Heavy Velar vs Meager Bilabial Articulations in Xiongnu Language

« previous post | next post »

[This is a guest post by Penglin Wang]

            The great difficulties we have with trying to study Xiongnu language persist from trying to glean Xiongnu words, especially the glossed ones, in early Chinese sources for comparison in order to know what linguistic affiliation it seems to have in the central Eurasian region. Since these difficulties cannot be overcome at all owing to its extinct status a millennium plus ago, an alternative approach could be to recognize that there are different components of language regardless of living or extinct and attempt to observe how different components can differ from one another yet still be entities that most researchers would want to treat as linguistic data or facts rather than imaginations for a comparative purpose. It could then be possible to open up a window to contribute to a solution of some classic problems in Altaic comparative studies. One such attempt is to examine the available Xiongnu words from the perspectives of articulatory phonetics and phonotactics. Concern for these is characteristic of Xiongnu studies. Pulleyblank (1962:242) has insightfully observed “only *b- initially, never *p-” in the Xiongnu transcriptions.

            Any close study of the diachronic change or synchronic variation from one sound to another will require the attention to place and manner of articulation from phonetic perspective. It is certain that articulatory phonetics exists as an independent linguistic mechanism for any languages on the parts of available data, despite the great difficulty for one to obtain glossed words. Below in List 1-2, the Xiongnu words are presented with their word-initial velar consonant being reconstructed generally according to Chinese phonologists. Please note the following: a) the sequence of one entry is the Chinese pinyin spelling, the hypothetical reconstruction in the parenthesis, Chinese transcription, and the source in which the word is attested; b) the letter x in pinyin is [ɕ], whereas the x in the reconstruction is [x]; c) the hyphen – following a word means that the word is immediately accompanied by a string of Xiongnu lexical elements in Chinese transcriptions and determined as such by my morphological attempts; d) I adhere to the phonetic integrity of Chinese characters. Despite exegetic intervention with the normal pronunciation of some characters, I do not follow the intervention because it was actually and logically intended to add alternative pronunciations to the foreign words transcribed in Chinese rather than to the concerned characters per se and thus has no solid basis for maintaining it; e) List 1 is from my collection of 80 plus words and List 2 from 10 plus items, yielding the ration of 7:1.


List 1: Xiongnu words with a velar consonant in the word-initial position

Gongnu (*k-) 恭奴, Hanshu 94.3828

Gudu (*k-) 骨都 Shiji 110.2890

Guli (*k-) 谷蠡, Shiji 110.2890

Gutu (*k-) 孤塗 ‘son’, Shiji 94.3751

Hailuo- (*h-) 醢落-, HHS 89.2939

Huchuquan (*x- ~ *h) 呼廚泉, HHS 89.2965

Hudu- (*x- ~ *h-) 呼都-, Hanshu 94.3828

Huhanxie (*x- ~ *h-) 呼韓邪, Hanshu 8.266

Hujie (*x- ~ *h-) 呼揭, Hanshu 94.3795

Hulanruo (*x- ~ *h-) 呼蘭若-, HHS 89.2962


            The Xiongnu words beginning with what is today’s bilabial or labiodental consonants in Chinese transcriptions are rarely found in Chinese sources. And these occurrences should be examined by going through the phonologists’ hypothetical reconstructions to understand aspects of the Xiongnu phonotactics in word-initial position. For the time being I am unable to see the bilabial stop p equivalent in its phonetic value to the reconstructed Proto-Altaic *p in the word-initial position in the Xiongnu language.


List 2: Xiongnu words with a bilabial consonant in the word-initial position

Bixu ~ bishu (*b-) 比余 ~ 比疏 ‘comb’, Shiji 110.2897, Hanshu 94.3758

Pulei (*b-) 蒲類, Shiji 22.1147


            The word bišu (bishu 比疏) or biśu (bixu 比余, 余 here = 徐 xu) is of great significance to understand Xiongnu contact with Yueshi [VHM:  Yuezhi] and also how a consideration of semantic change contributes to numeracy studies. In 174 BCE, according to Shiji (110:2897), in a letter to the Xiongnu shanyu, Emperor Xiaowen (孝文) of Han wrote: “As Han and Xiongnu have agreed to be brothers, I would like to deliver you, the Shanyu [VHM:  or Chanyu; also see bibliography below], rich gifts”. Among the gifts there is one biśu (比余) along with xupi (胥紕) as a Xiongnu word denoting a belt. Shiratori (1939:33) treats the word biśu as a Xiongnu word, transcribes it as piso, and connects it with Korean pis [pit] ‘comb’ and Hungarian fešü ‘comb’. The Hungarian fešü seems to be spelled as fésű. Although I have no knowledge of the Hungarian language, in search of the Hungarian word for ‘comb’ I found another word taréj or taraj meaning ‘comb’. I wonder if this Hungarian word has a connection with Tokharian B tarya and A tri (täryā-) ‘three’. It is highly likely that the Xiongnu word biśu had a connection with Tokharian B piś ‘five’ through Yueshi adjacent to Xiongnu and further with Old Turkic biš ~ beš and Turkish beş ‘five’, for some low cardinal numerals are related to a semiotic set of ǀǀǀǀǀ (Wang 2015:160). In this sense, the Xiongnu biśu must have been old, illuminating, and significant to numeracy studies, like the quinary formation based on the whole hand in Indo-European proposed by Polomé (1968:100).

            At this point, we should note that the phonemic systems of the two Tokharian languages have a single set of p, t, k stops (Adams 1988:10, 40), i.e., they were not paired with b, d, g; moreover, there was ‘a very much simplified obstruents system’ in Proto-Tokharian. If and only if the b of Xiongnu biśu may be reconstructed as *p as p in piso as shown in Shiratori’s transcription (which, as I believe, had no relevance to *p), the hypothetical *p remains /b/ in Turkic down to today and represents itself /f/ in Hungarian. Reminiscent of this Turkic initial /b/ is the connection of Uyghur baǰa and WMo baǰa ‘husbands of sisters’ with Tokharian A pācar and B pācer ‘father’, Wusun bujiu (布就) ‘caretaking father’ and Wahon boǰa ‘husbands of sisters’ derived from the Indo-European root *pəter– ‘father’ (Wang 2021). In addition to these two plausible etymologies, the Xiongnu word pulei (蒲類) may likewise be used to consider the Xiongnu response to Indo-European word-initial p– in a relatable etymology with Hittite paršanaš ‘leopard’ (Wang 1995:197, I wonder if this connection has been anticipated in the existing publications.) During the Xiongnu era the word pulei seemed to be popular in nomenclature and official titles both in Inner Asia and Han China. It was used for a small polity with 325 households counting 2032 people in the Western Regions, a body of water, an official position called Xiongnu dong pulei wang (匈奴東蒲類王 ‘Xiongnu east pulei king’ in word-for-word translation), and a Chinese military rank pulei jiangjun (將軍) ‘general’ recorded in Hanshu (96.3919, 3874, 94.3786). In addition, the Han dynasty had the military rank huya jiangjun (虎牙將軍) ‘tiger teeth general’ comparable to pulei jiangjun mentioned in Hanshu (94.3785), wherein we see Pulei Jiangjun Zhao Chongguo (趙充國) and Huya Jiangjun Tian Shun (田順), each commanding 30,000 plus horsemen. I am unaware whether the Xiongnu initial b and Indo-European initial p (or b) has anything to do with MMo h– in these three cases. In other words, the so-suggested Proto-Altaic *p could not evolve to f or h in Turkic and Mongolic. This is a telling failure.

            To understand how the phoneme /p/ emerged in Altaic, especially in the word-initial position, we should start with Old Turkic. Thanks to the Russian lexicographers of ДТС, we can see how the word-initial /p/ came about. ДТС has 180 entries beginning with p on pages 396-399: about 82 came from Sanskrit, 17 from Iranic (Persian and Sogdian), 10 from Chinese, 56 alternated with those beginning with b including one with f. These preliminary statistics show: language contact is responsible for 60% of the p-initial words, and the b ~ p alternation for 31%. The remaining 9% might be internally motivated according to the early Old Turkic consonant system organized by Erdal (2004:61), who does not include sounds found only in loan words. So, his system includes /p/ but not /f/. The lack of /f/ in Old Turkic reminds us of the similar phenomenon in Kitan. That was why Kitan responded to external f-initial words with p-initial. Furthermore, Kitan pilie (匹裂) ‘a small-mouthed wooden jar’ attested in the eleventh century is connected with and most probably adopted from Wahon pil ‘wooden bowl’ since /p/ has been a regular phoneme in Eastern Iranian languages.

            Since the Xiongnu velar concentration in two sets of consonants k/g and h/ɣ in contrast with relative inactivity in bilabial articulation of b holds up as more available data are amassed, it has given us a useful tool for prying into the phonetic variation and sound change from a velar stop to a velar fricative in Xiongnu. Furthermore, as we know that the attested Old Turkic and Middle Mongolian shared the Xiongnu velar concentration and that both languages existed in the former Xiongnu land, we can assume that both languages were either embedded in the speech of the Xiongnu people or genealogically affiliated with the Xiongnu language. Ramstedt’s hypothesis is about change from the Proto-Altaic bilabial *p to MMo h-. While Turkic is less active in bilabial articulation of /p/ in the word-initial position and Written Mongolian had not yet acquired /p/, how can one prove the change from the Proto-Altaic *p to MMo h– in its large-scale presence?



Adams, D. Q. 1988. Tocharian historical phonology and morphology. New Haven: American Oriental Society.

ДТС = Дreвнетюркский cловарь (Old Turkic Dictionary), 1969. Лeнингpaд: Наука.

Erdal, Marcel. 2004. A Grammar of Old Turkic. Leiden: Brill.

Hanshu (漢書 Book of Han) compiled by Ban Gu in 82. http://hanchi.ihp.sinica.edu.tw/ihp/hanji.htm.

HHS = Hou Hanshu (後漢書 Book of Later Han) compiled by Fan Ye in 445. http://hanchi. ihp.sinica.edu.tw/ihp/hanji.htm.

Polomé, Edgar C. 1968. The Indo-European numeral for ‘five’ and Hittite panku– ‘all’. Pratidānam: Indian, Iranian and Indo-European Studies Presented to Franciscus Bernardus Jacobus Kuiper on His Sixtieth Birthday, ed. J. C. Heesterman et al., 98-101. Janua Linguarum, Series Maior 34. The Hague-Paris: Mouton.

Pulleyblank, E. G. 1962. The consonantal system of Old Chinese. Asia Major 9:206-265.

Shiji (史記 Records of the Grand Historian) written by Sima Qian in 91 BCE. http://hanchi. ihp.sinica.edu.tw/ihp/hanji.htm.

Shiratori, Kurakichi. 1939. Xiongnu minzu kao (匈奴民族考), translated by He Jianmin. Kunming: Zhonghua shuju.

Wang, Penglin. 2015. Number Conception and Application. New York: Nova Science Publishers.

Wang, Penglin. 2021. Wusun Bujiu ‘Caretaking Father’ and Tokharian Pācar ‘Father’. Paper presented at the Western Conference on Linguistics (WECOL) hosted online by California State University, Fresno, November 13-14. https://cwu.academia.edu/PenglinWang.


Selected readings




  1. Victor Mair said,

    April 16, 2022 @ 10:14 pm

    From Axel Schuessler:

    If memory serves, 蒲類 (OC *ba-rus) is identical with the transcription of the Mongolian word bars ‘tiger’.

  2. Mehmet Oguz Derin said,

    April 17, 2022 @ 5:45 am

    Isn't Hungarian taréj comb related to fésű comb mainly through the overlap of these words in English, where this language uses the same word for both concepts? Otherwise, the comb word starting with taré- or tara- might not be that old of a contact. A derivative of Turkish tarak (comb) from tar- (to comb) appears in Bulgarian, at least according to Lakova (1970), as darača- (to comb), noting down the change of k.

  3. Philip Taylor said,

    April 17, 2022 @ 6:12 am

    I'm unclear what "both concepts" are, Mehmet — a comb for one's hair is clear, but what is the other ? Is it, perhaps, "comb" as in "honeycomb" ?

  4. Mehmet Oguz Derin said,

    April 17, 2022 @ 6:23 am

    @Philip sorry, the other should be anatomical: https://en.wikipedia.org/wiki/Comb_(anatomy)

    (BTW, irrelevant but, I use my Oguz name)

  5. Philip Taylor said,

    April 17, 2022 @ 6:53 am

    Ah, as in "cockscomb", Oguz — all now clear; thank you !

  6. Chris Button said,

    April 17, 2022 @ 6:55 am

    Among the gifts there is one biśu (比余) along with xupi (胥紕) as a Xiongnu word denoting a belt. Shiratori (1939:33) treats the word biśu as a Xiongnu word, transcribes it as piso, and connects it with Korean pis [pit] ‘comb’ and Hungarian fešü ‘comb’.

    For the semantics to work, presumably we’re talking about a comb as a device for holding hair in position (much like a belt) rather than something for brushing hair out?

  7. Victor Mair said,

    April 17, 2022 @ 10:37 am

    On Combs

    In my work on the mummies of Eastern Central Asia / Tarim Basin that date back as much as four thousand years ago, I have often observed and commented on / written about the presence of combs in the burials. Most women were accompanied with combs., which generally consisted of a row of sharpened wooden pegs / tines affixed to a wooden stick at 90º.

    It is also worth noting that artifacts for combing human hair often served a dual purpose for carding wool.

  8. David Marjanović said,

    April 17, 2022 @ 1:12 pm

    From Axel Schuessler:

    If memory serves, 蒲類 (OC *ba-rus) is identical with the transcription of the Mongolian word bars ‘tiger’.

    This word is also, as far as I'm (dimly) aware, all over Turkic as *barıs "leopard", and it was borrowed into Pre-Slavic early enough to give the widespread name Boris there.

  9. Mehmet Oguz Derin said,

    April 17, 2022 @ 1:24 pm

    @David is right, even attested in Old Turkic script corpus as "" in many places (but sometimes taken as leopard too in Turkic languages).

  10. Mehmet Oguz Derin said,

    April 17, 2022 @ 1:25 pm

    (Oops, this blog rejects Old Turkic script Unicode block; that's the ghost word inside double quotes!)

  11. Penglin Wang said,

    April 17, 2022 @ 2:53 pm

    I am very grateful to Professor Mair for his great effort to organize numerous scholarly discussions in Language Log including the present one and appreciate the foregoing postings. I am sorry I forgot to include the following item in “References”: Wang, Penglin. 1995. Tokharian words in Altaic regnal titles. Central Asiatic Journal 39.2:165-207.

  12. V said,

    April 17, 2022 @ 3:03 pm

    David Marjanović : The first recorded use of "Boris" as a Slavic name is after khan Boris converted and made Christianity Bulgaria's official religion and was subsequently canonized. It spread alongside Christianity.

  13. David Marjanović said,

    April 18, 2022 @ 3:37 pm

    Oh, then we need to assume some heavy etymological nativization, but nothing impossible, I think.

  14. Chris Button said,

    April 19, 2022 @ 10:57 am

    Regarding Hittite paršanaš ‘leopard’ and the Mongolian word bars ‘tiger’, it seems that we might be talking about the same etymon.

  15. Penglin Wang said,

    April 20, 2022 @ 3:41 pm

    It may not come as a surprise to learn that the semiotic set of ||||| intrinsic to certain things played a more significant role in triggering preliminary numeracy than thought. In his book “Sino-Tibetan Numeral systems: Prefixes, Protoforms and Problems” (1997:77, Pacific Linguistics) James Matisoff points out: “Several Kiranti languages of E. Nepal have numeral systems where the number FIVE has an explicit morphological relationship with their words for HAND or FINGER”. Likewise, in my view, the Chinese wu (五) ‘five’ and ya (牙) ‘tooth’ could be considered to have once been in cognate relationship in their initial existence. The dialectal pronunciations supported with Old Chinese reconstructions would be a better indicator of this early cognation. Cantonese ŋa and Hakka ŋa ‘tooth’ derived from Old Chinese *ŋa, and Cantonese ŋ and Hakka ŋ ‘five’ were from Old Chinese *ŋo. Thus in Old Chinese, the words for ‘tooth’ and ‘five’ differed in the vowels *a and *o. This difference could have developed in the course of semantic separation between the two notions.

  16. Chris Button said,

    April 20, 2022 @ 5:44 pm

    @ Penglin Wang

    牙 Is generally acknowledged as a Mon-Khmer loan (and can perhaps be ultimately related to 象)

    午 is homophonous with 五 all the way back in the oracle bone inscriptions.

  17. Andrew C Jones said,

    April 26, 2022 @ 10:35 am

    The site http://hanchi.ihp.sinica.edu.tw/ihp/hanji.htm seems to require an IP address in Taiwan. Anyone know a way around that?

RSS feed for comments on this post