I've been troubled by the different vowels in Written Mongolian altan 'gold' and its Turkic source altuun. Why isn't the Mongolian form altun? I found the answer on p. 409 of Claus Schönig's "Turko-Mongolic relations" in The Mongolic Languages (emphasis mine):

The correspondence Mongolic *A-A vs. Turkic *A-U, which can most easily be derived from Bulgharic (and Pre-Proto-Turkic) *A-O, as in Mongolic *altan 'gold' vs. Turkic *alton > *altun. The rules of vowel harmony in both Mongolic and Turkic have secondarily eliminated the synchronic possibility of the combination *A-O, but the comparative data confirm the previous existence of this combination in Turkic. It remains, however, questionable whether the combination also existed in Mongolic at the time of the contacts (*altan < Pre-Proto-Mongolic *alton?), or the borrowings were immediately adapted to the Mongolic rules of vowel harmony* (Pre-Proto-Mongolic *altan < Bulgharic *alton). It is also unclear what the exact chronology of the development *A-O > *A-U on the Turkic side was, for the Turko-Mongolic correspondence *A-U vs. *A-A is also attested in a number of words that may actually belong to a later (Common Turkic) layer of borrowings: e.g., Mongolic *tusa 'advantage, benefit' vs. Turkic *tuso > *tusu. There are indications that the combination *A-O may still have been synchronically valid for at least some forms of Old Turkic (as preserved in the Brahmi and Tibetan scripts).

If I understand this correctly, Mongolic altan is a loan from Bulgharic which preserved the pre-Proto-Turkic vowel sequence *a-o, unlike the rest of Turkic. So why not reconstruct Proto-Turkic *a-o on the basis of Bulgharic? If one reconstructs *a-u in spite of Bulgharic, that implies Bulgharic is a para-Turkic language - a sister of Proto-Turkic that did not undergo the *a-o > *a-u shift:

Pre-Proto-Turkic *a-o
Bulgharic *a-o Proto-Turkic *a-u

But I would rather reconstruct Proto-Turkic *a-o if the combination *A-O was attested in Old Turkic.



alcun 'gold'

has -u- instead of -a- or -o- in its last syllable, so it may ultimately originate from a non-Bulghar variety of Turkic filtered through some other ((para-)Mongolic?) language (Khitan? Xianbei?). The shift of *tu to cu is unknown in Jurchen and must be from some other language in the donor chain.

Tungusic aisin-type words for 'gold' may ultimately originate from yet another non-Bulghar variety of Turkic filtered through some other ((para-)Mongolic?) language (Khitan? Xianbei?). The shifts

*l > i

*c > s

*u > i

must have occurred earlier in the donor chain.

Next: More Jurchen Mountains

*23:10: An early Jurchen Jürcen might have been borrowed into as Mongolian Jürcen 'Jurchen'. It's also possible that a Jurchen Jurcen was, in Schönig's words, "immediately adapted to the Mongolic rules of vowel harmony" as Jürcen.

(The vowel sequence u-e violates Mongolian vowel harmony. u-e is acceptable in Manchu, but I don't know how far back that sequence goes in Jurchen.) HEAVY MOUNTAINS

The Jurchen characters


remind me of Chinese 重 'heavy' and 山 'mountain'. In the northeastern Chinese of the early first millennium AD, 重 was *tʃuŋ and 'mountain' was *ʃan. To Jurchen ears, those syllables sounded like juŋ and ʃan which are close to the syllables of Jin Qizong's (1984: 118, 141) reconstruction dʒuʃiɛn (= my Jurcen) 'Jurchen'.

Jin did not list derivations for either graph.

I think 重 *tʃuŋ 'heavy' is the likely source of the graph for Jurchen Ju (= Ligeti's early Jurchen Jür and later Jurchen Jü), even though I wouldn't expect an *-ŋ graph for an -r syllable.

However, I am not sure 山 *ʃan was the source of the graph for Jurchen šen.

First, the vowels don't match. Not a major issue. 仙 *sien with the phonetic  山 *ʃan had *-e-.

Second, what are the 乂 and ㅗ doing beneath 山? There are other Jurchen graphs with 乂 or ㅗ but I don't know of any other graphs with 乂 atop ㅗ. Are they just arbitrary additions?

Third, if Ligeti's early Jurchen reconstruction cen is correct, why would cen be written with a derivative of a Chinese graph for a ʃ-initial syllable? Why not use a Chinese tʃh-graph as a basis instead?

乂 atop ㅗ is vaguely similar to 尖,now pronounced jian [tɕjɛn] in modern standard Mandarin. Could 乂 atop ㅗ (or even the whole of  山 atop 乂 atop ㅗ) be derived from 尖? No, because 尖 was pronounced *tsiem with a final -m nine centuries ago. That syllable would have been pronounced with a Jurchen accent as jem, not cen. (There was no native Jurchen affricate ts- or dz-. Jin Qizong reconstructed ts- and dz- in loanwords from Chinese.)

12.10.18:00: 乂 atop ㅗ is also vaguely similar to 全 *tshyen, which was nativized in Manchu as ciowan. Its Sino-Jurchen reading would have been similar. But ciowan is still not very close to cen.

12.10.18:44: Could Khitan small script characters serve as phonetics in the Jurchen (large) script? If the Khitan small script character 山 'gold' were pronounced alcun, it might have served as a phonetic for cen. Or 山 'gold' could be  a semantic element atop the phonetic 乂 atop ㅗ derived from 全 *tshyen.

<cen> = something 山 golden that sounded like 乂 atop ㅗ < 全 *tshyen = '(Jur)chen'

One Khitan name for the Jurchen state was


If one changes the 冖 of <GREAT> to ㅗ and combines it with the 乂 of <GREAT> and 山 <GOLD>, the result is ...

Is the Jurchen character <cen> (above) really an altered combination of <GOLD> and <GREAT>? I doubt it. This addendum is just an example of the Rorschach quality of characters without known derivations. You can see almost anything in a character if you look at it long enough.

Here's another wild guess: 乂 atop ㅗ sort of looks like 金 'gold' minus four strokes:

乂 atop ㅗ < 소 < 仝 < 全 < 金

So is <cen> from Khitan small script 山 <gold> atop an abbreviation of Chinese 金 'gold'? Again, I doubt it.

I can only say that neither of the Jurchen graphs for alcu- 'gold'


resemble <cen>.

Most Chinese graphs are derived from other Chinese graphs. Nearly all Tangut graphs are derived from other Tangut graphs. But are there any Jurchen graphs derived from other Jurchen graphs? Similarly, are there any Khitan large script graphs derived from other Khitan large script graphs? Shared components do not entail derivation.

Next: Going An about Gold


Today is the 生日 birthday of the 詩人 poet mark Stratton. (He explains the unusual capitalization of his name in this interview.) 詩人 is literally 'poem-person'. 人 'person' is a stick figure. But what is 詩 'poem'? It is obviously not a drawing of a poem. It is a multiple play on words:

言 is a drawing of a flute. The Old Chinese word ŋan 'flute' was homophonous with ŋan 'words'. One can't draw words, but one can draw a flute.

寺 originated as a drawing of a 止 foot atop a 寸 hand; it meant 'hall' and later meant 'temple'

The Old Chinese word sdəs 'hall' sort of sounded like the word təʔ 'foot', so

止 'foot' vaguely indicated the sound

寸 'hand' vaguely implied the meaning since halls were places where work was done with the 寸 hands. (I'm not sure I buy this part.)

Together,  言+寺 signified something to do with words 言 that sounded like the word 寺 sdəs: namely, stə 'poem'.

One could think of 詩 poems as 寺 temples built with 言 words, though that was not the intent behind the design of the character.

I meant to honor mark by looking at poetry in 契丹 Khitan (a.k.a. Kitan or Qidan), one of the extinct languages that I write about on this blog. But I ended up looking for Khitan poetry instead. I'm very embarrassed.

According to Kane (2009: 190, 214), the epitaph of 耶律迪烈 Yelü Dilie (1026-1092) and the eulogy for 宣懿皇后 Empress Xuanyi (1040-1075) contain "lines of rhymed verse" that are preceded by empty spaces. (Khitan words are normally written without spaces between them.)

The trouble is that (1) empty spaces also precede "certain words to indicate respect" and (2) the rhymes aren't obvious to me.

In line 5 of the Yelü Dilie epitaph, empty spaces precede the word for 'heaven' (pronunciation unknown) twice and in line 16, there is a space before <mu.u.ji> 'sacred'. Those must be respectful spaces. But where are the poetic spaces?

There's a space before <da> (meaning unknown) at the end of line 17. So is <da> the start of a poem? A few of the following lines have possible rhymes that might just be coincidental:

18. ... <ui.en> cf. 24

19. ... <ha.ai> rhymes with 20

20. ... <a.ai> rhymes with 19 (and the 'rhyme' is suspiciously too easy!)

21. ... <?.al>

22. ... <s.ii> cf. 25?

23. ... <cu.ur.bun>

24. ... <ji.en> cf. 18

25. ... <x.ui.ci> cf. 22?


The lines are very long. Would anyone remember what the rhyme at the end of the 41 words of line 18 was? I doubt it.

On the other hand, there are much shorter word clusters in lines 37-38 that are separated by spaces. I thought these might be rhymed lines, but I found only one or two rhymes!

37.end: ... <na.as.bo.ń>

38a. ... <p.od.ún> loose rhyme with 37.end?

38b. ... <a.ai>

38c. ... <tủmu.úr>

38d. ... <a.en>

38e. ... <o.oi>

38f. ... <s.a.ar>

38g. ... <xe.se.ge.l>

38h. ... <tu> rhymes with 38j?

38i. ... (pronunciation unknown)

38j. ... <b.u> rhymes with 38h?

39a. ... <po.qo.ud.u.ji>

The eulogy for Empress Xuanyi has spaces as early as line 2, but they separate parts of the very long title of Yelü Gu, the official who composed the text. (He got to insert honorific spaces in his own description!)

There are other spaces in the eulogy which seem to be random. No rhyming jumps out at me from the text apart from a handful of matches that might be by chance: e.g., <te.xi.is> (line 22) : <c.i.is> (line 25). Do scholars assume the Khitan text must contain rhyming because the Chinese eulogy for Empress Xuanyi contains rhymed text? Am I going about this all wrong? Did Khitan have internal rhyming instead of end rhyming?

It doesn't help that, as Kane (2009: 185) wrote, "much can not" be understood in the epitaph for Yelü Dilie and the eulogy for Empress Xuanyi "can barely be understood at all". Still, in theory one should be able to detect rhymes because these texts were mostly written phonetically. Rhyming lines should contain the same characters or characters with similar readings.

I conclude with a quotation from György Kara (1987):

... it is still a long way to the heart of the jungle of the Khitan writing system [...] much hard work needs to be done before we shall be able to read, for instance, the Khitan rhymes in memory of Empress Xuanyi, the poetess who was put to death nine hundred and eleven years ago (1075).

Next: Heavy Mountains

Then: Going An about Gold LIGETI ON 'JURCHEN' (PART 3)

Ligeti (1953: 224, 226) reconstructed the Jin and Ming Dynasty Jurchen words for 'Jurchen' as

rcen > šen (cf. Manchu Jušen)

This implies that Jin Jurchen rc became Ming Jurchen (and Manchu?) š. I might expect Manchu to have no rc-words, but in fact I found a lot of them in Norman's dictionary: e.g.,

arcambi 'to block'

arcan 'cream'

bedercembi 'to retreat'

borcilambi 'to hang up to dry'

carcinambi 'to congeal'

circan 'a bright yellow pigment'

circinambi 'to freeze on the surface'

Skipping to j-, I found

jurcembi 'to disobey'

Jurcit 'Jurchen'

and even

jurcen (!) 'disobedience' (obviously derived from the root jurce- 'disobey')

How could such rc-words exist? They could be

- borrowings from languages/dialects with rc (e.g., Jurcit is probably from Mongolian Jürced with a Mongolian plural ending; -d is not a possible final consonant in Manchu)

- onomatopoeia (but I haven't seen any examples that fall into this category)

- archaisms (but could there really be so many archaisms?)

- from some other consonant cluster which became a new rc: XX > rc > š (but what would that cluster have been?)

Manchu is not the direct descendant of the recorded Ming Jurchen dialects which have innovations absent from Manchu: e.g., the dialect in the Sino-Jurchen Vocabulary of the Bureau of Interpreters lost g and h preserved in Manchu (Kane 1989: 107-108). Perhaps rc > š occurred in Jurchen dialects other than the ancestor of Manchu which preserved rc. And perhaps it is Jušen that is a borrowing in Manchu from an *rc > š dialect. (The Manchu variant Jusin may be a borrowing from an *rc > s dialect.)

Unfortunately, I cannot find any clear-cut cases of Manchu rc corresponding to clusters in other Tungusic languages in Dybo's database.

Next: A Poetic Pause

Then: Heavy Mountains

And Later: Going An about Gold

ADDENDUM: Janhunen (397-398) proposed *c > š as a sound change that occurred in the Para-Mongolic source of loanwords in Manchu like šanggiyan 'white' (cf. Proto-Mongolic *cagaxan). He also wrote that "the same development is also observed in native Manchu items" though he does not specify any examples.

Dybo's database includes three cases of Manchu š corresponding to c in other Tungusic languages:

Proto-Tungusic *mukca 'mallet' > Neghidal mụkocan 'mallet', M mukšan 'stick'

Proto-Tungusic *koci- > Ewenki kocut-, M hoššo- 'to deceive'

Proto-Tungusic *šimucken (why not *cimucken?) > Ewenki cimcikeen, M šumhun, simhun 'finger'

cf. the correspondence of J alcun to M aisin 'gold' and the M Jušen ~ Jusin 'Jurchen' variation

Kane (1989: 115) mentioned two instances in which the Jurchen dialect in the Sino-Jurchen Vocabulary of the Bureau of Interpreters has š corresponding to Manchu c:

J ešehe : M ecihe 'uncle'

J nuši : M necin 'harmony'

Did *(r)c become š in some Jurchen dialects due to Para-Mongolic influence? LIGETI ON 'JURCHEN' (PART 2)

Before I get to the title topic, Andrew West asked,

I wonder, did <n.i.gu> originally mean "gold" (the metal), and was later borrowed to refer to the "Gold" country (equivalent to the Chinese calling them the 金)? Or was <n.i.gu> originally a Khitan name for the Jurchen, unrelated to gold, and only means "Gold" as an equivalent of Chinesewhen refering to the Jurchen?

I think <n.i.gu> was originally 'gold' or 'yellow' (the color of gold). (I don't know whether the adjective or noun came first.)

In The History of the Liao Dynasty, the Khitan word for 'gold' was transcribed as 女古 *niuku, implying a Khitan nügu. (Did become <i> in the written Khitan dialect? A Khitan *nigu could have been transcribed as 尼古 *niku. See below for the mismatch between Khitan g and Chinese k.)

The transcribed Khitan names for the Xar Moron River (< Mongolian šar mörön 'yellow river') resemble <n.i.gu>. The pure Chinese names for the river were 潢河 which Kane (2009: 165) interpreted as 'yellow river'*. Another Chinese name is 潢水 with 水 'water' (which can also refer to rivers).

Thus I think

<GREAT n.i.gu g.úr> 'great 金 Jin state'

should be literally translated 'great golden state' like its Jin Chinese counterpart 大金國 *tai kim kui.

Kane (2009: 304) also translated the Khitan small script logograms

as 'gold' and 'gold' (masculine). His transcriptions <GOLD> and <GOLD♂> do not indicate pronunciations. I can think of four possible pronunciations:

1. nigu (native Khitan)

2. alcun (as in Jurchen; ult. < Turkic altuun)

3. aisin (as in Manchu; ult. < Turkic altuun)

4. a word unlike any of the above; possibly a second native Khitan word for 'golden' or even a third alt-type word ultimately from Turkic

György Kara (1996: 233) reconstructed** a fifth possibility: jürgü. I presume this reading is based on the Chinese transcription of the Khitan word for 'gold' as 女古:

女 = jür by analogy with 女真 *niu tʃin = Mongolian Jürcen

I'll deal with the *n : J mismatch problem in a later post.

古 = with a front vowel to match the front vowel of the first syllable

But why wouldn't a Khitan be transcribed as Chinese *kü instead of *ku?

If Kara is correct, the Jurchen must have borrowed their j-name** from Khitan because I doubt that the Khitan would borrow their word for 'gold' from their Jurchen subjects.

Now let's look at Ligeti's (1953: 224) reconstruction of the j-name in Jurchen again:


Ligeti: Jüšen (cf. Manchu Jušen)

1. The voiceless-voiced distinction in Manchu was transcribed in Qing Chinese as aspirated-unaspirated. Ligeti and others like Kane and I project this pattern back into Jurchen. Kane and I even think the pattern goes as far back as Khitan.

2.There was no distinction in Ming Chinese between *tʃu and *tʃü. So *tʃu could represent a Jurchen ju or jü. Ligeti may have chosen a front vowel ü because he assumed that Jurchen had vowel harmony that was lost in Manchu:

Jüšen (front vowel + front vowel; harmonic) > Jušen (back vowel + front vowel; nonharmonic)

(The nature of vowel harmony in Jurchen and Manchu is an issue I won't go into here. I want to focus on a single name.)

Jin Qizong (1984: 141) reconstructed the first syllable of 'Jurchen' as dʒu with a back vowel.

3. Ligeti projected Manchu -šen back into Jurchen. There was no syllable *ʃen in Ming Chinese. Modern standard Mandarin shen corresponds to Ming *ʃin, not *ʃen. So Ming *sien was the best possible equivalent.

Jin Qizong (1984: 118) reconstructed the Jurchen syllable transcribed by Ming 先 *sien as ʃiɛn, a combination of the Manchu onset š projected back into Jurchen with the Ming rhyme -iɛn (a more phonetically precise notation than mine; there was no -iɛn : -ien distinction in Ming).

A couple of pages later, Ligeti (1953: 226) reconstructed a Jin Dynasty reading rcen which is identical to the singular of "la forme mongole contemporaine (ǰürčen, pl. ǰürčet)." This reading entails the following sound change(s)****:

Jin Jurchen rc > Ming Jurchen š

or perhaps

Jin Jurchen r before a consonant > Ming Jurchen zero

Jin Jurchen c > Ming Jurchen š

Next: Did rc Become š in Jurchen?

*河 is definitely 'river'. 潢 'pool' is homophonous with 黄 'yellow' in the much more famous 黄河 Yellow River. Since 'pool river' makes no sense, I suspect that the 氵 in 潢河 was added to differentiate that yellow river from the 黄河 Yellow River.

**I've rewritten Kara and Ligeti's reconstructions to be consistent with my Manchu and Mongolian notation.

***"j-name" refers to Jurchen-type names rather than readings of Chn 金 'gold': e.g., modern standard Mandarin Jin.

****12.7.2:42: The sound change rc > š reminds me of *rt > š in Avestan: e.g., A aməšəm : Skt amṛtam 'immortal' (Jackson 1892: 53).. LIGETI ON 'JURCHEN' (PART 1)

There are at least three kinds of names for 'Jurchen'.

1. The n-exonym

The Khitan name

<n.i.gu> 'Jurchen' (possibly with a variant nügu?)

shares an initial with Chinese 女真, now pronounced Nüzhen in modern standard Mandarin, but the second syllables are difficult to reconcile.

2. The a-endonym

The Jurchen called their state the

<amba.an alcu.un guru*.un> = amban alcun gurun 'great gold state'

At present I think Jurchen alcun is indirectly related to Manchu aisin 'gold':

Turkic *altuun Intermediary 1: *altuun > alcun Borrowed into Jurchen dialect ancestral to written Jurchen
Intermediary 2: *altuun > *altın > *aisin Borrowed into Jurchen dialect ancestral to Manchu

In the above scenario, the sound changes that produced J alcun and M aisin occurred in the history of the unknown intermediary languages (Xianbei? Khitan?), not Jurchen and Manchu.

3. The j-endonym

Ligeti (1953: 224) first reconstructed


Ming Chinese transcription: 朱先 *tʃu sien

as Jüšen which is very close to Manchu Jušen. Why didn't he reconstruct a back u for the first syllable? And why is this name so reminiscent of Chinese 女真 Nüzhen?

Next: Ligeti's Quest for Golden Pronunciation LIGETI ON 'GOLD'

Back on November 16, I wrote,

As far as I know, [according to Kane 1989: 350] Ligeti (1953) was the first to reconstruct this [Jurchen] word with -lc- as alcu. I have not seen that article, so I don't know why Ligeti reconstructed -l- instead of -n-. (Acta Orientalia Academiae Scientiarum Hungaricae is not easy to find around here in Nowhere, NJ.)

Thanks to Nathan Hill for finding that article for me. Here's the key section from p. 225 of "Le déchiffrement des 'petits caractères' joutchen":

[Chn transcription] ngan-tch'ouen-wen* »or«: [Jurchen] alču-un, lire alčun (p. 30, nº 568), ma[ndchou = Manchu]. aisin < *alšin < *alčin < *alčun, nanaï** aysin < ma.

(The final -n of Ligeti's alčun is absent from Kane 1989: 350. I've rewritten his č as c.)

Unfortunately, Ligeti did not explain his reason for reconstructing *alcun instead of *ancun (as reconstructed by Jin Qizong 1984) even though the Chinese transcription 安春温 (?) was *anchunwen (in a Pinyin-based romaniztion). A Chinese *-nch- could have represented a foreign -lc- or -nc- since a Chinese *-lch- was not possible.

I used to be worried that Ligeti's -l- was a guess based on -l- in Turkic and Mongolic words for 'gold' rather than on any Jurchen or even Tungusic-specific data. But the -l-transcriptions of Jurchen names for the Ashi river that I discovered yesterday make me more confident that at least some Jurchen dialects had an -l- in 'gold'.

Contra Ligeti, I still don't think Jurchen alcun is the direct ancestor of Manchu aisin, since I don't know of any other examples of the sound changes

J *l > M i

J *c > M s

J *u > M i

However, I still cannot rule out an indirect relationship between two words: e.g., that two words *alcun and *aisin sharing a common ancestor were borrowed into Jurchen from different languages (Xianbei? Khitan?) which may or may not have been related.

Next: Ligeti on 'Jurchen'

*12.5.2:01: Ligeti used the EFEO romanization of Chinese which has ng- as well as zero corresponding to what I reconstruct as a zero initial. Ligeti's ng- is not evidence for initial ng- in Jurchen; I think it reflects later Mandarin dialects that have nonetymological ŋ- in words that once had *ʔ- (like 安, the first graph of the transcription).

**12.5.2:03: Would Ligeti regard all non-Manchu aisin-type words for 'gold' in Tungusic as loans from Manchu?

Tangut fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2011 Amritavision