Chinese: a couplet of four words preserved in a seventh-century chronicle, and some fifty terms glossed in the Han histories — titles, a few objects, and a great many personal names.[1] Out of that material a century and a half of scholarship has produced Turkic readings, an Eastern Iranian reading, a Yeniseian reading and the verdict that the language cannot be classified at all. The question matters here because the Zhou Shu derives the Ashina from the Xiongnu, and because the title of the Xiongnu ruler carries two words that Turkic texts use six centuries later.
The Jie couplet
The only connected text ever assigned to the language stands in the Jin Shu, a history of the Jin dynasty composed in the middle of the seventh century. The couplet is given in a passage about the years 307 to 311 in the lands of the Jie, whom the chronicle counts a branch of the Xiongnu; it is written in Chinese characters, it is four words long, and the chronicle supplies a Chinese translation.[1] Two obstacles stand between that and a reading. The text is too short to constrain anything, and recovering the sounds behind the characters requires a reconstruction of Old Chinese phonology, of which several competing versions are in use.[1]
More than a dozen readings exist, and most of them make the language an early form of Turkic.[1] The oldest of the often-cited ones is by Gustaf John Ramstedt, building on Kurakichi Shiratori and followed by Louis Bazin and Annemarie von Gabain. Alexander Shervashidze advanced the question by using more recent reconstructions of both Old Chinese and Proto-Turkic, and Anna Dybo elaborated his interpretation and identified the language of the couplet with Late Proto-Turkic.[1] A further Turkic reading by Andrew Shimunek and his colleagues argues on historical grounds that the Jie were not Xiongnu at all, and is held to sit badly with what is known of Turkic historical phonology.[1]
Alexander Vovin read the couplet instead as the nearest relative of Pumpokol, an extinct Yeniseian language. The objection to that reading is arithmetical. Of the four words, two can be construed from Pumpokol material only tentatively, and both involve lexical and grammatical elements that Pumpokol does not in fact attest; the remaining two have no parallel in Yeniseian at all, so that the reading has to posit a loan into the Xiongnu language from a language that is not identified.[1]
Glossed words and the two strata
Beside the couplet stand the isolated words the Chinese sources gloss, collected by Edwin Pulleyblank, who connected a number of them with Yeniseian lexemes. Most of the collection consists of personal names, which admit almost any interpretation; the discussion therefore turns on a few dozen terms.[1] Of Pulleyblank’s Yeniseian etymologies, some are held unconvincing outright, and the more plausible ones belong to a shared cultural vocabulary that is of foreign origin in Yeniseian as well as in Xiongnu — which makes them evidence of contact rather than of descent.[1]
Anna Dybo went through the fifty-six most transparent etymologies and found two strata, distinguished by date as well as by meaning. The Eastern Iranian words appear in the earlier sources, mostly those of the Western Han, and are almost entirely titles and terms for dairy products, with the word for comb beside them. The Turkic words appear in early and late sources alike and run across every field of meaning.[1] That distribution carries a historical claim: crucial parts of the Xiongnu package — dairy pastoralism and elements of the state organisation — came in with Eastern Iranian speakers who were in time absorbed by the Turkic-speaking majority of the confederation.[1] Isolated Tocharian and Mongolic forms may be present too, and they are the least secure part of the material.[1]
The ruler’s title belongs in this picture. The Han Shu gives it as Chengli Gutu Chanyu and glosses it as son of Heaven.[2] Chengli has long been read, with little dissent, as tengri, the sky-god of the Turkic and Mongolic world, and the word is plainly not the Chinese one; whoever borrowed the idea of a heaven-given sovereignty did not borrow the word for heaven with it.[3] Gutu, which the Chinese renders as son, has never been resolved etymologically to general satisfaction; this work reads it as Kut. Chanyu has been compared by Pulleyblank with a later Altaic dārγā, which again points away from the Chinese vocabulary.[3]
What the genomes add
Population genetics bears on the question only indirectly, and this work keeps the rule that a genome carries no language. Three Xiongnu genomes from Mongolia have been published; they are genetically diverse, one of them largely eastern Eurasian and the other two clearly mixed between western and eastern sources, and they carry, on the modelling of the geneticists, some eighteen per cent of an ancestry related to Iranian, Caucasian and Turanian populations.[1] That figure sits beside the Eastern Iranian stratum in the vocabulary without proving it.
The Yeniseian hypothesis is the one that genetics cuts against. If the Ket of today stand for the historical Yeniseian speakers of southern Siberia, then the decisive fact is a negative one: the Ket lack precisely the Iranian-related ancestry that marks the Iron Age steppe populations and the Xiongnu, and they show instead a strong affinity with their Samoyedic-speaking neighbours.[1] Linguistics and genetics arrive at the same place here by separate roads, which is the strongest form the argument can take.
The positions
The traditional and still prevailing view makes the Xiongnu, or their ruling stratum, Turkic speakers; it is held by Kurakichi Shiratori, Johannes Benzing, Edhem Tenişev, Claus Schönig, Anna Dybo and Juha Janhunen, and Omeljan Pritsak placed them more broadly among Altaic speakers.[1] Harold Bailey argued for Eastern Iranian. Lajos Ligeti, Edwin Pulleyblank and Alexander Vovin argued for Yeniseian. Gerhard Doerfer held that the material does not permit a classification at all, and Andrew Shimunek has recently taken the same position.[1]
The judgement this work adopts is that the confederation was mixed and that its ruling stratum spoke Turkic. That rests on the royal title, on the rulers’ names, on the heartland of the confederation and on what the Chinese histories say about Ashina descent, and it is carried in the data as an inference and not as a finding. A more guarded reading has it that the early Turkic peoples fell at one point or another under Xiongnu sway, and that their part in Turkic ethnogenesis cannot be assessed in full.[4] The debate on Turkic origins carries the wider question, and the making of the first steppe empire carries what the vocabulary implies about where the institutions came from.
References
Sources
Literature
- Alexander Savelyev; Choongwon Jeong, Early Nomads of the Eastern Steppe and Their Tentative Connections in the West (2020)
- Nicola Di Cosmo, Ethnogenesis, Coevolution and Political Morphology of the Earliest Steppe Empire: The Xiongnu Question Revisited (2011)
- Peter B. Golden, An Introduction to the History of the Turkic Peoples (1992)
- Emre-E. Yavuz, The Gokturk Empire: 2nd Edition (2025)
- Juhyeon Lee et al., Genetic Population Structure of the Xiongnu Empire at Imperial and Local Scales (2023)
What links here4
- Historiography
- Making of the first steppe empire · The Turkic ethnogenesis debate
Cite this article
Yavuz, Emre-E. “The language of the Xiongnu.” Turkic Atlas, revised 13 September 2026. https://turkicatlas.business-5d4.workers.dev/historiography/xiongnu-language-debate/
Last revised: 13 September 2026