Until two days ago, I assumed that Khitan small script character 342,

variously read or transcribed as

[en] (Chinggeltei et al.1985: 99; Chinggeltei 2002: 29, Chinggeltei 2010: 421)

[ən] (Aisin Gioro 1999)

[æn] (Aisin Gioro 2004)

<iên> (Kane 2009: 63)

<üan> (Kane 2009: 305)

<üen> (Wu and Janhunen 2010: 268)

was only in Chinese borrowings.

But when I was looking for combinations of

033 <is> 'nine'

with 'yang' characters for my last post, I found this set of seven blocks combining 324 with the yang character 151 <ghu> which never appears in Chinese borrowings:

Block Character numbers Transliteration Attestations
151-324 <ghu.?n> 興宗 34.13, 蕭仲恭 30.49
151-324-033-090 <ghu.?n.is.ó> 許王 21.2
151-324-033-090-097 <ghu.?n.is.ó.úr> 道宗 26.19, 蕭仲恭 40.31
151-324-033-261-051-122 <ghu.?n.is.l.gha.ai> 道宗 6.6
151-324-033-261-051-189-098 <ghu.?n.is.l.gha.a.al> 道宗 24.18

151-324-033-261-051-290 <ghu.?n.is.l.gha.án> 仁懿 18.11, 蕭令公 5.7
151-324-033-311 <ghu.?n.is.b> 許王 18.31, 許王 22.10

If 324 can be in the same block as <ghu> and <gha>, it is either yang or neutral. Which is a surprise if it was <üan> or <üen> because the combination of gh plus ü is alien to Mongolian (and Turkic). I think its vowel may have been [ɛ] which is the same height as the yang vowel [ɔ]. Could <ó> in the second and third blocks be [ɔ] - and úr in the third block be [ʊr] with the yang vowel [ʊ]?

Was Khitan [ɛ] was a yang vowel like the nonhigh front vowel *e, which I've regarded as the Old Korean yang counterpart of originally yin (later neutral) *i?

I already tentatively read

073 <ên>

as [ɛn], and 324 doesn't look like a variant of 073 (i.e., a character with an identical reading), so I think the reading of 324 must have been slightly different: e.g., [jɛn]. So perhaps the seven blocks above were read as something like

1. [ʁʊjɛn]

2. [ʁʊjɛnɪsɔ]

3. [ʁʊjɛnɪsɔʊr] or [ʁʊjɛnɪsɔːr]?

4. [ʁʊjɛnɪs(ɔ)lʁɑ(ː)j]

5. [ʁʊjɛnɪs(ɔ)lʁɑ(ː)l]

6. [ʁʊjɛnɪs(ɔ)lʁɑ(ː)n]

7. [ʁʊjɛnɪs(ɔ)b]

8.6.23:36: Do all those words share a root [ʁʊjɛn]? Are [ɪsɔ] and [b] heretofore unknown (derivational?) suffixes? Could 3 have a masculine perfective suffix /ɔr/? Are the [lʁɑ]-forms passives or causatives followed by converbs [ɑj], [ɑn], and the feminne perfective [ɑn]?

I have projected the vowel [ɔ] from 2 and 3 into what would otherwise be consonant clusters in 4-7.

I don't know if graphic double or triple vowels in 4-6 represented long vowels. <IS> 'NINE' NEUTRAL?

In "Emperor Nine", I proposed that Khitan <i> was phonemically /i/ with a lowered allophone [ɪ] in words with 'lower' series vowels: [ɑɔʊ].

Since vowel series correlate with degrees of consonant backness (higher : velar and lower : uvular) and possibly pharyngealization (as in modern Khalkha [see Janhunen 2003] and my Old Chinese reconstruction, though this detail is difficult to reconstruct), I will speak of 'yin'and 'yang' syllables

Syllable type Pharyngealization Velar or uvular? Vowel height Examples
Yin No Velar Higher [tə kə gə xə]
Yang Yes? Uvular Lower [t(ˁ?)ɑ qɑ ʁɑ χɑ]

to emphasize that whole syllables - not just segments - belong to one class or the other. And as many Khitan small script characters represent syllables, I will speak of 'yin characters' (e.g., <k g ge>) and 'yang characters' (e.g., <qa gha xa>).


<is> 'nine'

from "Emperor Nine" a yin character or a yang character? Its Proto-Mongolic cognate *yersü/n 'nine' has 'feminine' (i.e., yin) vowels, but that does not guarantee that it too is yin.

I propose a simple (and hence probably not foolproof test) of yin/yang classification:

- A character is probably yin if it appears in a block with the yin characters par excellence

<k g ge>

- A character is probably yang if it appears in a block with the yang characters par excellence

<qa gha>

- If a character appears in a block with both types of characters, it is neutral.

Note that cooccurrence with the invariable accusative/instrumental suffix <er> is not evidence for classifying a character as yin. The yang characters par excellence <qa> and <gha> appear before the suffix <er> in

<qa.gha.ad.er> 'qaghan-PL-ACC' (found in Kane 2009: 132 which does not specify the source; not in the texts in Qidan xiaozi yanjiu)

but that does not mean that <qa> and <gha> are yin or neutral because <er> has no allomorphs; it appears after syllables of all types. (Contrast that <er> with its homophone, the perfective suffix which has <er> after yin stems, <ar> after yang stems, and <or> after <o>-stems [yin/yang to be determined].)

Also note that this test only applies to words not borrowed from a language without harmony (e.g., Chinese). Chinese transcription combinations like

<k.ai>, a transcription of Chinese 開 *kʰaj 'open'

contain velar-lower vowel (i.e., yang consonant-yin vowel) sequences that I wouldn't expect in native Khitan words (or words borrowed from languages with harmony: e.g., <qa.gha> 'qaghan' which might be borrowed from Turkic).

I regard <is> as neutral because it appears with both yin and yang characters:

033-334-144 <is.g.ún> [isgun] (a title or name transcribed in Chinese as 乙室謹 *iʂikin; 蕭仲恭 1.10, 50.17) <g> is yin

The Chinese transcription may indicate that /s/ was palatalized to [ɕ] after [i].

The mismatch of 144 <ún> with Chinese *-in here and with -an in foreign renditions of <qid.ún> 'Khitan' is unexplained; the interchangeability of 311-144 <b.ún> with 288 <bun> (transcribed in Chinese as 本 *pun) points to an u-vowel (Kane 2009: 52).

073-033-051-123 <ên.is.gha.ar> [ɛnɪsɣɑr] '?' (興宗 31.14); <gha> is yang, and presumably <ên> is either yang or neutral

8.5.1:50: I am guessing that /s/ did not palatalize after [ɪ]. Cf. Ukrainian which has [sʲi] and [sɪ] but not *[sʲɪ].

I am now skeptical about pharyngealized consonants in Khitan. If yang syllables had pharyngealized segments, then <ên.is.gha.ar> would have been pronounced [ɛˁnˁɪˁsˁɣˁɑˁrˁ], and the character <is> would have stood for both [is] and [ɪˁsˁ]. Although Middle Old Chinese phonetic elements could represent both pharyngealized and plain syllables, that flexibility was due to sound changes that complicated an earlier, simplier system in Early Old Chinese: e.g., the phonetic series 壹 GSR 395 (Schuessler 29-13):

Sinograph Early Old Chinese: 壹 for *...ʔit(s) syllables Middle Old Chinese: 壹 for *ʔit(s) and *ʔˁiˁtˁ(sˁ) syllables
㦤懿饐 *ʔits *ʔits
曀㙪殪 *Cʌ-ʔits *ʔˁiˁtˁsˁ

The Khitan small script had no long history (though the possibility of changes in spelling over time should be explored), and so I assume the fit between sound and symbol was fairly close during its three centuries of use. EMPEROR NINE

If my hypotheses about Khitan consonant/vowel harmony are correct, was the first syllable of Chinese *xɔŋti 'emperor', written as


in the Khitan large and small scripts, [χɔŋ] with a uvular and lower series vowel or [xoŋ] with a velar and a higher series vowel? Or was it [xɔŋ] as in Chinese with an un-Khitan velar plus lower series vowel sequence?

I don't know for sure, which is why my transliteration <xong> is agnostic: <x> could be velar or uvular, and transliteration vowel symbols generally do double duty for the upper and lower series (except for <e> and <a> which form an upper-lower series pair). But maybe looking at combinations of <xong> in the small script might point in one direction or the other.

Qidan xiaozi yanjiu lists two such combinations for what I assume to be native (or at least non-Chinese) words:

075-151 <hong.ghu> (仁懿 6.1)

I think <gh> was uvular [ʁ], so <u> had to be lower series [ʊ]. My harmony hypothesis would then require <hong> to be [χɔŋ] (or in very strict notation, [χɔɴ] with a uvular nasal).

I suspect Khitan had a neutral vowel /i/ with higher and lower series allophones: [i] and [ɪ] depending on the context. So

075-033 <hong.is> (許王 52.8)

might have been pronounced [χɔŋɪs].

(The post title refers to this character block which does not mean 'emperor nine', though its components do mean 'emperor' and 'nine'.)

Next: Was 'nine' really neutral? DID KHITAN HAVE GH- IN BORROWINGS FROM CHINESE?


But I think my first blog post in seven and a half months should be longer than that, so here's the reasoning behind my answer.

For some time I've revised Kane's (2009) transliteration of the Khitan word for 'emperor' (below in the large and small scripts) <hoŋ di> [ɣoŋ di]


as <xong di>: e.g., here. The reasoning for this reading is as follows:

1. The large script spelling looks exactly like Liao Chinese 皇帝 *xɔŋti 'emperor'. This fact by itself does not guarantee that Khitan 皇 帝 also meant 'emperor' or sounded like the Chinese word for 'emperor', as there are many large script characters which have meanings and/or readings unlike their Chinese lookalikes.

2. The context of these two spellings indicates that they indeed did represent a Khitan word for 'emperor'.

3. In the Langjun inscription, the first character of the small script spelling is also used to write the Chinese surname 黃 *xɔŋ, a homophone of the first syllable of 皇帝 *xɔŋti 'emperor'. Therefore it is likely that the Khitan word for 'emperor' was something like *xɔŋti: i.e., <xong di>, which may have phonetically been something like [χɔŋti]*.

4. Although it is true that the Chinese word for 'emperor' once had a voiced back fricative onset *ɣ/*ʁ/*ɦ, all other Chinese loans in Khitan postdate the devoicing of initial voiced obstruents: e.g., 唐 'Tang dynasty' in the Langjun inscription appears as

229-199-140 <ta.ang.en>

from *tʰaŋ plus the Khitan genitive ending <en>, not

*171-199-140 <da.ang.en>

from earlier*daŋ. There is no reason to assume that 'emperor' was borrowed before the rest of the Chinese vocabulary in Khitan. Therefore it is most likely that the initial of the Khitan word for 'emperor' was voiceless <x>, not voiced <gh> (= Kane's <h>).

Yesterday I realized that

'empress' (in the small script; what is the large script equivalent?)

in Khitan - also a loan from Chinese - must be <xeu> and not <gheu> (= Kane's <heu>) for the same reason that <xongdi> 'emperor is not <ghongdi>. The correct reading of this character is crucial because it is part of the spelling for

<?.úr> 'spring'

which Kane transliterated as <heu.ur> (= <gheu.ur> in my system) and which I transliterate with a voiceless initial as <xeu.úr>. The Khitan word is cognate to Written Mongolian qabur:

Written Mongolian q a b u r
Khitan (Qinggeltei 2002) <ɣou> <ur>
Khitan (Kane) <heu> [ɣ-] <úr>
Khitan (this site) <xeu> [xəw] <úr>

Although my voiceless x is a closer match to WM voiceless q than Qinggeltei and Kane's voiced initials, it is still a fricative rather than the stop <q> that I would expect to correspond to WM q (as in

<qa.gha> : WM qaghan).

'Qaghan' is an areal political term that could have been borrowed independently into Khitan and Mongolian. Perhaps *q in the common ancestor of Khitan and Mongolian weakened to <x> in Khitan, and <qa.gha> was borrowed after that weakening. Are all <q>-words in Khitan late borrowings? No, because nothing indicates that the Khitan autonym


<qid.ún> (in the large and small scripts)

is of foreign origin. Did Khitan preserve a two-way contrast lost in Mongolian?

*q- > K <q>, M q: e.g., 'qaghan', 'Khitan'?

*x-> K <x>, M q: e.g., 'spring'

Proto-Mongolic *x is unrelated to Khitan <x>; it is from pre-Proto-Mongolic *p: e.g., PM *xon 'year' corresponds to Khitan


<po> ~ <p.o> 'time' (in the large and small scripts)

Moving on to the first vowel of 'spring', I would expect it to be <a> in Khitan, but the fact that the first syllable of 'spring' is homophonous with a borrowing of Chinese 后 *xəw 'empress' rules out <a>. I side with Kane and transliterate with <e> (for [ə]). Chinggeltei's <o> could be explained as a Khitan assimilation of the first vowel to the second (or lacking a dissimilation that occurred in Mongolic). However, I think <o> might be an anachronism since 后 did not have a rounded vowel in northern Chinese until recently; that syllable was borrowed into Manchu centuries later as heo [xəw].

Could Khitan have retained a *e lost in Mongolic?

*qebur > K <xeu.úr>, M qabur

Although *qe was impossible in Mongolic which permits qa and ke but not *qe or *ka, Mongolic might have developed harmonic restrictions under the influence of early Turkic (which also permitted qa and ke but not *qe or *ka). Khitan underwent a parallel development after *q shifted to <x>; was not possible in Khitan, and <ka> could only occur in loanwords and transcriptions: e.g.,

<k.ai>, a transcription of Chinese 開 *kʰaj 'open'

The alternate spelling


may represent a [χɑj] that was harmonic and hence less un-Khitan than <k.ai>. <x> does appear instead of <k> even in syllables where <k> would be harmonic: e.g.,

<x.i>, a transcription of Chinese 起 *kʰi 'rise' and 期 *kʰi 'period'

The lenition of *-b- in the Khitan word for 'spring' is also found in


<tau> 'five' (in the large and small scripts)

whose Written Mongolian cognate tabun retains the medial consonant.

*I hypothesize that Khitan had two series of vowels, 'higher' and 'lower', and that they appeared in complimentary distribution after back consonants in native words:

Native Khitan vowels (excluding the vowels ɨ and y which are only in Chinese loanwords)

higher i (no [e]? merged with *i?) ə <e> o u
lower ɪ e [ɛ] <ê> (> initial [ja]?) a ɔ ʊ

Native Khitan KV-combinations

Consonant class 'higher' 'lower'
k-, g-, x- X
q-, ʁ-, χ- X

Velar-'lower' vowel combinations were possible in loanwords and transcriptions: e.g., <k.ai> above.

Khitan had two series of obstruents that are transcribed by Kane as if they were voiceless and voiced. I follow Kane's practice, though I am not sure what the phonetic distinction was. Kane's voiced series usually but not always corresponds to the Chinese voiceless unaspirated series, so it may have been phonetically voiceless aspirated: e.g., <d> = [t]. The possibility that Khitan had nonphonemic voicing like Korean should be explored: e.g., <qid.ún> 'Khitan' may have been /qitun/ = [qɪdʊn] with intervocalic voicing. The seemingly inconsistent transcriptions of Chinese syllable initials may show patterns in polysyllabic contexts: e.g., <t> initially and <d> in intervocalic position.

Tangut fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2015 Amritavision