Reinterpretation of Xianbei qifen ("grass") and its reflection in Mongolic
« previous post | next post »
[This is a guest post by Penglin Wang]
The Chinese transcription of foreign words has made a unique and valuable contribution to our understanding of linguistic situations in early Inner Asia, but it was sometimes inevitably fraught with logographic confusion and scribal errors. Even given quite advanced word-processing and printing in modern times, one can hardly prevent miswriting or misspelling from happening. In ancient China, presumably, it was historians and other authors who heard foreign words spoken and jotted them down, and then further changes developed through the involvement of scribes, typographers, and printers, with each possibly committing their own miswritings and infelicities. It is therefore necessary to reinterpret certain transcriptions on the basis of the known philological and linguistic relevance of what came to be written down.
The choice of the character 俟 (pronounced qi and si in Putonghua depending on its meaning or contextual use) in 俟斤, 俟利發, and 俟汾 corresponding respectively to Old Turkic erkin~irkin (official title), elteber~ilteber (official title), and Written Mongolian ebesü(n) ‘grass’, must be an erroneous substitute for 埃 (ai). I have discussed the wrong use of 俟 in 俟斤 due to its phonetic deviance from the word-initial vowel e~i in Turkic erkin~irkin as a result of the logographic similarity between 俟 and 埃 (Wang 2020). In this short note I want to extend my points to the Xianbei 俟汾.
The earliest known mention of Qifen (俟分) as an ethnonym for one of the twelve hordes of Gaoche (高車) is in Weishu (103.2310) completed in 554 and then in Beishi (98.3273) completed in 659. In their ethnographic account of the Yuwen (宇文) tribe the authors of Xin Tangshu (71.2403) completed in 1060 consider that the ethnonym Yuwen originated in 俟汾 through gradual phonetic changes and try to justify their etymological solution with the meaning of 俟汾 as ‘grass’. The authors imply that Yuwen came forth from China’s Shennong (神農) ‘divine farmer’: “After Shennong was defeated by the Yellow Emperor, his descendants fled to the North for habitation. The Xianbei customarily called grass 俟汾, and since Shennong was able to taste grass, they called themselves 俟汾, which evolved to yuwen through phonetic erosions”.
By reinterpreting 俟汾 to 埃汾 (aifen) and reconstructing the ancient pronunciation of the latter as *eben, we can locate its etymological and cognate reflection in Written Mongolian as ebesü(n)~ebüsü(n) ‘grass, hay, herb’, of which ebe-~ebü- is the root and –sü(n) is the suffix. It is worth noting that the stem ebesü-~ebüsü- is deep-rooted in Written Mongolian and is productive in its derivation: ebesüči(n) ‘mower, hay-maker, seller of grass or hay’, ebesüd- ‘to become weedy, be[come] covered with grass or herbage’, ebesüle- ‘to feed cattle with grass or hay; to feed on growing herbage; to lead to pasture’ (Lessing 1960:287, 291).
References
Beishi (北史 History of the Northern Dynasties) compiled by Li Yanshou in 659. http://hanchi.ihp.sinica.edu.tw/ihp/hanji.htm.
Lessing, F. L, General Editor. 1960. Mongolian-English Dictionary. Berkeley: University of California Press.
Wang, Penglin. 2020. The Turkic influence on Kitan: Kitan bori (the name for an evil person) and Old Turkic böri ‘wolf’. Языки стран Дальнего Востока, Юго–Восточной Азии и Западной Африки: материалы XIV Международной научной конференции, edited by А. Ю. Вихрова, 55-59. Москва: Ключ-С. (See http://www.cwu.edu/anthropology/penglin-wang)
Weishu (魏書 Book of Wei) compiled by Wei Shou in 554.
http://hanchi.ihp.sinica.edu.tw/ihp/hanji.htm.
Xin Tangshu (新唐書 New Book of Tang) compiled under Ouyang Xiu in 1060. http://hanchi.ihp.sinica.edu.tw/ihp/hanji.htm.
Selected readings
- "'Mulan' is a masculine, non-Sinitic name" (7/15/19)
- "An early fourth century AD historical puzzle involving a Caucasian people in North China" (1/25/19)
- "Of jackal and hide and Old Sinitic reconstructions" (12/16/18)
- "Of honey, bee, mead, and Old Sinitic reconstructions" (11/1/18)
- "China and Rome" (2/24/19)
- "Of reindeer and Old Sinitic reconstructions" (12/23/18)
- "Of horse riding and Old Sinitic reconstructions" (4/21/19)
- "What a prehistoric pair of pretty pants can tell us about the spread of early languages" (4/3/21)
Victor Mair said,
August 23, 2021 @ 5:04 am
From Juha Janhunen:
This sounds like a correct interpretation and provides evidence that the Xianbei (or some part of them) spoke (Pre-Proto)´-) Mongolic.
Marcel Erdal said,
August 23, 2021 @ 6:14 am
Perfect connection. It should, also for this word, be noted that there is no Turkic cognate.
David Marjanović said,
August 23, 2021 @ 9:28 am
I'm wondering about the f of fen, though. At the time it wasn't [f], but was [pj] or whatever it was really a good fit for a Mongolic b?
Also, ai for what seems to have been [ə] strikes me as an odd choice (the [ɛ] of modern Khalkh, but not e.g. Chakhar, is most likely an innovation).
Peter B. Golden said,
August 23, 2021 @ 9:31 am
See Andrew Shimunek, Languages of Ancient Southern Mongolia and North China. A Historical-Comparative Study of the Serbi or Xianbei Branch of the Serbi-Mongolic Language Family, with an Analysis of Northeastern Frontier Chinese and Old Tibetan Phonology. (Wiesbaden: Harrassowitz Verlag, 2017), who notes (p.338) Yuwen (a Middle Serbi [Xianbei] dialect): *ǝbun/*ǝbǝn “cognate to Middle Mongol ebesü ~ ebesün, Common Serbi-Mongol *ǝbe-n “grass” and pp. 454-458 with discussion of the *sU suffix, which he views (p.338, n. 213) as a morphological innovation in Pre-Proto-Mongolic distinguishing the Mongolic branch from Serbi.”
Chau said,
August 23, 2021 @ 11:37 am
Thank you for the enlightening post.
"By reinterpreting 俟汾 to 埃汾 (aifen) and reconstructing the ancient pronunciation of the latter as *eben, we can locate its etymological and cognate reflection in Written Mongolian as ebesü(n)~ebüsü(n) ‘grass, hay, herb’, of which ebe-~ebü- is the root…"
Latin herba means 'grass'. Assuming H-dropping occurring in Northern China during the Serbi/Xianbei times, the L. herb- loan might develop into Serbi *eben 'grass'.
Penglin Wang said,
August 23, 2021 @ 3:45 pm
I am thrilled with and highly appreciate all the comments on my post and want to have additional remarks. Chau’s insightful tracing of WMo (Written Mongolian) ebesü(n) ‘grass’ to Latin herba ‘grass, herb’ is well supported by the loss of the Latin initial h in Mongolic. Moreover, the instability of the postvocalic liquid r~l in Mongolic is another factor for us to connect WMo ebesü(n) with Latin herba, i.e., r of herba dropped off in Mongolic as well. Thus, we can reconstruct the root *herbe- for Xianbei and Mongolic. Such reconstructions can have a support from the connection between Old English hyhtan~hihtan ‘to hope, trust’ and WMo itege- (< *hilte-) ‘to believe, trust, rely on, hope’. In this case, the early Mongolic form *hilte- is at present echoed with two variants of Eastern Yugur ļdeğe- or hətege- ‘to trust’, with the former having the initial hə deleted and the latter losing the ļ in the postvocalic position in the initial syllable (Wang 2000).
The ancients disagree on the origin of the ethnonym yuwen. Zhoushu (1.1) completed in 636 holds that the Xianbei customarily called heaven yu (宇) and monarch wen (文), implicitly glossing yuwen as ‘heavenly monarch’. I do not believe yuwen consisting of two morphemes and instead treat it monomorphemic. But I consider the sense of ‘heavenly monarch’ more or less relevant to yuwen. Phonetically, there is no problem at all reconstructing *man for wen of yuwen. And I prefer to reconstruct *ü or *ȵü for yu of yuwen. Thus, we have the form *üman or *ȵüman meaning ‘eight’ and connected with WMo naiman and Baoan nimaŋ ‘eight’. If the former, the initial nasal dropped off.
Why ‘eight’? First, the people in Inner Asia had been accomplished in using numerals in ethnonyms including Naiman meaning ‘eight’. Second, the interested historians seem to be united in their opinion that Kitan was ethnically developed from Yuwen. Third, in its ethnographic and ethnogenetic account of Kitan Liaoshi (32.377) is characteristic of its thick discourse of babu (八部) ‘eight hordes’ or gu babu (古八部) ‘ancient eight hordes’. And finally, at ancient times the Chinese numeral ba ‘eight’ is culturally and symbolically associated with Bagua (八卦) or eight diagrams. According to Zhouyi (周易 Book of Changes), Baoxishi (包牺氏), legendary Chinese progenitor and emperor, created Bagua to comprehend the virtue of divine brilliance and patternize the phenomena of myriad things. The octonary ‘divine brilliance’ here is compatible with ‘heavenly monarch’. That is why I am arguing for more or less relevance of ‘heavenly monarch’ to yuwen. This means: I do not agree with the proposition of the authors of Xin Tangshu to connect qifen with yuwen; qifen should be aifen; WMo ebesü(n) is cognate to Xianbei aifen; WMo naiman ‘eight’ is again cognate to yuwen.
References
Liaoshi (遼史 History of Liao) compiled by Tuotuo in 1344. http://hanchi.ihp.sinica.edu.tw/
ihp/hanji.htm.
Wang, Penglin. 2000. Lexical connections between Germanic and Mongolic. Interdisciplinary Journal for Germanic Linguistics and Semiotic Analysis 5.1:71-91.
Zhoushu (周書 Book of Zhou) compiled by Linghu Defen in 636. http://hanchi.ihp.sinica.edu.tw/
ihp/hanji.htm.
Zhouyi (周易 Book of Changes). https://ctext.org/text.pl?node=46944&if=gb&show=parallel&remap=gb.
Alexander Vovin said,
August 23, 2021 @ 7:06 pm
I am sorry, but your logic for replacing 俟 with 埃 here and in your 2020 article you refer to escapes me. Ditto for the same replacement that Shimunek suggested as Peter pointed out. What is the evidence besides the necessity to connect Mong. ebe(sü)n with this word? But I trust we might have an additional problem here: The book 103 of Wei Shu was largely reconstructed n the basis of the book 98 of Bei Shi by the Song period scholars. And this means that we cannot use LHC reading Ɂə of 埃, which is sorely needed for Mong. ebe(sü)n ( e = ə). At the earliest, it is EMC Ɂậi, that you mention, but it would underlie OT *ärkin, not ėrkin > erkin ~ irkin (leaving aside the problem whether it is Turkic, Mongolic, etc. title).
Also MM h- is quite stable. MM ebesün is amply attested in both EMM and WMM, while MM *hebesün is not attested at all.
Penglin Wang said,
August 23, 2021 @ 11:01 pm
Although I respect phonologists’ efforts to reconstruct ancient Chinese words as homogeneously as possible, in my discussion of the erroneous logographic substitution I rely on the bona fide established pattern of Chinese transcription of foreign words. In the case of the Chinese 埃 (ai), we have recurring and established pattern for 埃 to correspond to the foreign e: Eddington 埃丁頓, Edgar 埃德加, Edmund 埃德蒙, Edwin 埃德溫, Eleanor 埃莉諾, Emma 埃瑪, and so on and so forth. However, the pronunciation of 埃 is definitely not homogeneous in Chinese: Putonghua ai, Cantonese āi, oi, ai, Hakka ai, yai, and Wu ei, a, e, ɛ. Hypothetical reconstruction often in isolation and out of context pales in significance in the face of such a solid pattern. It will be wrong if the future phonologists try to reconstruct a single homogeneous form for 埃. On the other hand, it will be preferable if they reconstruct ai for Putonghua by recognizing the existence of regional dialects.
It is, moreover, irrefutable that such dialectal differences existed throughout history. Confucius instigated the ideology of yayan (雅言) ‘gracious speech’ and became a role model in using it in education and ceremonies. Mencius once ridiculed the speech of southern dialect speakers as shrike twittering. It is my understanding that the Confucian concept of ‘gracious speech’ refers to one mainstream dialect which rose to the level of a common language serving as a cross-dialectal tool for communication and which eventually evolved to be what is today’s Putonghua. And Mencian ‘shrike twittering’ is a piece of strong evidence for the dialectal differences in ancient China.
The attestation of the initial fricative h in Middle Mongolian is in itself indicative of its continuation from and does not preclude its loss in earlier Mongolic. With increases in transcontinental flow of human population and exchange of ideas in many parts of ancient central Eurasia, it is becoming increasingly more difficult to think of vocabulary of a language as an isolated, insulated entity. In this sense, Chau’s idea of connecting Latin herba with Xianbei *eben ‘grass’ could be a clue to the loss of word-initial h- in Xianbei and Mongolic *herbe- ‘grass’.
I thank Professor Golden in drawing my attention to the reconstruction of yuwen in the form of *ǝbun/*ǝbǝn ‘cognate to Middle Mongol ebesü ~ ebesün, Common Serbi-Mongol *ǝbe-n ‘grass’, but I do not see ‘Ditto for the same replacement’ in his posting. I will check this issue later.
Andreas Johansson said,
August 24, 2021 @ 1:06 am
Also MM h- is quite stable. MM ebesün is amply attested in both EMM and WMM, while MM *hebesün is not attested at all.
MM h- may be stable, but Imperial era Latin h- is not, so I don't think that's a problem.
Why grassslands-dwellers in the far east would borrow a word for "grass" from the far west is another question of course.
David Marjanović said,
August 24, 2021 @ 3:46 am
What Andreas Johansson said.
I mean, it's possible that the speakers of Mongolic didn't "always" live in grasslands, and that, when they entered that landscape, they borrowed the vocabulary to describe that landscape, maybe even something as utterly basic as "grass", from a language that was already there. The list of choices for such languages is pretty long – but why in blazes would it include Latin? Likewise, how would a Germanic word get there that early, and why?
What happened to vowel harmony?
This pattern, quite reasonably, uses [æ] – the pronunciation in ai in many even near-standard Mandarin accents – to approximate foreign [ɛ].
But early Mongolic, by all appearances, had no such sound at all, neither [æ] nor [ɛ] nor for that matter [e]. What Mongolists traditionally denotate as e appears to have been [ə], which it still is in about half of the Mongolic language family today, and which fits the theoretical expectations for a vowel-harmony system based on neutral vs. retracted tongue root (as observed today in all Mongolic languages that haven't lost vowel harmony altogether).
This is why Prof. Vovin just commented that the Late Han Chinese "reading Ɂə of 埃 […] is sorely needed for Mong. ebe(sü)n ( e = ə)".
The reasons for the traditional denotation as e are as follows:
1) The modern Mongolic idiom most Mongolists are most familiar with is Khalkh, which has shifted [ə] to [ɛ] (and spells it accordingly). As part of the same shift, it has also shifted [o] to [ɵ] – but, IIRC, just the short one; the long one is still [oː], which is noteworthy because vowel length is a pretty recent development that is still lacking in Written Mongolian.
2) The vowel-harmony systems most phonologists are most familiar with are those found in and around Europe: the backness-based systems of Uralic and Turkic languages (recently lost in some of them without replacement, e.g. Estonian and Uzbek). The systems found in Mongolic (and most of Tungusic, and other languages of the region) are different, but similar enough that most Mongolists preferred to assume they were derived from a backness-based system by a few very odd sound shifts. This made it possible to assume that early Mongolic had practically the same vowel system as (early) Turkic. (This assumption used to be so widespread that it underlies the Cyrillic orthography of Khalkh and Buryat, and almost every Latin transcription I've ever seen.) Only in the last ten years has RTR-based harmony really been understood.
3) The availability of e on typewriters. It's really hard to write ə on a typewriter designed for English, French or Russian: take the paper out, put it back in the other way around, finagle with the typewriter for ten minutes to find the exact spot, type e…
4) In order to maintain backwards-compatibility, historical linguists tend to be very conservative in their choice of notation; compare the continuing use of *h₂ by Indo-Europeanists despite widespread agreement that it was a plain old [χ].
(There is also a widespread myth in historical linguistics that sounds cannot be reconstructed anyway, only abstract phonemes – so if there's no evidence for a contrast between [ə] and [ɛ] or [e] in the language you're reconstructing, you can just write e without misleading anyone. That's true sometimes. In other cases, it is easier to reconstruct sounds very precisely than to figure out which phonemes they belonged to.)
Chris Button said,
August 24, 2021 @ 8:28 am
@ Alexander Vovin
埃 is EMC Ɂəj (LMC Ɂaj) in Pulleyblank’s reconstruction. I wonder what might have been more appropriate? Incidentally, the -j coda that remains unaccounted for in many Old Chinese reconstructions comes from an earlier velar glide coda in his system.
Penglin Wang said,
August 24, 2021 @ 12:48 pm
Wonderful discussions.
Andreas had made a very good point concerning the phonetic variation of h- and raised again a very good question: Why grassslands-dwellers in the far east would borrow a word for ‘grass’ from the far west? There are two schools of thought how to deal with the similarities in languages. One school holds a genealogical approach and organizes language families. The late Professor Joseph Greenberg put forward a very large family called the Eurasiatic Language Family including Indo-European and Altaic among others. Given this approach, many words of similar phonetics and relatable meaning found in Altaic and Indo-European could be considered as cognates. So, we do not need to bother ourselves who borrowed from whom and why people borrowed words from other languages afar. The other school treats linguistic similarities as the result of language contact. I prefer to use this approach. In my article (Wang 2000) I argued:
"My assumption is that the historical long-term contact between Altaic and Indo-European in the Eurasian steppes could lead to repeated circulation of basic words. Ancient vocabulary was not as complex and abundant as today’s. Moreover, the proportion of basic words in ancient vocabulary was significantly larger than today’s. For example, if basic words represent fifteen percent of today’s total vocabulary, they were eighty percent of the ancient one. So what circulated were mostly basic words."
The ethnonym Yuwen can be traced back to the third century according to Weishu (1.5) recording the Bigman Mohuan (莫槐) of the Xiongnu Yuwen tribe. Mohuan lived in the third century up to the year of 293. I have no knowledge of if vowel harmony existed in the language of those who initiated the name yuwen. Suggested by David, I shall now consider the vowel harmony to adjust my hypothetical reconstruction of *üman or *ȵüman to *ïman or *ȵïman for yuwen.
How would a Germanic word get there that early, and why? Because the Xiongnu people in Inner Asia in general and in historical Mongolia in particular were interacting with their western neighbors – the Yueshi and Wusun groups speaking Indo-European languages, the former of which, according to some researchers’ suggestion, was what would become the Tokharian, we must assess the impacts that the Xiongnu and Yueshi were having on each other in spite of their confrontation with each other. From that point it was a natural development to follow vocabulary pooling into each other’s language to see how the two languages interacted. The Altaic groups essentially inhabited the Xiongnu territory and could be componential parts of Xiongnu. According to Douglas Adams, the Tokharian and Germanic share lexicon (by memory). Furthermore, Ronald Ringer with a coauthor argue for Indo-European homeland in the area lying between the Black Sea and Caspian Sea (by memory, I’m responsible for any inaccuracies). This area is very close to Central Asia and henceforth easy for Indo-European to radiate in the eastern, Altai direction. Some people see the Altai mountains as natural barriers for transcontinental exchange as if they insurmountable. In reality, according Russian archaeologists, there was crowded traffic in the Altai mountainous area at ancient times, and some people built up walls to block the traffic (by memory). Apparently, many early people saw mountains as connections between peoples and wanted to explore the other sides of the mountains.
David Marjanović said,
August 24, 2021 @ 2:30 pm
Oh, I'm sorry, this is a fundamental misunderstandings of both the "long-rangers" and the "splitters".
Both schools agree that the null hypothesis, the default explanation, for any similarity between any languages is random chance. Only when chance is shown to be a very improbable explanation for a similarity can we look to other explanations and try to test them against each other and against the null hypothesis.
Latin [(h)ɛrba] and Proto-Mongolic *[əb̥ə(su)n] are barely even similar. Why do you think what little similarity they have requires any explanation at all?
If that language was Mongolic or closely related to Mongolic, the simplest hypothesis is that it had RTR-based vowel harmony. This is both because this is so widespread (if not universal) among Mongolic languages today and because it is so widespread among other language families of northeastern Asia that it must be quite old in that area.
Not more than any other two branches of Indo-European. Tokharian and Germanic are not closely related with Indo-European either, and they have literally never been in direct contact (Iranian may have been in direct contact with both, but evidence for contact with Germanic before the Alans of the 5th century is very sparse at best).
Donald Ringe.
Yes, this is the current consensus (at least for the last common ancestor of Tocharian and Indo-Actually-European, not necessarily for the last common ancestor of that and Anatolian).
But Germanic formed only after an IE expansion in the opposite direction (northwest). There is ample loanword evidence showing that Germanic and Finnic have been in contact for thousands of years.
In short, I expect future research (assuming past research hasn't done that already) to uncover plenty of Tocharian and Iranian words to show up in the reconstructed lexicon of Proto-Mongolic, and perhaps vice versa (even if these words were perhaps passed on by intermediaries). But Germanic or Italic – I see no reason to consider this in the continued absence of much stronger evidence than you have yet provided.
Penglin Wang said,
August 25, 2021 @ 9:32 am
This is an addendum to my posting: Shimunek (2017:338) identifies the substitute of ‘俟 for 矣’ and reconstructs the form *əbun/*əbən for Xianbei [俟汾] ‘grass’ cognate to Middle Mongol ebesü ~ ebesün ‘grass’.
I am sorry that Shimunek’s book escaped me when I posted my note. I thank Professor Golden for drawing my attention to Shimunek’s findings and Professor Vovin’s follow-up reminder of ‘Ditto for the same replacement’. I apologize for my misunderstanding of their points. (Shimunek, Andrew Eric. 2017. Languages of Ancient Southern Mongolia and North China: a Historical-Comparative Study of the Serbi or Xianbei Branch of the Serbi-Mongolic Language Family, with an Analysis of Northeastern Frontier Chinese and Old Tibetan Phonology. Wiesbaden: Harrassowitz Verlag).