In my last entry, I proposed that Middle Korean (MK) hwar 'bow' might be related to Khitan and Mongolic words for 'to shoot'. Martin (1996: 95) mentioned two other etymologies:

lYu lYel* 1990: 160-1, while noting the [雞 林] Kyeylim (#300) phonogram [活] ˙HHWALQ [for Korean 'bow'], would reconstruct an earlier doublet *pala/pele, associating it with a verb 'spread, stretch' on the basis of the use of the verb for the noun in very early itwu [= idu; no source cited**]. He says the Jurchen word for 'bow' is bere, but that should be amended to beri (Kane 1989: 251). Yu's derivation would be

*pala > *hala > *hal[a] > hwal,

much like (later)

han soy > hwan soy > hwāng-say [= my hwāngsae] 'stork'


˙han ˙sywo > hwang-so 'bull'

I don't know whether Yu thought hwar was borrowed from Jurchen

beri 'bow'

or some Tungusic cognate of it, or if he regarded Jurchen beri and MK hwar as reflexes of a Proto-Koreo-Tungusic (or Proto-Altaic?) word for 'bow'.

In any case, the shift of *p to h is unknown in both Korean and Jurchen/Manchu. p weakened to f (not h) in Ming Jurchen after hwar was first attested in Korean during the Song Dynasty.

Moreover, the -w- in the first syllables of hwāngsae 'stork' and hwangso 'bull' might be due to rounded vowels in the second syllables***, whereas there is nothing labial in *hal(a) to condition a glide after *h-.

I assume this next etymology is Martin's, though I am uncertain:

Another possibility: [弧] HHWO 'wooden bow' (+ [發] ˙PELQ**** 'release an arrow', or the northern Chinese suffix [兒] -r?).

I can't think of an example of a Sino-Korean noun-verb combination like 弧 becoming a noun in Korean. If 弧發 is the source of hwal, the *-op- must have contracted to  -w-:

*hopar > *hobar > *hoβar > hwar

Are there any other examples of that change?

I also don't think is likely since that would have led to MK *hor or *hur (if borrowed after shifted to *xu in northern Chinese), not hwar. Does any northern Chinese dialect have a word 弧兒 *hur 'bow'?

*8.5.19:20: Martin romanized the name 류렬 <r.yu r.yŏ.l> as lYu lYel (= my Ryu Ryŏl) with superscript l to indicate that the initial liquids are silent in the South Korean version of standard Korean pronunciation.

**8.5.19:39: *pala/pele 'spread, stretch' corresponds to *para/pŏrŏ in my romanization.

The only Korean para- I know of is 'wish'.

Ryu's *pŏrŏ is presumably ancestral to Kor pŏr- 'spread' (coincidentally resembling Old Japanese pirə- 'wide'?).

The "very early itwu" that Ryu had in mind might be the early spellings of the Paekche names for 羅州 Naju on p. 352 of his 1983 book 세 나라 시기 의 리두 에 대한 연구: 發羅 ~ 保羅 which he interpreted as *para.

***8.5.21:36: The derivations of 'stork' and 'bull' contain some unusual changes.

Yu (1964) does not list any MK sʌ̀y (= Martin's soy) 'bird'. The earliest attested form of 'stork' in Yu (1964: 744) is MK hàn sǎy with sǎy 'bird'. I would expect hán 'great' instead of hàn '?'. Is the absence of a dot indicating a high tone an accident?

MK hán syó (= Martin's ˙han ˙sywo), on the other hand does have the expected hán 'great' before syó 'ox'.

The modern Cheju vowel corresponding to MK ʌ is rounded ɔ. Martin's han soy (= my hàn sʌ̀y) could have become hàn sɔ̀y and the -w- before h- could be due to assimilation with the ɔ in the following syllable.

The long vowel of modern Korean hwāngsae 'stork' should be from an MK rising tone, but the first syllable of MK 'stork' had a low tone.

The change of MK -n to -ng before s- is unexpected.

I think native hàn and hán could have been replaced with a Sino-Korean hwang. Son and Yasuda (1989: 2066)'s dictionary lists hwangso as a Sino-Korean hybrid 黃소 'yellow ox'. But it lists hwāngsāe with two long vowels as a native word. I hypothesized that hwāngsāe might be from 皇 hwang 'emperor' plus 'bird'. However, 皇 is hwang with a short vowel, and the only hwāng in Martin et al. (1967: 1864) is 況 hwāng 'all the more/less' which would not make any sense before 'bird'. Could the long vowel of hwāng be due to confusion with a Sino-Korean hwān? None of these hwān in Martin et al. (1967: 1859) make any sense before 'bird' and none are prefixes:

幻 'deception'

宦 'government official'

患 'trouble'

換 'exchange'

逭 'escape' (so rare it's not even in Microsoft's Korean IME!)

****8.5.21:39: Martin wrote prescriptive Sino-Korean readings from 東國正韻 (1448) in capital letters. The reading ˙PELQ = my pə́rʔ for 發 in 東國正韻 does not match the later attested reading pal which corresponds to the alternate reading párʔ = his ˙PALQ in 東國正韻

I don't know why Martin didn't derive MK hwar from his 東國正韻 弧 HHWO + 發 ˙PALQ which wouldn't require a vowel change of E to a. BIRCH BOW OF POTTERY VILLAGE (logonotes*)

Last night I was skeptical about Song Chinese 晝里 *tʃiu li being an error for 畵里 *xwa li, a transcription of Khitan qori or xori 'twenty' (cf. Written Mongolian qorin 'twenty' and Para-Mongolic *xorïn 'twenty' in Janhunen 2003).

Today Andrew West pointed out a more solid case of SC *xw corresponding to WM q in 遼 史 53:

SC 陶里樺 *thau li xwa 'rabbit shooting' : WM taulai 'rabbit', qarbu- 'to shoot'

He also noticed that Shimunek (2007: 48) had also identified 樺 as Khitan X(w)a-, a cognate of WM qarbu-.

I hypothesized that pre-Khitan *qo might have become qwa in 'twenty', but that begs the question of where Khitan got its

<qo> and <qó>

from: either something shielded them from the shift or they once had different vowels and participated in a chain shift.

*qV > qo > qwa

Does the Khitan word for 'shoot' (qwa? xwa?) go back to *qo (even though its WM cognate lacks -o-), or does it contain a primary -w- lost in Mongolic?

Common ancestor of Khitan and Mongolic Khitan Written Mongolian Example word containing this syllable
*qo qwa (merger of qwa with secondary -w- from *qo and qwa with primary -w- from *qwa) qo 'twenty'
*qwa qa 'shoot'

Khitan qwa- (or xwa-?) and WM qarbu- 'shoot' remind me of Korean hwal < Middle Korean hwar 'bow' which was first definitely transcribed as 活 *xwar in 雞林類事 (1103-1104 AD)**. Korean hw- is rare in native words, and Middle Korean (MK) hwar has a low pitch instead of an expected high pitch. Could MK hwar be a loanword from some Mongolic-type language? Was there a final -r in Khitan 'shoot' that was not transcribed in Chinese?

Starostin's database links hwal to other 'Altaic' words with s-: e.g., Old Japanese sa 'arrow'. But I think there may be two different word families:

- the q-family:

WM qarbu- 'shoot'

Khitan Xwa(r)- 'shoot'

Kor hwal 'bow'

- the s-family:

WM saɣali 'crossbow'

Kor sal < MK sar 'arrow' (transcribed in 雞林類事 as 薩 *sar)

OJ sa 'arrow' (borrowed from a Paekche cognate of MK sar?; the unrelated native Japanese word is ya)

The two families are united in Kor hwasal 'arrow' < Middle Korean hwar sar 'bow [and] arrow'.

Unlike Starostin, I do not regard this shared terminology*** as evidence for an Altaic family (as opposed to an Altaic convergence zone).

Next: Two More Etymological Proposals for Korean Hwal 'Bow'

*8.4.1:10: The title refers to the meanings of the transcription characters


'pottery village birch'

and a possible connection between the Khitan word for 'shoot' represented by 樺 'birch' and the Korean word for 'bow'.

Wikisource has the spelling 陶 裏樺 with 裏 'inside' instead of 里 'village' (8.5.19:23: probably due to automated simplified to traditional conversion), but I have chosen to follow the text as presented in Academia Sinica's Scripta Sinica database.

**8.4.23:38: The dialect of Chinese in 雞林類事 has a final *-r from Middle Chinese *-t, whereas the dialect of Chinese reflected in the Khitan small script had shifted that *-r to a glottal stop in poetry (which might have been lost in everyday speech; see Kane 2009: 252).

***8.4.23:46: I have excluded Starostin's Turkic cognates for MK hwar because they are not attested before Middle Turkic and they lack -r. (8.5.00:10: I can't find an entry for this root in Clauson 1972.)

Starostin's proposed Proto-Tungusic cognate *sug- has an *-u- that doesn't match the -a- elsewhere in Altaic and its meaning is broad: 'spear; fish fork; a kind of knife' as well as 'arrow'. Perhaps it meant 'sharp tool'. YOUNG SHEEP KHAN?

In my last post, I proposed that


in 契丹國志 represented something like *juR(I)HUn qan 'twenty khan' (singular!): i.e., a khan associated with the number twenty. (I've rewritten all the uncertain segments in capital letters.) Since a Para-Mongolic word like *juR(I)HUn is the source of Jurchen

jirhon 'twelve' (not 'twenty'!),

I ended up hypothesizing that the same sequence of the same two morphemes referred to two different numbers in two different Para-Mongolic varieties: written Khitan and the source of loans in Jurchen (eastern Khitan? a non-Khitan but still Para-Mongolic language?).

The written Khitan word for 'twelve' would be 'ten-two' (as implied by the large and small scripts) rather than 'two-ten'. This situation would be like that of English twenty-one which corresponds to Dutch eenentwintig and German einundzwanzig which have 'one' and 'twenty' in the opposite order even though they are both related to English. (Of course English has no equivalent of en and und 'and' in twenty-one.)

Today Andrew West proposed that Song Chinese 昏 *xun could represent the Khitan cognate of Written Mongolian qoni(n) 'sheep'. Phonetically *xun is a near-perfect fit for a Khitan qon. (An exact match is not possible since Song Chinese did not have a syllable *qon.) Andrew suggested that Khitan

<iam.a> 'sheep'

is a borrowing from Chinese 羊 'id.' which was *yɨaŋ in Middle Chinese*.

However, that leaves SC 晝里 *tʃiu li unexplained. It can't be 'two', though it resembles Janhunen's (2003: 16, 399, 400) Proto-Mongolic *jiri/n 'two' and Khitan cur ("more probably jur or jir") 'two'.

I thought it might be some unrelated (near-)homophone like jul- describing the 昏 sheep. Andrew suggested that it was the Khitan cognate of Written Mongolian juljaɣa(n) 'young of an animal' and juljaɣala- 'bring forth young' and interpreted 晝里昏 as 養羊 'sheep-raising' since 契丹國志 mentioned that 晝里昏呵 had 養羊二十口 'raised twenty sheep'.

What if the reading of the Khitan small script character

'young' (the Chinese 子 'child' in the graph is probably not a coincidence)

was jul? Would 晝里昏呵 have been written as


in the Khitan small script?

I then reconsidered my 'twenty' hypothesis. What if 晝里昏 were 'twenty sheep' with haplology?

*jurqon-qon > jurqon?

Andrew found another explanation in Kane (2009: 35) that I had forgotten:

Ji Shi suggested that 晝 zhou [the modern Mandarin reading] might be a mistake for 畵 hua [another modern Mandarin reading], thus giving the name as 畵里昏阿 hua-li hun-a [still more modern Mandarin; note the different final character]. This is reminiscent of Mo. [= Mongolian] qori ~ qorin 'twenty' and qoni ~ qonin 'sheep'.

If Ji is correct, then Khitan and Jurchen may share a word for 'twenty' after all. But it's still not certain whether Khitan had 'two-ten' or 'ten-two' for 'twelve'; perhaps

'ten-two' (large script; 北大王墓誌 6.28-29)

'ten-two' (small script; 興宗皇帝哀冊 5.61-62 - note how the 'small' script characters have more strokes than their 'large' counterparts)

were read as 'ten-two': cf. how Dutch and Germans read 21 as 'one-and-twenty' even though the 2 comes first in writing.

The one odd thing about Ji's interpretation is that a Chinese consonant cluster *xw- corresponds to Written Mongolian *q- and presumably a single consonant in Khitan (since there is no evidence that Khitan had initial clusters). Could *qo have broken to *qwa in Khitan? But if Khitan had Cw-clusters, surely we would have seen them in the small script by now unless the characters for *Cw(V) all happen to be unidentified.

*8.3.2:36: Given that the Khitan borrowed a Chinese form like 龍 *lyuŋ 'dragon' as


<lu> (large script) ~ <lu> (small script)

perhaps 羊 *yaŋ was borrowed as ya, and -ma is a suffix.

The simplification of *ly- to l- in 'dragon' is expected if Khitan had no initial consonant clusters.

Kane (2009: 67) also transcribed

as <êm.a>, and if this transcription is correct, then it is unlikely that this Khitan word for 'sheep' is a borrowing from Chinese unless *ya became ê in Khitan. HEAD, PIG, TWENTY: THE THREE KHANS?

The 契丹國志 (Record of the Khitan State compiled by 葉隆禮 Ye Longli during the Southern Song; complete text) mentions three legendary early chiefs of the Khitan:

廼呵, who "was nothing but a skull"

喎呵, who "wore a boar's head and was clad in pigskin"

晝里昏呵, who "had raised twenty sheep" that would magically regenerate after he ate nineteen of them

(Translations from Xu Elina-Qian 2005.)

Last night I realized that the three names share a common structure if 晝里 Song Chinese *tʃiu li is ignored. The first half might correspond to the distinguishing characteristic of each ruler:

廼 Song Chinese *nai transcribing Khitan nai 'head'? 呵 Song Chinese *xɔ transcribing Khitan Qo '?'
喎 Song Chinese *(kh)wai transcribing Khitan uil 'pig'??
昏 Song Chinese *xun for transcribing Khitan QUN 'twenty' (cf. Written Mongolian qorin)??

Capital Q represents q or x. Song Chinese *x could be a transcription of Khitan q (a consonant absent from Song Chines) as well as Khitan x.

Capital U represents u or o and capital N represents n or r. Song Chinese *-un could be a transcription of Khitan -un or Khitan rhymes absent from Song Chinese: -on, -ur, -or.

I am most confident about nai 'head'*. However, Song Chinese *(kh)wai is not a good match for uil 'pig'**, and I know of no other evidence for reconstructing the Khitan word for 'twenty'***.

Tonight it occurred to me that if these transcriptons were very old, 呵 could have been originally meant to be read as Middle (i.e., Tang) Chinese *xa(h) which might have been a transcription of Khitan qa 'khan' (as reconstructed by Kane 2009: 103). So the names would be titles: 'Head Khan', 'Pig Khan', and 'Twenty Khan'.

Was Song Chinese ɔ was low enough to be acceptable as a transcription of Khitan a? If not, was there a Khitan noun or suffix -qo or -xo?****

*8.1.2:19: I had forgotten that Kane (2009: 101) mentioned 廼呵 in his entry for Khitan nai 'head' without comment. It is not clear whether he thinks 廼呵 means 'head + X'  or  something like 'leader'.

**8.1.2:33: Kane (2009: 66) transcribed the Khitan small script character



'matter, affair'

the transcription of Liao Chinese 衛尉 *ui and *-ui in 國 *kui and 隨 *sui

as <ui>, noting that

Some scholars have suggested it is read uil or uile, with which one can compare Mo. and Ma. weile 'affair, business'. Cf. Ju. ulgian 'pig', where -gian is a nominal suffix, perhaps leaving the stem as *ul-.

He then suggested that the dot of

also used for the transcription of Liao Chinese *ui could "perhaps [be] indicating a reading slightly different from" the dotless version.

I think the dotless version might have originally been read uil, whereas the dotted version was read ui without -l. However, the two could have been confused over time.

***8.1.3:00: I now wonder if Song Chinese 晝里昏 *tʃiu li xun represented a Khitan word like *jur(i)hun resembling Janhunen's (2003: 16, 399, 400) Proto-Mongolic *jiri/n 'two', Khitan cur ("more probably jur or jir") 'two', and Pre-Proto-Mongolic *jï.r.ku/n (> Para-Mongolic *jir.hon) 'twelve' (not 'twenty'!). PPM *-ku/n > PM *-hon was a '-teen' suffix. Did Khitan have very similar words for 'twelve' and 'twenty': i.e., 'two-teen' and 'two-ten'?

Unfortunately, the Khitan spellings for 'twenty' give no clues to pronunciation:

large script 廿 (as in Chinese; non-Chinese-like variants:


small script 丁 (二 'two' with the bottom stroke rotated?)

8.2.00:06: However, Janhunen (2003: 400) wrote,

The Khitan Small Script would suggest that the teens were expressed in terms of additive compounds (10 + digit), rather than as lexicalized derivatives.

Perhaps Khitan had 'two-ten' for 'twenty' and 'ten-two' for 'twelve', unlike the Para-Mongolic source of Jurchen


jirhon 'twelve' (< 'two-ten') and orin 'twenty' (< Para-Mongolic *xorïn; not related to jir- 'two'; perhaps not even related to Proto-Mongolic *koxar ~ *koyar 'two'; Janhunen 2003: 397 regarded Proto-Mongolic *korïn 'twenty' as "irregular" and "perhaps influenced" by *koyar)

Did Jurchen borrow from a non-Khitan Para-Mongolic language or from a nonstandard (eastern?) dialect of Khitan?

****8.2.1:30: The only Khitan QU-suffix I know of is an adjectival suffix


<qu> (f.?) ~ <qú> (m.?) (Kane 2009: 48, 64)

Could this also appear in nouns?

I thought

might be a noun suffix <qó> on the basis of two Chinese transcriptions of the name of a Khitan tribe:

阻卜 *tsu pu (short form?; corresponding to a hypothetical <c.bu> in the Khitan small script)

朮不古 *tʃu pu ku (long/suffixed form?;

<c.bu.qó> in the Khitan small script)

but now I think both transcriptions might have represented a Khitan cubuq. (I reconstructed the first vowel on the basis of the Chinese transcriptions.) There was no final *-q or even *-k in Liao Chinese, so a scribe could either ignore Khitan -q (e.g., in 阻卜) or represent it with a graph for a Chinese *kV-syllable (e.g., in 朮 不古). I am assuming that <qó> could also represent <q>, though there is no other evidence for this. It is simpler to assume that 阻卜 is an incomplete transcription.

Another example of <qó> is

<ja.ri.qó> 'prime minister'

which has no short form <ja.ri> and which may be cognate to Written Mongolian jerge 'rank' (Wang Hongli 1986 as reported in Kane 2009: 52). If Khitan had prototypical Altaic vowel harmony, 'prime minister' should be jarïqó with back vowels - yet WM jerge has front vowels.

A case of the reverse correspondence may be 'dog':

Khitan <ńi.qo> (in Kane's transcription) : WM noqai.

This reminds me of the correspondence between Ukrainian i and Russian o: e.g., U nich : R noch' 'night'.

But Kane (2009: 93) also lists others' reconstructions of 'dog' that lack front vowels in the first syllable: nəxi, noxi, nəxəi, njaqa, ñaxa, ñaha, ñaqa.

The problem of reconciling Khitan and Mongolian vocalism will have to wait until Khitan phonology is better understood. Khitan appears to have had at least five vowels in native words. Kane's diacritics indicate potential vowel distinctions: e.g., <ó> differed from <o> in some as yet unknown way. The reconstruction of Khitan vowels is complicated by their occasional absence from the small script and the limitations of Chinese transcription data: i.e., the restricted number and distribution of Chinese vowels. 'LAB-IEW' SYLLABLES IN TANGUT (PART 2)

I covered the first of two PIw syllables in my Tangut reconstruction in part 1. The only other PIw syllable is


which Homophones defined as a 'person's name'. Have any Tangut people with this name been found? The only nondictionary attestation in Li (2008: 553*) is in this transcription of Sanskrit through Tangut period northwestern Chinese (TPNWC)**:

ʔa 2dəəu 1tõ 1lo < TPNWC 阿耨多羅 *ʔa ndəu to lo < Middle Chinese *ʔa nəuh ta la < Skt anuttarā 'supreme'

1sã 2mieu 1sã 1po 1tiẽ < TPWNC  三藐三菩提 *sã mieu sã phəu thi < Middle Chinese *sam mieuʔ sam bo de < Skt samyak sambodhi 'correct enlightenment'

I don't mind 2miew as an exception to my prediction that Tangut would have no PIw syllables if it represents

- foreign syllables (Chinese or Sanskrit through Chinese) in languages without a constraint against PIw

- syllables in Tangut names from nonstandard dialects that did not undergo -w loss after labials

Maybe the other unexpected syllable

1piew 'poor'

is also a borrowing from such a nonstandard dialect.

*7.31.00:28: The fifth tangraph 1sã is missing from Li's (2008: 553) entry for 2mieu; it is, however, in Li's (2008: 640) entry for 1sã.

**7.31.4:13: The Tangut version of this Sanskrit phrase has features of TPNWC absent from both Middle Chinese and Sanskrit (in bold):

Tangraph Tangut Sinograph TPNWC MC Sanskrit Notes
ʔa *ʔa *ʔa a The most straightforward syllable.
Tangut tone unknown.
2dəəu *ndəu *nəuh, (*ɲuok, *nouk) nu(t?) If the Tangut had been borrowed directly from Sanskrit, it would have been nəu. (There is no nu in my Tangut reconstruction.)
Less common MC readings in parentheses.
1t̃ *to *ta (t?)ta If the Tangut had been borrowed directly from Sanskrit, it would have been ta. The nasalization in Tangut defies explanation; was this syllable initially borrowed into a dialect that had denasalized -õ?
1lo *lo *la ra If the Tangut had been borrowed directly from Sanskrit, it would have been raʳ. (There is no ra in my Tangut reconstruction; r- must be followed by a retroflex vowel except in rhyme 2.37 -iẽ.)
1sã *sã *sam sam
2mieu *mieu *mieuh, (*mæuh, *mɔk) myak None of the MC readings (including the less common readings) is a good match for Skt myak. Other Chinese transcriptions of this syllable are
貌 MC *mæuh, *mɔk
耶 MC *jæ
1sã *sã *sam sam
1po *phəu *bo bo Could Tangut 1po-1tiẽ be a borrowing from a Chinese dialect with *po tie instead of TPNWC *phəu thi for 'bodhi'? If the Tangut had been directly borrowed from TPNWC, I would expect phəu thi; if it had been directly borrowed from Sanskrit, I would expect bo di.
Skt dhi corresponds to MC *de because there was no MC *di.
As with 1tõ above, the nasalization in Tangut 1tiẽ defies explanation.
1tiẽ *thi *di dhi

I am tempted to revise my reconstruction and replace nasalization in rhyme groups VIII ̣and XI ̣(following Gong's numbering) with Gong-style glides:

1tõ > 1tow (cf. Gong's 1tow)

1tiẽ > 1tiej (cf. Gong's 1tjɨj)

There is no doubt that rhyme groups III and V are the nasal counterparts of groups II and IV. (Capital letters represent generic vowel types. Y represents central nonlow vowels.)

Oral Nasal
I. -U (XII. -Ũ - loanwords only)
II. -I III. -Ĩ
IV. -A V. -Ã
VI. -Y (none; did *-Ỹ denasalize and merge with *-Y?)
VII. -E VIII. -Ẽ (-EJ?)
IX. -IW (none)
X. -O XI. -Õ (-OW?)

I had assumed that VIII and XI were to the preceding rhyme groups VII and IX what III and V were to the preceding rhyme groups II and IV. VIII and XI tend to transcribe Chinese nasal rhymes, though there are exceptions. Is it possible that TPNWC

- was losing its nasal vowels

- had glides corresponding to earlier nasals; cf. how Sino-Japanese -i and -u correspond to MC nasals

- had nasal vowels in syllables that had no nasals in standard dictionary MC; cf. the unexpected nasal vowel in Taiwanese 他 thã corresponding to an oral vowel in MC *tha

On closer examination, the distribution of vowel types within VIII and XI does not match the pattern shared by the uncontroversial nasal rhyme groups III and V:

rhyme class \ vowel class short long tense retroflex
III and V -Ĩ, -Ã


Do VIII and XI really constitute a set? Should they be reconstructed with features distinguishing them from each other as well as from III and V? XI patterns like IX, suggesting that they both shared final -w.

Tangut fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2012 Amritavision