The Xiongnu wrote nothing that survives. What can be said about their language rests on two bodies of evidence, both of them Chinese. The first is a couplet of four words preserved in a seventh-century chronicle. The second is 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, Eastern Iranian and Yeniseian identifications of the language, and the verdict that it cannot be classified at all. The question matters here for two reasons. The Zhou Shu derives the Ashina from the Xiongnu. The title of the Xiongnu ruler also contains 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 and is four words long. The chronicle supplies a Chinese translation.[1] Two obstacles stand between that and an interpretation. The text is too short to constrain anything. Recovering the sounds behind the characters also requires a reconstruction of Old Chinese phonology, of which several competing versions are in use.[1]

More than a dozen interpretations exist. 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. Anna Dybo elaborated his interpretation and identified the language of the couplet with Late Proto-Turkic.[1] Talat Tekin also interprets the couplet as Turkic, in the sense ‘go out against the army, seize Bukuk’. He takes the l in its first verb, where later Turkic has š, as a sign of an l-language.[2] Andrew Shimunek and his colleagues offer a further Turkic interpretation. They argue on historical grounds that the Jie were not Xiongnu at all. This interpretation is held to sit badly with what is known of Turkic historical phonology.[1]

Glossed words and the two strata

Edwin Pulleyblank collected the isolated words that the Chinese sources gloss. He 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] Some of Pulleyblank’s Yeniseian etymologies are held unconvincing outright. The more plausible ones belong to a shared cultural vocabulary that is of foreign origin in Yeniseian as well as in Xiongnu. They are therefore evidence of contact rather than of descent.[1]

Pulleyblank’s own case of 1962 rests on sounds as well as words. The Han transcriptions show no aspirated stops. Common Mongol is thought to have had strongly aspirated ones, and Turkic is not. To that extent he found the sounds closer to Turkic than to Mongolic.[3] He judged nevertheless that the language was not closely like any form of Turkic or Mongolic that is known, and was unlikely to have been Altaic at all.[3] He ruled out Sino-Tibetan, and found no sign of a genetic tie to Indo-European.[3] The Yeniseian languages remained. One of his comparisons was the Hu word for ‘boot’, sak-dak, which Ligeti had already set beside Ket sagdi.[4] Harold Bailey offered an Iranian origin for the same word, which Pulleyblank gave in a note and did not rule out.[5] A further one set the reconstructed Xiongnu word for ‘son’ beside Yeniseian words for ‘son’.[6]

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][7] In her own summary, Turkic words already prevail among the common nouns of the Han period. The only common nouns of secure Iranian origin are the names of dairy products and the comb.[8] Titles and personal names lean the other way. Some of the Iranian titles seem to have passed through a Turkic-speaking milieu before the Chinese wrote them down. An example is the word for queen, γāt-tə̄j, from an Iranian hva-tuń.[7] She regards the Iranian titles as a possible legacy of the time when the Xiongnu were vassals of the Yuezhi, among whom Bagatur had been held as a hostage in his youth.[9] In the later transcriptions she found no word of secure Iranian origin. She offers the result as tentative, to be set beside the other hypotheses on the language.[8] That distribution contains a historical claim. Crucial parts of the Xiongnu package, namely dairy pastoralism and elements of the state organisation, came in with Eastern Iranian speakers. In time, the Turkic-speaking majority of the confederation absorbed these speakers.[1][10] Isolated Tocharian and Mongolic forms may be present too. 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.[11] Chengli has long been identified, with little dissent, with tengri, the sky-god of the Turkic and Mongolic world. 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.[12] Gutu, which the Chinese renders as son, has never been resolved etymologically to general satisfaction. Some scholars identify it with Kut. Pulleyblank has compared Chanyu with a later Altaic dārγā. This again points away from the Chinese vocabulary.[12]

The Turkic interpretations of the title go further. F. W. K. Müller identified chengli gutu with Old Turkic täŋri qut, ‘holy majesty’.[13] Owen Lattimore followed him in identifying gutu with qut.[14] Jean-Paul Roux sees the Turkic idea of kut, the ruling power that Tengri gives, already in this title. He doubts, though, that the Han Shu is right to gloss gutu as ‘son’.[15] Pulleyblank judged that Müller’s identification becomes less plausible once the Han-period form of the word is reconstructed. He reconstructed the Han form of the word as something like *kwala, with a foreign l in the second syllable.[16] Kurakichi Shiratori rejected Müller’s identification too. He regarded gutu as a Tungusic word, hutta, ‘son’.[13] Pulleyblank doubted that as well, because the phonetic resemblance is vague once the Han form is reconstructed.[16]

Two further Turkic comparisons concern the title shanyu and the Xiongnu word for the ceremonial sword. Gerard Clauson suggested that the title shanyu is the Turkic yabgu.[17] Togan made the same equation, and Esin interprets the whole title as the Turkic phrase ‘Tengri Kut Yabgu’.[18] Pulleyblank rejected the equation. The Han form is a poor phonetic match for yabgu. The Han Chinese also had a good transcription of yabgu of their own, xihou, which they used for officers of the Wusun and the Yuezhi.[17] Bahaeddin Ögel holds that the title yabgu was already a first-rank office among the Xiongnu, as it later was among the Göktürks. He holds that the Kushan rulers took it from them.[19] The sword word, Chinese jinglu, has been set beside Turkic kıŋrak, ‘large knife’, and has been called the oldest Turkic word on record.[20] Pulleyblank found the match in meaning and sound striking. He pointed out that the Turkic word probably has a connection with Iranian words for ‘sword’ as well.[21]

The genetic evidence

Population genetics bears on the question only indirectly, because a genome contains no language. Three Xiongnu genomes from Mongolia have been published. They are genetically diverse. One of them is largely eastern Eurasian, and the other two are clearly mixed between western and eastern sources. On the modelling of the geneticists, they carry 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. They show instead a strong affinity with their Samoyedic-speaking neighbours.[1] Here linguistics and genetics reach the same result independently. This is the strongest form the argument can take.

The Old Arin hypothesis

Svenja Bonmann and Simon Fries argue that the Xiongnu spoke an early form of Arin. Arin is one of the Yeniseian languages, recorded between the eighteenth and twentieth centuries and now extinct. The two authors argue that the Huns of Europe spoke the same language.[22] They build the case in four steps, each resting on a different kind of evidence.[23]

The first step is vocabulary borrowed into Turkic and Mongolic. A comparison of twenty-nine Turkic, eleven Mongolic and twelve Yeniseian varieties yields five words, for lake, rain, lizard or frog, silver and birch. These words cannot be explained from Turkic or Mongolic roots. They belong to the inherited Yeniseian vocabulary. They show a word-initial k- where Arin has it and the other Yeniseian languages have none.[24] Earlier work treated Yeniseian mainly as the recipient of Turkic and Mongolic loans. Here the direction is reversed for these five words. They occur in every main branch of both families and share in sound changes of the daughter languages. They therefore entered at the stage of Proto-Turkic and Proto-Mongolic.[25] Two consequences follow. Those stages are no older than about the time of the Xiongnu empire. Old Arin was also a prestigious language of Inner Asia.[26]

The second step is the Jie couplet and the glosses. The couplet is interpreted as two Yeniseian verb phrases. This interpretation holds that the couplet fits Arin better than the neighbouring Pumpokol. The syllable interpreted as ‘army’ resembles the Arin word for army, where the related words in the other languages mean ‘war’. The verb ending -taŋ is attested as a third-person plural only in Arin.[27] Of the glosses, the second element of the ruler’s title, kwala, ‘son’, is regarded as an early stage of the Arin word for son. The Xiongnu word for milk is regarded as an Arin form with a voiceless initial.[28]

The third step is Hunnic names. Attila is derived from an Arin adjective for ‘quick’ with a suffix that softens the quality. The name was perhaps an epithet at first. The authors call this derivation plausible but by no means necessary.[29] Atakam and Eskam are interpreted as compounds whose second element is the Arin word for ‘wife’. The first element means ‘alive’ or ‘quick’ in Atakam and ‘sky’ or ‘god’ in Eskam. This would make Eskam a woman’s name, ‘god’s wife’.[30]

The fourth step is place names. Of 171 river names and place names with the element -kul, ‘water’, 131 lie north of the Altai and Sayan mountains and 40 south of them. In a second sample of 108 Yeniseian river names, 43 are of the Arin type, 25 Ket–Yugh, 20 Kott–Assan and 20 Pumpokol. The Arin names predominate north-west of the range.[31] The authors interpret this as a westward movement of Arin speakers from the Altai–Sayan region and set it beside the westward movement of the Huns.[32] For the link between the two peoples they cite a genetic study of 2025 that reports long shared genomic tracts between some of the highest-ranking individuals of the Xiongnu period and persons of the fifth and sixth centuries in the Carpathian basin.[33]

Against the Turkic and Mongolic interpretations they argue that nothing shows a Turkic or Mongolic presence in Inner Asia between the third century BC and the second century AD. The earliest Turkic words are of the seventh century AD, and the earliest Mongolic fragments are of the fifth and sixth.[34] They count an Iranian component in the empire possible and quite likely. They hold, however, that the Iranian etymologies proposed so far do not meet the standards of comparative linguistics. They also hold that an Iranian-speaking ruling dynasty has not been shown.[35] They offer the interpretation as a hypothesis that texts found in the future could test.[36]

The positions

The traditional and still prevailing view makes the Xiongnu, or their ruling stratum, Turkic speakers. It is held by Johannes Benzing, Edhem Tenişev, Claus Schönig, Anna Dybo and Juha Janhunen. Kurakichi Shiratori held it at first, in his work of 1900 on the Xiongnu language, and later changed his view.[37][38] By 1926 he regarded the Xiongnu as a mainly Mongolic people with some Tungusic admixture. He regarded shanyu as a Chinese expression, and yanzhi and gutu as Tungusic words, and called the attribution to the Turks a long-established habit of Western scholars.[39] Omeljan Pritsak placed the Xiongnu more broadly among Altaic speakers.[1] Harold Bailey argued for Eastern Iranian. Lajos Ligeti and Edwin Pulleyblank argued for Yeniseian. In 1962 Pulleyblank set that case out, and also described the Turkic position as the most widely held one in Western scholarship. He noted that the Chinese statement of descent has little evidential value in itself. By the time the Göktürks appeared in the sixth century, the Xiongnu had long since disappeared.[40] He named the equation of chengli with tängri as the main support of the Turkic position.[41] Bonmann and Fries argue for it in the narrower form of an early Arin, for the Xiongnu and the Huns alike.[42] Gerhard Doerfer held that the material does not permit a classification at all. Andrew Shimunek has recently taken the same position.[1] A further group of scholars states the Turkic position as a settled fact, or close to it. In Turkish scholarship the Xiongnu are widely held to be at least proto-Turks.[43] Talat Tekin holds that they spoke an archaic form of Turkic, Proto-Turkic or even Pre-Turkic. He rests that on the Chinese statement that the Göktürks descended from the Xiongnu, and on the few glossed words and the Jie couplet. He grants that the language material that could prove it is almost nothing.[20] Owen Lattimore held in 1940 that the Xiongnu spoke a Turkic language, and rendered Modu as Bagatur. He added that the language must come from the original stock from which Turkic and Mongolic both later derived. He also said that his interpretation of the names should not be claimed as correct.[14] Zeki Velidi Togan regarded the Xiongnu as a Turkic people, perhaps the ancestors of the Oghuz and the Kipchaks.[44] Faruk Sümer counted them among the Turks and took the Chinese derivation of the Göktürks, Uyghurs and Kirghiz from them as one of the main proofs.[45]

An argument of this group is that the vocabulary known from Chinese notices is Turkic. László Rásonyi lists kız, ‘girl’, katun, ‘queen’, bügü, ‘magic’, tengri, ‘sky’, kırsa, ‘steppe fox’, tinglig, ‘squirrel’ and tok, ‘thick, strong’.[46] İbrahim Kafesoğlu adds sky, power, wolf, realm, camp, standard and sabre. He takes the lists back to Shiratori’s work of 1900 and De Groot’s of 1921, and holds that the ruling stratum and the dynasty spoke Turkic.[47] He also holds that the rulers called themselves by a word that means ‘people’ in Turkic. Otto Maenchen-Helfen objected to this equation, and Shafer defended it.[47]

A second argument draws on culture, not on words. Kafesoğlu points to the steppe way of life, the cult of Heaven and the family built on the law of the father. He adds that Turkic herders kept horses and sheep, and that the Mongols and the Tungus kept pigs.[48] Ögel likewise holds that the Xiongnu and the Turks kept no pigs, and that the Donghu kept many.[49] The inventories of the settlement at Ivolga, however, list pigs among the animals.[50] Both authors also draw a conclusion about language from food and livestock.

The Turkic position comes in qualified forms too. Rásonyi counts the Xiongnu the first Turkic people to enter written history. He holds that a single purely Turkic-speaking Xiongnu people cannot be assumed, and that the name stands for a union of Mongolic, Tungusic and other tribes under Turkic leadership.[51] Roux does not commit himself on the language of the rulers. He grants that it is disputed who spoke an Altaic language in these centuries, and that it is harder still to separate Proto-Turks, Proto-Mongols and Proto-Tungus without a text.[52] It has also been said that knowledge of the Xiongnu language is not yet enough to settle the form of their titles. On the same account, nothing certain can be said of the language before about AD 250.[53]

Paul Pelliot took a different line on the titles. He found the element gutu again in the title of a Rouran khan. He regarded the endings in -fa of several Göktürk titles as a Rouran legacy of Mongolic form.[54] A further position is agnostic on all the early peoples at once. It holds that whether the Xiongnu, the Wusun of Central Asia or the Huns of Eastern Europe spoke Turkic is not known. A description of the language that is actually recorded can therefore leave them aside.[55]

The best explanation is that the confederation was mixed and 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. It is an inference, not a finding. A more guarded position has it that the early Turkic peoples fell at one point or another under Xiongnu sway. In that case their part in Turkic ethnogenesis cannot be assessed in full.[56] The debate on Turkic origins covers the wider question. The making of the first steppe empire covers what the vocabulary implies about where the institutions came from.

References

  1. ^ a b c d e f g h i j k l m n o Savelyev 2020
  2. ^ Tekin 2003, p. 12–13
  3. ^ a b c Pulleyblank 1962, p. 242
  4. ^ Pulleyblank 1962, p. 242–243
  5. ^ Pulleyblank 1962, p. 243–244
  6. ^ Pulleyblank 1962, p. 244–245
  7. ^ a b Dybo 2007, p. 114–115
  8. ^ a b Dybo 2007, p. 115
  9. ^ Dybo 2007, p. 81, 115
  10. ^ Gong 2026, p. 61–63
  11. ^ Han Shu, juan 94
  12. ^ a b Cosmo 2011
  13. ^ a b Kurakichi 1926, p. 38
  14. ^ a b Lattimore 1951, p. 450–451
  15. ^ Roux 1994, p. 91, 133–134
  16. ^ a b Pulleyblank 1962, p. 244
  17. ^ a b Pulleyblank 1962, p. 256
  18. ^ Hassan 1985, p. 296–297
  19. ^ Ögel 2000, p. 152
  20. ^ a b Tekin 2003, p. 12
  21. ^ Pulleyblank 1962, p. 222
  22. ^ Bonmann 2025, p. 29–30
  23. ^ Bonmann 2025, p. 30
  24. ^ Bonmann 2025, p. 35–38
  25. ^ Bonmann 2025, p. 38–40
  26. ^ Bonmann 2025, p. 40
  27. ^ Bonmann 2025, p. 41
  28. ^ Bonmann 2025, p. 41–42
  29. ^ Bonmann 2025, p. 43
  30. ^ Bonmann 2025, p. 43–44
  31. ^ Bonmann 2025, p. 46–47
  32. ^ Bonmann 2025, p. 47–48
  33. ^ Bonmann 2025, p. 34
  34. ^ Bonmann 2025, p. 31–32
  35. ^ Bonmann 2025, p. 32
  36. ^ Bonmann 2025, p. 48
  37. ^ Kafesoğlu 1997, p. 58
  38. ^ Kurakichi 1926, p. 9–12, 16, 38
  39. ^ Kurakichi 1926, p. 9–12, 16, 36–38
  40. ^ Pulleyblank 1962, p. 239
  41. ^ Pulleyblank 1962, p. 239–240
  42. ^ Bonmann 2025, p. 29
  43. ^ Hassan 1985, p. 216
  44. ^ Togan 1981, p. 40
  45. ^ Sümer 1972, p. 2
  46. ^ Rásonyi 1993, p. 12
  47. ^ a b Kafesoğlu 1997, p. 59
  48. ^ Kafesoğlu 1997, p. 58–59, 318
  49. ^ Ögel 2000, p. 171
  50. ^ Houle 2011
  51. ^ Rásonyi 1993, p. 65
  52. ^ Roux 1994, p. 12–13, 52
  53. ^ Hassan 1985, p. 283–284
  54. ^ Pelliot 1929, p. 228
  55. ^ Erdal 2004, p. 2
  56. ^ Golden 1992

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)
  • Marcel Erdal, A Grammar of Old Turkic (2004)
  • Svenja Bonmann; Simon Fries, Linguistic Evidence Suggests that Xiong-nu and Huns Spoke the Same Paleo-Siberian Language (2025)
  • Shubin Gong, Lexical Stratigraphy and Turkicization in the Xiongnu: An Iranian Prestige Layer in Chinese Transcriptions (2026)
  • Anna V. Dybo (А. В. Дыбо), Лингвистические контакты ранних тюрков. Лексический фонд: пратюркский период (2007)
  • Shiratori Kurakichi, A Study on the Titles Kaghan and Katun (1926)
  • Talât Tekin; Mehmet Ölmez, Türk Dilleri: Giriş (2003)
  • Edwin G. Pulleyblank, The Consonantal System of Old Chinese, Part II (1962)
  • Paul Pelliot, Neuf notes sur des questions d'Asie Centrale (1929)
  • Owen Lattimore, Inner Asian Frontiers of China (1951)
  • László Rásonyi, Tarihte Türklük (1993)
  • İbrahim Kafesoğlu, Türk Millî Kültürü (1997)
  • Ümit Hassan, Eski Türk Toplumu Üzerine İncelemeler (1985)
  • Zeki Velidi Togan, Umumi Türk Tarihine Giriş I (1981)
  • Faruk Sümer, Oğuzlar (Türkmenler). Tarihleri, boy teşkilâtı, destanları (1972)
  • Jean-Paul Roux, Türklerin ve Moğolların Eski Dini (1994)
  • Bahaeddin Ögel, Türk Kültür Tarihine Giriş I: Türklerde köy ve şehir hayatı (2000)
  • Jean-Luc Houle; Lee G. Broderick, Settlement Patterns and Domestic Economy of the Xiongnu in Khanui Valley, Mongolia (2011)

Related

What links here16

States
Xiongnu
Places
Noin Ula
Peoples
Eurasian Nomads · History of the Turkic peoples · Jie · The Great Empires of the Steppe
Lineages
Luandi
Languages
Chinese transcriptions of Turkic · The Hunnic language · The Turkic languages
Controversies
Archaeogenetics of the Steppe Empires · The Göktürks and the Xiongnu · The Huns and the Xiongnu · The origin of the title Khan · The Turkic ethnogenesis debate · Xiongnu: Making the First Steppe Empire

Cite this article

Yavuz, Emre-E. “The language of the Xiongnu.” Turkic Atlas, revised 8 November 2026. https://turkicatlas.org/historiography/xiongnu-language-debate/

Last revised: 8 November 2026