14.6.21.23:54: NAMES OF THE TANGUT CAPITAL (PART 1)

This article made me wonder about the name of the Tangut capital. Today I found this article by Ruth Dunnell about it.

The first Tangut capital was 靈州 Lingzhou, captured in 1002 and renamed 西平府 Xipingfu. I don't know how these names would have been written in the Tangut script invented roughly three and a half years later. 州 -zhou 'prefecture' was borrowed into Tangut as

=+

0707 1tʂɨew =

bottom left of 1408 1lhioʳ 'place' +

left ('earth') of 2627 2lɨə̣ 'earth'

but beyond that I can't make any further guesses.

The capital then moved to 興州 Xingzhou around 1020. Its Tangut name was

2635 0707 2xiõ 1tʂɨew

The first half is puzzling for three reasons.

First, the analysis of its character is unknown. The left side is 'earth' (like the right side of 0707) but the right side (Boxenhorn code dol) could be from one of eight characters which all have inappropriate semantics:

2582 2niaa 'mud'

2623 2ʂɛ̃ (second half of 2khə 2ʂɛ̃ 'servant')

3870 1pi (second half of 2sɨə 1pi 'Xianbei')

3896 1pi (second half of 1ʂɨə 1pi 'tick')

3916 2si (perfective and nominalizing suffix)

3931 2si (first half of 2si 1thwao 'phantom')

3979 1veị 'stupid'

5384 2nwiə̣ 'rush used as a candle wick'

Second, 2xiõ has an o absent from Tangut period northwestern Chinese 興 *x(ɨ)ĩ. In theory the Chinese word could have been borrowed as *1xĩ, a theoretically possible Tangut syllable that does not exist.

There is no Tangraphic Sea entry for 2xiõ, but the Homophones states it is a clan name 'Hon' (in my lay Tangut transcription). Did the Tangut name the capital after the Hon, and was the Chinese name 興 *xɨĩ 'rise, flourish' chosen because it sounded vaguely like Hon?

Third, that theoretical *1xĩ would have distinguished the Tangut name for Xingzhou from the homophonous Tangut name

2567 0707 2xiõ 1tʂɨew

of 雄州 Xiongzhou. Could 2567 2xiõ be a transcription of Chinese 雄 *xɨũ, unlike 2635 2xiõ which might have been Sinified as 興 *xɨĩ? (A  Tangut syllable *xɨũ that was a perfect match for Chinese *xɨũ was not possible since Tangut lacked the rhyme -ɨũ.)

The analysis of 2567 is unknown; it looks like

+

all of 3485 1lạ 'hand, arm' +

left ('earth') of 2627 2lɨə̣ 'earth'

Was Xiongzhou somehow associated with hands: e.g., did it have a hand-shaped landmark?

Sometime in the 1030s the capital was renamed 興慶府 Xingqingfu 'Flourishing Celebration Prefecture'. I don't know the Tangut version of the name, but my most straightforward guess is

2635 3118 2xiõ 1khĩ 1xou

with transcription characters for 慶 *khĩ and 府 *fu. (Tangut had no *f- or a simple vowel *-u.)

Even later - by the mid-12th century if not earlier - the name of the capital was changed to 中興府 Zhongxingfu 'Central Flourishing Prefecture'. The Tangut form of this name was

2635 3118 1tʂõ 2xiõ 1xou

with a transcription of Chinese 中 1tʂɨũ whose rhyme had no exact Tangut equivalent.

Next: Other names for the Tangut capital.

14.6.20.23:20: G-*R-ADATION IN TANGUT (PART 1)

Now that I've established in the last two posts what I think the four grades might have been in Chinese, I can move on to guessing what they might have been in Tangut.

I am assuming Tangut had grades. I am not absolutely sure because I don't know of any Tangut text that mentions them, whereas I do know the Tangut terms for 'rhymes' and 'tones':

0433 1biu 'rhyme' and 1586 1ɣɤị 'tone'

Nonetheless there are strong correlations between Tangut rhymes and Chinese grades. Tangut rhymes come in groups of three or four, and each rhyme may tend to be transcribed by a particular Chinese grade. Here are the frequencies of the Chinese grades of the Chinese rhymes corresponding to Tangut rhymes 8-11 in Tai (2008: 206-208):

In this set of four Tangut rhymes,

- Tangut rhyme 8 only corresponds to Chinese Grade I (though a sample of two is too small to be meaningful)

- Chinese Grade II is unique to Tangut rhyme 9 and is the most common Chinese grade (though again the sample is too small)

- Tangut rhyme 10 only corresponds to Chinese Grade III

- Chinese Grade IV is unique to Tangut rhyme 11

The figures for Chinese Grade III and IV after the slash are mine. The Chinese syllables 移息鼻脾琵名荑夕膝 whose rhymes correspond to Tangut rhyme 11 are considered Grade III by most modern scholars even though they are in Grade IV in the Yunjing rhyme tables. I follow the Yunjing rather than the usual classification, so I count them as Grade IV rather than as Grade III. (息 and 名 occur twice each.) Even if one rejects my figures, Chinese Grade IV remains unique to rhyme 11.

While some exceptions may be random errors, yet others may be explained upon closer examination: e.g., the Tangut period northwestern Chinese Grade III syllable 備皮被 *ph(ɨ)i (with two different tones) was transcribed as Tangut Grade IV

0749 1phi

because Tangut had no Grade III syllable *phɨi.

It is also possible that the Grade III medial *-ɨ- had been lost in Tangut period northwestern Chinese, making the resulting *phi a perfect segmental match for Tangut 1phi.

Tangut undoubtedly has at least two strata of Chinese loanwords; the earlier stratum (which may turn out to contain multiple substrata) has archaic Chinese traits going back centuries.

I conclude that Tangut developed a Chinese-like grade system under heavy Chinese influence. The Tangut grades may not have been identical to those of Chinese, but they were close enough for frequent matches like those in the table above.

Here is my attempt to adjust my previous reconstruction of the Tangut grades to fit my current reconstruction of the Chinese grades:

There are many differences between this scheme and the one I reconstruct for Chinese.

One key difference is that Tangut Grade I was characterized by a shared medial vowel /a/ whereas Chinese Grade I was characterized by the absence of a medial vowel.

Another key difference is that all six basic Tangut vowel types occurred in all four grades, whereas Chinese vowels had restricted distributions: e.g., there were no Chinese Grade I or II *-i rhymes corresponding to Tangut Grade I /ai/ and Grade II /ɤi/.

Next: The origin of the Tangut grades.

14.6.19.23:49: G-*R-ADATION IN CHINESE (PART 2)

Part 1 was largely a recap of what I've written about early Chinese phonological history here and in this book. This part introduces some new twists.

Late Old Chinese (LOC)

Middle Old Chinese allowed sequences of vowels with mixed height types within a sesquisyllable: e.g.,

埋 *mˁʌˁ-ʀˁəˁ (lower vowel + higher vowel)

離 *tɯ-raj (higher vowel + lower vowel)

In LOC, the root vowels 'bent' (in bold) if they did not already match the height class of a preceding presyllabic vowel:

(The concept of bending is from Axel Schuessler; the details are mine.)

Emphatic *aˁ backed to *ɑˁ; its reflexes in modern Chinese languages are usually back.

Emphatic *əˁ may have lowered and backed to *ʌˁ, but I will continue to use the symbol *əˁ to distinguish the emphatic central higher root vowel from the emphatic higher presyllabic vowel *ʌˁ.

Other emphatic vowels may have changed slightly in ways not indicated in the table: e.g., *eˁiˁ might have been *[*ɛˁɪˁ], etc.

Nonhigh vowels after *ɯ became rising diphthongs beginning with high vowels.

High vowels after *ʌˁ became falling diphthongs beginning with mid vowels.

These bent vowels were originally allophonic but became phonemic after presyllabic vowels (and entire presyllables) were lost along with 'emphasis' in later LOC:

卯 */mʌ-ruʔ/ *[mˁʌˁ-ʀˁoˁuˁʔˁ] > */mʀouʔ/ [mʀouʔ]

離 */tɯ-raj/ [tɯ-rɨaj] > /lɨaj/ [lɨaj] (MOC *r- became LOC *l-.)

MOC dentals and alveolars became LOC retroflexes before rhotics:

宅 *dˁʀˁɑˁqˁ > *ɖʀɑq

知 *Tɯ-te > *Tɯ-tie > *rtie > *trie (or *rʈie?) > *ʈrie

MOC *sˁ may have become retroflex after *qˁ- which was lost in LOC:

山 *qˁsˁɑˁnˁ > *ʂɑn

In later LOC, medial rhotics lenited, resulting in new diphthongs:

宅 *ɖʀɑq > *ɖʁɑq > *ɖɣɑq > *ɖɤɑq

知 *ʈrie > *ʈɨie >  *ʈɨe

Most retroflexes developed before rhotics, so most came to be followed by an achromatic vowel (*ɤ before lower vowels and *ɨ before higher vowels). Such back vowels also spread by analogy after those few retroflexes that had not developed before rhotics:

*ʂɑn > *ʂɤɑn

The 'grade' system was developed for LMC. The interpretation of the grades has long been controversial. I have long thought that emphatic syllables developed Grades I and II whereas nonemphatic syllables developed Grades III and IV. However, I used to think Grade II conditioned lowering. But now I think it might have had its own characteristic medial:

I. no medial: 高 LMC *kɑw < MOC *qˁɑˁwˁ

II. medial *-ɤ- (< usually from *-ʀˁ-)  交 LMC *kɤɑw < MOC *qˁʀˁɑˁwˁ

III. medial *-ɨ- (< *-r- and *-ɨ- from bending) 嬌 LMC *kɨew < MOC *Cɯ-kaw

IV. medial *-i-: 澆 LMC *kiew < MOC *Cɯ-kew

Those four example words are from the 守温 Shouwen manuscript from Dunhuang (Pulleyblank 1984: 75).

I got the idea of a medial *-ɤ- for Grade II from Zhengzhang Shangfeng who reconstructed a medial fricative -ɣ- in Grade II: e.g., his *kɣau corresponds to my *kɤɑw.

- In the southern ancestor of Taiwanese and the source dialects of Sino-Japanese Go-on and early Sino-Vietnamese, *ɑ after *-ɤ- fronted (to dissimilate from it?) and then *-ɤ- was lost: e.g.,

夏 EOC *gras > MOC ɢˁʀˁɑˁχˁ > LOC *ɣɤɑʰ > early MC *ɣɤæʰ > *ɣæʰ >  Taiwanese hē, Go-on ge, early Sino-Vietnamese hè

- In Cantonese, *-ɤ- has left a trace as vowel length: e.g.,

交 LMC *kɤɑw > kaːw

- In Mandarin, *-ɤ- fronted to *-e- which raised to *-i- and merged with *-ɨ- and original *-i-. This new *-i- conditioned palatalization of velars:

II. 交 LMC *kɤɑw > *keɑw > *keaw > *kiaw > jiao [tɕjaw]

III. 嬌 LMC *kɨew > *kiaw > jiao [tɕjaw]

IV. 澆 LMC *kiew > *kiaw > jiao [tɕjaw]

Perhaps Cantonese had a similar intermediate stage with *-e- reflected by Sino-Vietnamese giao [zaːw] < *kjaw:

交 LMC *kɤɑw > *keɑw > *keaw > kaːw

(The ancestor of Cantonese was not the source of late Sino-Vietnamese, but the two were probably close.)

I am not entirely committed to reconstructing *-ɤ- for many reasons: e.g., I don't know of any language that has ɤV-diphthongs.* Nonetheless, as is often the case on this blog, I wanted to try out an idea by seeing how far I could go with it.

Next: Regrading Tangut.

*UPDATE: 6.20:0:24: UPSID lists a single language (Parauk) with only two ɤV-diphthongs: ɤi and breathy ɤ̤i. Neither is in my LMC reconstruction which is full of *ɤɑ.

14.6.18.23:57: G-*R-ADATION IN CHINESE (PART 1)

At the end of my last post, I asked,

How can I salvage my latest [Tangut vowel] reconstruction?

I have an answer, but I'd like to explain how I got it first. And that entails going back to ...

Early Old Chinese (EOC)

This is the earliest stage reconstructible using internal evidence. There are no foreign transcriptions at this point, and going back any further would require comparison with other Sino-Tibetan languages.

EOC had none of the features that were prominent in later stages: 'emphasis' (i.e., pharyngealization), complex vowels, tones, or 'grades'.

Roots were monosyllabic with six different vowels: three higher (*i, *ə, *u) and three lower (*e, *a, *o). The addition of prefixes with unstressed short vowels made them sesquisyllabic.

There were at least two vowels in prefixes, a higher vowel that I write as *ɯ and a lower vowel that I write as *ʌ. (These symbols were inspired by the higher and lower minimal vowels of Middle Korean.)

In EOC, prefixes with either vowel could precede roots with any vowel, so there were twelve possible vowel combinations in a sesquisyllable.

2 (*ɯ/*ʌ) x 6 (*i/*e/*ə/*a/*u/*o) = 12

Middle Old Chinese (MOC)

Syllables and sesquisyllables whose first vowel was lower (*e/*a/*o or *ʌ) became emphatic (i.e., pharyngealized): e.g.,

捭 *rpeʔ > *pˁʀˁeˁʔˁ (*r- backed to uvular *ʀˁ and metathesized if it was in preinitial position)

都 *ta > *tˁaˁ

雊 *kos > *qˁoˁχˁ (emphatic velars and final *-s backed to uvulars)

埋 *mʌ-rə > *mˁʌˁ-ʀˁəˁ (higher vowels became 'emphatic' after *ʌˁ)

Syllables and sesquisyllables whose first vowel was higher (*i/*ə/*u or *ɯ) remained unchanged: e.g.,

離 *tɯ-raj (lower vowels remained unchanged after *ɯ)

MOC had fourteen phonetic vowels: seven emphatic and seven plain:

- *ʌˁ was always emphatic

- *ɯ was always nonemphatic

- root vowels could be emphatic or nonemphatic:

*iˁ *eˁ *əˁ *aˁ *oˁ *uˁ

*i *e *ə *a *o *u

The emphasis of the root vowel was predictable and not yet phonemic:

- Lower root vowels were always emphatic if preceded by no prefix or *ʌˁ

- Lower root vowels were always nonemphatic if preceded by *ɯ

- Higher root vowels were always nonemphatic if preceded by no prefix or *ɯ

- Higher root vowels were always emphatic if preceded by *ʌˁ

The pronunciation of /r/ was also predictable in a similar way:

[ʀˁ] if preceded by *ʌˁ or if in word-initial position and followed by a lower vowel:

*/CʌrV/ *[CˁʌˁʀˁVˁ]

*/re/ *[ʀˁeˁ], */ra/ *[ʀˁaˁ], */ro/ *[ʀˁoˁ]

*/rʌCV/ *[ʀˁʌˁCˁVˁ]

[r] elsewhere:

*/CɯrV/ *[CɯrV]

*/ri/ *[ri], */rə/ *[rə], */ru/ *[ru]

*/rɯCV/ *[rɯCV]

These two pronunciations of /r/ could have been like the pronunciations of Old Turkic 𐰺 <r¹> and 𐰼 <r²>.

*/V/ above could be any of the six root vowels.

I will get back to Tangut eventually. Stick with me.

14.6.17.23:54: FOUR MEDIAL VOWELS FOR FOUR GRADES IN TANGUT?

I ended my last post by proposing that each of the four grades of Tangut rhymes was characterized by a medial vowel of a different height. If each medial vowel could be followed by six types of main vowels (as in the reconstructions of Arakawa and this site for years), and the medial vowel's phonetic realization was conditioned by the following main vowel, Tangut would have twenty-four basic phonemic diphthongs.

Here is a comparison of four reconstructions: Gong's, Arakawa's, and two of mine: my previous reconstruction and my new four-medial vowel reconstruction.

Gong's reconstruction (his ij/ej corresponds to e in other reconstructions and his III corresponds to my III and IV)

Arakawa's reconstruction (his III corresponds to my III and IV; he reconstructs a single Grade IV rhyme -jaː not included in this table)

My old reconstruction

My four-medial vowel reconstruction

That last solution is elegant, but does it fit the facts?

Here's one fact: in Tangut as well as in Chinese, shibilants occur before Grade II and III grades, but sibilants occur before Grade I and Chinese Grade IV (and Tangut Grade IV in my reconstruction).

Therefore Grades II and III had to be shibilant-friendly in some way.

In Gong's reconstruction, Tangut shibilants were palatals and could only occur before the Grade II palatal vowel -i- and the Grade III palatal glide -j-. This makes phonetic sense, but there is very little support for -i- or -j- in Tibetan transcriptions: e.g.,

Gong's Grade II -iej always corresponds to -e or -eH in Tibetan transcription except for one case of -ye and one case of -i.

Gong's Grade III -ja always corresponds to -a in Tibetan transcription except for five cases of -aH, even though -ya is possible in Tibetan transcription.

In Arakawa's reconstruction, Tangut shibilants could only occur before the Grade II palatal glide -j- and Grade III long vowels.

First, there is very little support for Grade II -j- in Tibetan transcriptions: e.g., Arakawa's -je always corresponds to -e or -eH in Tibetan transcription except for one case of -ye and one case of -i.

Second, I don't know why Arakawa reconstructed Grade III as long. He reconstructed long Grade III vowels for characters transcribing Sanskrit short as well as long vowels: e.g.,

1863 1taː (transcription character) for Sanskrit ṭa, ṭā, ta, and tā.

Third, I don't know of any language in which shibilants precede long vowels but not short vowels. Perhaps this anomalous distribution could be explained by reconstructing *iV-diphthongs as sources of Grade III long vowels: e.g., *ia > aː, etc. (Cf. Gong's reconstruction in which iV-diphthongs are in Grade II, not III!)

In my old reconstruction, Tangut shibilants were retroflexes that could only occur before lowered Grade II vowels and Grade III -ɨ-. That was also the distributive pattern of retroflexes (but not palatals!) in Middle Chinese. There was no way to transcribe lowering or -ɨ- in Tibetan, so it is not surprising that Grade II and III rhymes were usually transcribed as simple vowels without -y-. (The transcription dgye for

5057 1ɣɛ 'true'

may simply be an error; other transcriptions are se [due to confusion with

4751 1se 'clean'],

Hgi [the only example of this rhyme being transcribed with -i], dge, dghe, and Hge.)

In my new reconstruction, Tangut shibilants occur before the Grade II glide /ɜ/ and the Grade III glide /ɘ/. This makes no phonetic sense.

Moreover, there is no reason to reconstruct such glides in Chinese. Although there is no guarantee that the Tangut and Chinese grade systems were identical, the fact that they strongly correlate with each other in transcriptions as noted by Kychanov and Sofronov 1963, Nishida 1964, and Gong 1995 implies they were similar. Unfortunately, there is no consensus on how to interpret the Chinese grade system.

Next: How can I salvage my latest reconstruction?

14.6.15.23:50: DID TANGUT HAVE A VOWEL SYSTEM LIKE MARSHALLESE? (PART TWO)

I thought I'd be able to move on, but I can't let go of this subject just yet ...

At first it is difficult to believe that Tangut - with simple-looking Tibetan transcriptions like ma for

'mother'

or mo for

'heaven'

could have sounded anything like Marshallese with complex vowel sequences in words such as

[mʲææ̯] 'breadfruit'

[mʲæ͡ɑɑ̯] 'but'

[mʲæ͡ɒɒ̯] 'taboo'

Then again, in Marshallese orthography, those words are mā, ma, and mo̧ - and stripped of diacritics, they look just like the transcriptions of the Tibetan transcriptions ma and mo. Those three words have been phonemically analyzed as consonant-vowel-glide sequences (spaces have been added for clarity):

/mʲ a j/

/mʲ a ɰ/

/mʲ a w/

Although Tangut fanqie do not support a rich, Marshallese-type system of consonants with secondary articulations, what if many Tangut syllables had the structure

consonant + (w) + vowel + (glide)

Suppose each of the four Tangut grades had a vowel of a different height like Marshallese (ɨ, ɘ, ɜ, a) and could be followed by zero or various glide codas (w, j, ɰ, ɥ). There could be twenty different combinations - not too far from the twenty-four of my 6 x 4 grid:

The above system could be enlarged with the addition of a third dimension: frontness:

2 (front/back) x 4 heights x 5 (zero + 4 glides) = 40

Each rhyme group could be defined in terms of frontness and a glide:

I regard rhyme 104, III, V, VIII, and XI as nasalized counterparts of I, II, IV, VII, and X. VI has no nasal counte)rpart.

Here is what rhymes 8-11 and 17-20 (both ending in Ø) could look like:

This cannot be right, as the transcription data clearly does not match this scheme: e.g., rhyme 8 is almost always transcribed as i in Tibetan and rhyme 20 is almost always transcribed as a in Tibetan. (The exceptions still contain i and a: two instances of iH and one of ing for rhyme 8 and one of aH for rhyme 20.)

The transcription data seem to indicate a vowel shared across grades rather than one vowel per grade. Perhaps the grades stood for the height of the first part of a vowel or diphthong: e.g.,

In that system, the twenty-four slots would consist of four grade markers of different heights (/ɨ ɘ ɜ a/) followed by six vowels (/u i ɑ ɯ e o/). The latter color the former: e.g., front vowels front /ɜ/ to [ɛ] and back vowels back it to [ʌ] (see rhymes 9 and 18 above). The grade elements could have been very short and were hence ignored in the Tibetan transcriptions: e.g., rhyme 8 [æ̆i] was transcribed as i, etc.