10.5.8.23:58: GRADING TANGUT SYLLABLES: INTRODUCTION

One of the many challenges of reconstructing Tangut is figuring out how it could have 105 rhymes without final consonants. Clausen (1964: 66-67) wrote,

It does seem impossible that a Tangut phonetician, however acute his hearing, could have distinguished sixty-five different open vowel sounds [in Sofronov's 1963 reconstruction of Tangut], even if some of these were in fact diphthongs. [...] It must surely be the case that in its original state the table of 105 rhymes included a number of sounds ending in consonants ending in consonants or consonant clusters. It remains to be seen whether they can be reconstructed.

There is no doubt that pre-Tangut had final consonants. Last night I mentioned

phɔ̃ 'white'

whose nasality is a trace of a final *-m still present in rGyalrong. The question is whether these final consonants still existed when the Tangraphic Sea and other native works of Tangut phonology were written. I have been assuming that codas were already lost by that point, but I could be wrong.

How can a language have 105 open-syllable rhymes, particularly if its ancestor had a much simpler vowel system? I reconstruct only two vowels in pre-Tangut minor syllables (presyllables) and twelve vowels in pre-Tangut major syllables that develop into a far larger number of vowels in Tangut.

Pre-Tangut minor syllable vowels

(The symbols are used purely to distinguish minor syllable vowels from major syllable vowels. Minor syllable vowels may have been phonetically identical to major syllable vowels. There may have been more than two minor syllable vowels, but so far only two types are reconstructible.)

Pre-Tangut major syllable vowels

Pre-Tangut vowels belong to two classes:

High: *i(i), *ə(ə), *u(u) (six-way distinction reduced to *ɯ in presyllables)

Low: *e(e), *a(a), *o(o) (six-way distinction reduced to *ʌ in presyllables)

Note that 'high' and 'low' are not precise phonetic terms. The 'high' vowel *ə(ə) and the 'low' vowels *e(e) and *o(o) are all mid vowels, but they developed differently in my reconstruction.

Tangut vowels are classified according to 'grades' and 'cycles'. Each pre-Tangut vowel could have up to twelve different reflexes depending on preceding consonants and lost presyllables. Below are the reflexes of one pre-Tangut high class vowel and one pre-Tangut low class vowel.

The one unchanged reflex is in bold.

Tangut reflexes of pre-Tangut low class vowel *-a (Gong's rhyme group IV)

The one unchanged reflex is in bold.

You may have noticed that I now reconstruct the Grade III and IV members of group II with mid vowels -ə- and -e- instead of high vowels -ɨ- and -i-. I'll explain why in a future post.

Not all possible grade/cycle combinations exist: e.g., there is no Grade II Cycle II rhyme *-æ̣. Such 'missing' combinations may be accidental gaps or may have merged with other rhymes: e.g., *sCra may not have existed or may have merged with *sCa, becoming Cạ R66.

Next: Grade I.

10.5.7.20:36: UVULAR, RHOTIC, AND ALVEOPALATAL SOURCES OF TANGUT GRADE II

I spent my last two posts looking at Dayang Pumi phonology. The Dayang Pumi autonym pʂhɛ̃ mi 'White People' from Pre-Qiangic *pram (Matisoff 1997: 172) reminded me of Tangut phɔ̃ 'white' in the name of the Tangut state

phɔ̃ bie lhiẹ liẹ

'white high state great'

'The Great State of the White and High'

phɔ̃ in turn is cognate to these rGyalrong forms

Japhug ku-ɣrum < *βr- < *pr-

Somang kə-prám

Zbu kə-prúm, phrúm

from Proto-rGyalrong *prom (my reconstruction in Guillaume Jacques' system).

In "Qaput?", I derived Grade II from an earlier uvular initial, but no uvular is reconstructible in 'white'. Right now I think Grade II has six different sources: two with uvulars and four without them:

1a. Pre-Tangut uvular initial *Q(r)- + any vowel

1b. Pre-Tangut *Cʌ-Q(r)- + any vowel

2a. Pre-Tangut *Cr- + nonhigh vowel

2b. Pre-Tangut *Cʌ-Cr- + any vowel

3a. Pre-Tangut *Sh(r)- + nonhigh vowel (*Sh- = any alveopalatal initial)

3b. Pre-Tangut *Cʌ-Sh(r)- + any vowel

'White' has source 2a: *phr- + the nonhigh vowel *o which lowered to ɔ in Grade II.

Conversely,

4. Pre-Tangut *Cɯ-Q- + any vowel

5a. Pre-Tangut *Cr- + high vowel

5b. Pre-Tangut *Cɯ-Cr- + any vowel

lead to Grade IV (with i) and

6a. Pre-Tangut *Sh(r)- + high vowel

6b. Pre-Tangut *Cɯ-Sh(r)- + any vowel

lead to Grade III (with ɨ).

Two forms are listed after presyllables:

1. Forms which lost their presyllables before lenition: e.g., *Cʌ-q(r)a > *kæ

2. Forms which lost their presyllables after lenition: e.g., *Cʌ-q(r)a > *Cʌ-ɣæ > *ɣæ

Actual examples (from Jacques 2006; reconstructions are mine)

LFW = Li Fanwen 2008 tangraph number

= follows a presyllable that triggered lenition of the initial of the following syllable.

*C-: unknown consonant

*-H: glottal consonant conditioning rising tone; may ultimately be partly from *-s: e.g., 'louse' might have once been *srik-s.

*K-: back consonant conditioning aspiration

*-N: unknown nasal

*P-: labial consonant conditioning medial -w-

*S-: coronal consonant conditioning tense vowels

I have revised my reconstruction of R6 as Grade II -ʊʊ instead of Grade III -ɨuu. The two initials in R6 (kh- and ʒ-) normally only coexist in Grade II. Thus R6 is the long counterpart of R4:

Rhyme group I (large cycle rhymes only)

R14, R47, R58, R64, R70, and R92 have no grade III/IV distinction, but I treat them as Grade III after alveopalatal and labiodental initials and as Grade IV after other initials, following the general correlations between grades and initial classes:

There are a few cases of Grade IV after alveopalatal initials: e.g.,

dʒiə R31 1.30 'skin' instead of dʒɨə R30 1.29

Perhaps the front i in such syllables was conditioned by a lost presyllable or preinitial. Or these syllables may reflect a sound change in progress: -ɨə > -iə.

Problematic correspondences

R94 and R100 have no grade III/IV distinction, but I treat them as Grade IV after velar initials.

LFW0860, LFW2768, and LFW3582: Medial *-r- should result in Grade II, not Grade I with vowel retroflexion. Is the retroflexion from a suffix *-r unrelated to the medial -r- of rGyalrong, or did a coda *-r result from sporadic metathesis?

*CrV > *CVr

LFW1752: There is no Tangut rhyme -ʌʌʳ. Perhaps there was a merger of -ʌʌʳ with -ɨəəʳ.

LFW2768: There are no Tangut rhymes -ɛʳw or -ɪʳw. Perhaps *-ɪʳw merged with -iʳw, but I would expect *-ɛʳw to merge with -eʳw instead of somehow 'skipping' over -eʳw to merge with -iʳw.

LFW3582: There is no Tangut rhyme -æ̣. Perhaps there was a merger of -æ̣ with -ạ. If so, then kwạ is expected.

LFW4935: Did a presyllable or preinitial block the shift of *a to the extremely open vowel that I write as æ? Or could the Grade I vowel be an archaism?

LFW5109 and LFW5990: rGyalrong does not support the reconstruction of PT medial *-r- that would be the expected source for Grade II. Could a PT uvular preinitial *Q(ʌ)- have conditioned Grade II?

LFW5109: Nasality is unexpected. Maybe R42 and other rhyme group VII members ended in -j instead of nasal vowels. (If so, then rhyme group XI should be reconstructed with -w instead of nasal vowels.) Could the nasality be from a suffix *-N absent in rGyalrong?

10.5.6.23:46: MATISOFF (1997) ON DAYANG PUMI VOWELS AND TONES

199: Dayang Pumi (DP) has far fewer rhymes than Tangut because it has not lost presyllables and complex initials:

Both languages have completely lost final stops and nasals.

DP has a four-way distinction among sibilants (s : ʂ : ʃ : ɕ) whereas Tangut only has a two-way distinction.

DP retains uvulars whereas Tangut has lost them, though uvulars may have conditioned rhyme development (see my next post).

The DP basic vowel system is not quite symmetrical:

There is only one nasalized high vowel:

In UPSID, ĩ is slightly more common than ũ, and both are far more common than ɨ̃.

The lack of -ũ is reminiscent of Tangut which has no -ũ and has -əũ in only three words:

xəũ R104 1.96 'red' < Chinese 紅

təũ R104 1.96 'winter' < Chinese 冬

tsəũ R104 1.96 (a surname) < Chinese 宗?

-j can only follow -e. Were ever any other -Vj rhymes? If so, what happened to them?

200: The backing of DP /i/ after palatal fricatives and affricates reminds me of how Tangut alveopalatals normally occur in Grade III before ɨ rather than in Grade IV before i. I assume that Tangut ʃɨi < *ʃi, and for typing convenience, I may delete the predictable -ɨ- from my reconstruction after alveopalatals.

201: What is the etymology of DP ʃɛ 'Chinese'? I also have no idea where three Tangut terms for Chinese (zaʳ, khwa, and gwi) come from. Tangut xã is obviously from 漢.

Is DP xjɛ̃ 'seven' from *sj-? Cf. Tangut ʃạ and Mawo Qiang stə. I doubt the nasality has anything to do with an earlier root-initial *n still in Written Burmese khu-nac 'seven'.

201: DP has no -aw. Could -ow be from *-aw? Perhaps the development of [w] before -o predates the rounding of *-aw:

202: The lowest back vowel is ɒ instead of ɔ corresponding to ɛ. I presume *ɔ lowered to ɒ (cf. the lowering of *ɔ in some English dialects).

204: ʉ is the only rounded central vowel. I wonder what its origin is. Here are a few examples with possible Tangut cognates which have unrounded vowels:

205: Why is -ə̃ only after retroflex initials?

206: DP has two basic tones, high level (55) and low-to-mid rising (13 ~ 24; Matisoff's 'low').

Similarly, Tangut also has two basic 'tones', 'level' and 'rising'. It is unknown whether these were actually distinct pitches, phonations, or combinations of the two. The terms 'level' and 'rising' may have been taken from Chinese tonal terminology and may not describe tonal contours. The second volume of Precious Rhymes of the Tangraphic Sea contains 11 legible tangraphs with an 'entering tone'. I looked at these eleven years ago on this blog but should reinvestigate them.

207: DP has "sporadic" tone sandhi in noun compounds. The pattern of sandhi is not predictable.

Could this complexity reflect a more elaborate earlier tonal inventory? E.g.,

Hence

HH < *H1H1

LH < *H2H1

HL < *H1H2

but what would the current tones of *H2H2 be?

L tones may be mid level (33), low falling (21), or low level (11) in second syllables. Could these variants reflect earlier distinctions? Here's an extreme scenario in which an earlier four-way contrast was neutralized in isolation but maintained in compounds:

Since I think tones are more likely to lose contrasts in compounds (where the tones of parts become redundant), I suspect that the variants of the low tone are 'neutralized', as Matisoff suggested.

I have long suspected that irregular Tibetan transcriptions of Tangut tones may reflect tone sandhi as well as errors: e.g., tsho ngaH instead of tsho dngaH with a d- for the level (= first) tone of the second syllable of

tshọ^̣2 ŋa¹ 'void' < 'empty' + 'hollow'

might reflect a tonal assimilation rule

2 1 > 2 2

within disyllabic compounds. See Tai Chung-pui (2008: 35) for the entire transcribed passage.

209: L tones of verbs may assimilate to H-toned prefixes:

H- + L > H-H

This is like the affix-to-root assimilation I have proposed for Old Chinese, though the latter involves (non)emphasis rather than tones:

CV- + CVC = CVCVC (root deemphasizes after nonemphatic presyllable)

CV- + CVC = CVCVC (root emphasizes after emphatic presyllable)

210: *-a has several degrees of raising or 'brightening' in Pumi which are reminscent of reflexes of *a in Tangut:

Tangut also has other reflexes of *-a: -u, -əi, -ie, -ɪ, -e, -ə, etc.

Old Chinese nonemphatic *a has similar raised reflexes and more: e.g., [y]: 居 Md [tɕy] < OC *ka 'dwell'.

*APPENDIX: TANGUT INITIALS

My modification of Gong Hwang-cherng's Tangut reconstruction has 30-31 basic initials and 24 w-clusters:

I am uncertain about whether ɲ- needs to be reconstructed for the so-called 'retroflex'-initial syllables of Homophones.

Only six or seven initials cannot be followed by -w-: the five labials, r-, and perhaps ɲ-.

gw- is only in Grade IV. I wonder if any Cw-clusters do not occur in Grades II-IV.

Others have reconstructed even more initials, but no reconstruction of Tangut has as many initials as Dayang Pumi.

10.5.5.23:55: MATISOFF (1997) ON DAYANG PUMI CONSONANTS

Around the fall of 1996, after my first summer of Tangutology, I attended a lecture by James Matisoff about Dayang Pumi, a living relative of Tangut. I found an article expanding on his lecture in Mon-Khmer Studies 27. It is neat to look at Dayang Pumi again now that I have years of Tangut under my belt. I am amazed at how much data Matisoff collected in just 22 days. A few quick notes in order of page numbers:

172: I'll discuss the etymology of Pumi and its Tangut cognates in a future post.

172: Dayang Pumi (DP) has a lot of initial clusters like my reconstruction of Pre-Tangut (PT).

Tangut has few clusters because of 'feature shuffling': earlier initial distinctions conditioned rhyme distinctions that became phonemic after the simplification of initials. PT *Cr-clusters conditioned Grade II (see next post) and PT obstruent clusters conditioned tense vowels (indicated with subscript dots):

DP has labial-fricative clusters from earlier labial-liquid clusters. These clusters may resemble those of Old Vietnamese (though the OV clusters originated from labial-j sequences):

Perhaps PT had fricative-obstruent clusters like those of DP:

Matisoff (1997: 187) compared ʃtʃ- to Russian щ.

I think ʃtʃh- and ʃtʃhj- would be difficult to distinguish. The latter occurs in only one word in the data:

ʃtʃhju 'thief' (does it have Tangut cognates?*)

Matisoff (1997: 176-177, 195) transcribed the voiced uvular/velar preinitial cluster in four different ways.

Note the absence of velar xk(h)(w)- which might have been absorbed into the uvular clsuters.

175: The reduction of PM BəR- to BR- in noncareful speech reminds me of how I used to think that

Old Chinese *CəC- > *CC-

Old Chinese *CC- > *C-

e.g.,

各 OC *kəlak > *klak > MC *kak 'each'

洛 OC *krak > *rak > MC *lak (name of a river)

Sagart (1999) changed my mind. I now reconstruct

各 OC *klak > MC *kak

洛 OC *kʌ-lak > MC *lak

However, I cannot rule out the possibility that different OC dialects collapsed sesquisyllables in different ways: cf. the two pathways to collapse within a single language (Nha Heun; Ferlus 1971 as presented in Sagart 1999: 15-17).

176: I have not seen the shifts

*pw- > ɖw-

*ml- > ɖ-

outside DP. Perhaps they involved intermediate stages with a retroflex liquid *-ɭ-:

*pw- > *pɭw- > *ɭw- > ɖw-

*ml- > *mɭ- > *ɭ- > ɖ-

177: DP has only one voiceless nasal hm- < *sm-. I presume other *sN-clusters merged with regular voiced nasals.

If a language has a single voiceless nasal, would hm- be it? Only 18 languages in UPSID have voiceless nasals, and only one has a single voiceless nasal: Trumai with hm. (But the text of the UPSID Trumai entry mentions a dental voiceless nasal.)

193: What is the etymology of khɒ 'emperor'? A Sinification 可汗 of Mongolian qaɣan?

193: Matisoff derives DP uvulars from *velars and DP velars from *velar clusters. I wonder if the uvular/velar distinction is actually a retention rather than an innovation.

Like Qiang languages but unlike rGyalrong, DP has q- in 'head': qhu 'head'.

194: Why do Qiang languages have a fricative generally corresponding to stops in other languages in 'needle'?

Mawo Qiang χe

Taoping Qiang χe

Ronghong Qiang χa

but

Dayang Pumi qho

Japhug rGyalrong ta-qaβ

Somang rGyalrong ta-káp < *-q-

Written Tibetan khab

Written Burmese ʔap < *q-?

and

Zbu rGyalrong tɐ-ʁâv < *-q-

Tangut ɣa R17 1.17 < *Cʌ-qa

196: DP mr- (sometimes mʐ-) is the only Cr-cluster in the language. Languages are more likely to have Cr-clusters than Cʐ-clusters, but DP is an exception.

196: DP ɦ (which "sometimes varies wtih the voiced velar spirant [ɣ]") can sound like [ʕ]. I have hypothesized that Old Chinese and even Proto-Japonic might have had *ʕ, but I've never heard of it in a living East Asian language before.

Next: Dayang Pumi vowels.

APPENDICES

Possible examples of *ʕ- in Old Chinese (added 5.6.0:45)

GSR 589 is a pure emphatic series:

医 OC *ʕis > *ʔis 'quiver'

殹 OC *ʕis > *ʔis (a particle)

翳 OC *ʕi(-s) > *ʔi(-s) 'screen'

繄 OC *ʕi > *ʔi 'brown'

鷖 OC *ʕi > *ʔi 'goose'

My current hypothesis about emphasis requires me to reconstruct a low vowel presyllable as the source of emphasis in syllables with nonlow vowels:

*Cʌ-CI > *CI

Reconstructing *ʕ- allows me to derive emphasis from an initial without a presyllable regardless of the height of the main vowel:

*ʕV > *ʔV

Possible examples of *ʕ- in Proto-Japonic (added 5.6.0:50)

PJ *ʕana > *ànà 2.3 'hole'(voiced *ʕ- conditioned low pitch of later 'zero initial' syllables)

PJ *ʕita > *ìtá 2.4 'board' (ditto)

PJ *ʕahma-i or *ʕaʔma-i > *àmê 2.5 'rain' (ditto)

cf. PJ *ʔizo-i > *ísì 2.2 'stone' (voiceless *ʔ- conditioned high pitch of later 'zero initial' syllables)

(2.3, 2.4, 2.5, and 2.2 refer to Japanese pitch accent classes.)

*Is DP ʃtʃhju 'thief' cognate to Tangut

tʃɔ̃ɔ̃ R59 1.57 'to steal'

and

tʃhɔ̃ɔ̃ < *k-tʃ-? R59 1.57 'to steal'?

DP has no -ũ, so DP -u might partly be from *-ũ, and DP -u : Tangut -ɔ̃ɔ̃ might be a valid correspondence.

10.5.4.23:59: QAPUT?

In my last post, I implied that 'bitter' was originally uvular-initial in Proto-Sino-Tibetan, citing uvular-initial forms from two branches (Bodic and Qiangic) and reconstructing a possible uvular in a third branch (Sinitic).

Here's another uvular-initial etymology that I've mentioned before:

PST *m(ʌ)qo 'head'

Old Chinese 后 *ɢoʔ < *ɴɢoʔ < *ɴqoʔ < *m(ʌ)qoʔ 'ruler' (< 'head')

Classical Tibetan mgo

Tangut ɣʊ R4 1.4 < *Cʌ-q/ɢu (Grade II which may [partly] originate from uvulars and/or emphasis; R4 has mostly velar initials which might be from earlier uvulars; see below)

Modern Qiangic languages preserve the hypothetical original point of articulation:

Mawo Qiang qə-patʂ

Taoping Qiang qə-potʂɿ

Ronghong Qiang qə-pɑtʂ

(pɑtʂ is 'round')

Pumi qho

The problem is that rGyalrong languages have a velar instead of the expected uvular. Forms from Jacques (2004):

Proto-rGyalrong *ku (my reconstruction based on his system)

Japhug tɯ-ku (root ku: e.g., nɤ-ku 'your head')

Somang tə-kó (p. 153, 171) or ta-kó (p. 233)

Zbu tə-kuʔ

Is the rGyalrong velar an innovation or a retention? If it is the latter, how can the uvulars in Qiang be explained?

后 has a uvular in Baxter and Sagart's Old Chinese reconstruction system since it belongs to a mixed stop-fricative phonetic series:

Is there any Tibetan-internal reason to reconstruct a uvular ancestor for mgo? Zhongu has a velar instead of a uvular in ŋgo 'head'. (But Zhongu does not have ɴɢ- which might have merged with velar ŋg-.)

The Tangut word belongs to a low-frequency rhyme. That made me initially think that the pre-Tangut form was not simply *Cʌ-k/gu or *Cʌ-k/go which would have become ɣəu R1 1.1. (Tangut ɣ could be a lenition of a voiceless or voiced stop.)

Frequencies of R1 and R4 compared

There is no ɣəu R1 1.1. The 15 : 0 ratio of ɣʊ R4 to ɣəu R1 must be significant. Oddly, kʊ R4 is more common than kəu R1, yet there are ten gəu R1 but no gʊ R4. Why do kh-, g-, and ʔ- cluster in R1 whereas k- and ɣ- cluster in R4?

{k-, x-, ɣ-} do not form an obvious natural class opposed to {ʔ-, kh-, g-}. Here is a convoluted attempt to explain this bizarre distribution pattern:

1. Pre-Tangut *Cʌ-ku and *(C(ʌ))-qu without lenition merged as kʊ R4. But why is the common word kəu R1 1.1 'therefore' not R4? Is it an archaism, or conditioned by an unusual preinitial?

(5.5.00:40: Could kəu R1 1.1 'therefore' simply be a loan from Chinese 故 'therefore'? I'm surprised that neither Gong Hwang-cherng nor Li Fanwen noted this possibility.)

LFW5890 kʊ R4 1.4 'loose' may be from PT *qu if it is cognate to Japhug rGyalrong kɯ-ɴɢu.

2a. PT *Cʌ-khu without lenition > khəu R1

2b. PT *C-qhu > khʊ R4 (Cf. Baxter and Sagart's derivation of Old Chinese velars from uvular clusters.)

3. PT *Cʌ-gu without lenition> gəu R1.

4. PT *Cʌ-ʔu > ʔəu R1

5a. PT *Cʌ-xu without lenition > xəu R1

5b. PT *(Cʌ-)qhu without lenition > xʊ R4

6. PT *(C-)ɢu or PT *Cʌ- + any velar or uvular + -u with lenition > ɣʊ R4

Possible sources of ɣʊ R4

The above table excludes PT *-H which is my hypothetical source for the Tangut rising tone. Syllables without PT *-H develop level tones in Tangut.

The 15 ɣʊ R4 (9 level tone, 6 rising tone) may have all been nonhomophonous in pre-Tangut.

Although R1-R3 have long counterparts, there is no long counterpart of R4.

(5.5.1:05: I have reconstructed R6 as -ɨuu, the long counterpart of -ɨu R2. However, I am no longer certain that R6 corresponded to R2 since R6 can be preceded by the non-Grade III initial kh-, and some Grade III initials occur before R7: e.g., tʃ- and ʃ-.)

Perhaps some R4 syllables originally had long vowels (or whatever was the defining characteristic of R5-R7 which Gong reconstructed as long).

APPENDIX: The initials of R4

m- and d- are out of place among the velars. There are no məu R1 or dəu R1, though other labial and dental initials can occur with R1:

This is suspicious. I am tempted to say that mʊ R4 and dʊ R4 were accidentally placed under the wrong rhyme, but I can't believe 14 tangraphs with those readings were misplaced. One or two, maybe. But 14?

10.5.3.23:30: VOWEL HARMONY IN OLD CHINESE AND TANGUT POLYSYLLABIC WORDS

I have hypothesized that the four grades of Tangut reflect earlier presyllables:

I have excluded Grade II for now.

The grade of Grade III/IV syllables is generally determined by initial class:

Exceptions include foreign syllables.

That was based on a similar hypothesis I've had for Late Middle Chinese grades since 2002*:

Grades I and II were *emphatic in Old Chinese whereas Grades III and IV were *nonemphatic:

(Grade IV here refers to 'primary' high front vowel syllables which had high vowels in Early Middle Chinese and Late Old Chinese, not 'secondary' high front vowel syllables which developed high vowels in LMC like 天 LMC *thien < EMC/LOC *then.)

Although Chinese Grade II is from *r, I still do not know what conditioned Tangut Grade II. I used to hypothesize three types of pre-Tangut presyllables:

I have since rejected the above table. Alveopalatals are correlated with Grade II. There is no reason that such initials would be associated with low vowel presyllables but not mid vowel presyllables.

In both Chinese and Tangut, main syllables might have harmonized with presyllables.

OC syllabic harmony may have operated within disyllabic as well as sesquisyllabic (1½-syllable) words.

Five years ago, I found that 83% of Old Chinese disyllabic noncompounds were

*emphatic-emphatic: e.g.,

螻蛄 *roka [ʀˁɔˁqˁɑˁ] 'mole-cricket'

鵜鶘 *liga [lˁɪˁɢˁɑˁ] 'pelican'

*nonemphatic-nonemphatic: e.g.,

麒麟 *gərən 'kirin'

貔貅 *bixu 'pixiu, a mythological animal'

If Tangut were like OC, I would expect Tangut disyllabic noncompounds to also mostly have syllabic (grade) harmony: e.g.,

na-raʳ 'tomorrow'

contains a Grade I-Grade I sequence, and could be from pre-Tangut *(Cʌ)-na-ra with low vowels. (I am assuming it is not a compound from 'night-pass'.)

However, yesterday I came across the disyllabic noncompound

bi-vəiʳ 'abundant' (< reduplication *bi-bir or *bi-bri or even *bi-rʌ-bi, all with lenition of medial *-b-?)

containing a Grade IV-Grade I sequence instead of

*bi-vɨiʳ (Grade IV-III; all high vowels)

*bəi-vəiʳ (Grade I; all nonhigh main vowels)

17% of Chinese disyllabic noncompounds are also disharmonic: e.g.,

鸕鶿 *ra-dzə [ʀˁɑˁdzə] 'cormorant'

蝙蝠 *pen-pək [pˁɛˁnˁpək] 'bat'

Some may be morphologically complex: e.g.,

扶桑 *ba-saŋ [basˁɑˁŋˁ] 'mythical tree'

looks like 桑 *saŋ 'mulberry' tree' plus a plant prefix *ba- which has a different spelling in

芙蕖 *ba-ga 'lotus'

芙蓉 *ba-loŋ 'hibiscus'

Others may be loanwords from languages without emphatic harmony.

Are disharmonic disyllabic noncompounds a minority in Tangut as well as Chinese? Can the above explanations also apply to harmonic exceptions in Tangut?

*For simplicity, I left out main syllables with uvular initials (*Q-) which always became *emphatic (> LMC Grade I/II) unless preceded by a high vowel presyllable:

This table predicts that *Q-initial phonetic series should mostly be emphatic, since it is unlikely that *Cɯ-QV outnumber *QV and *Cʌ-QV combined.

A possible *Q-series may be Schuessler's (2009) 1-1 which is mostly emphatic:

古 *(Cʌ-)qaʔ 'old'

苦 *(Cʌ-)qhaʔ 'bitter' (cf. Zhongu Tibetan qhɐ, Mawo, Taoping, and Ronghong Qiang qhɑ)

胡 *(Cʌ-)ɢa or N(ʌ)-qa 'dewlap'

居 *Cɯ-qa 'dwell'

Note that Baxter and Sagart would reconstruct a velar initial for this series since it lacks Middle Chinese *ʔ- and/or *x- that would go back to their Old Chinese *q- and *qh-.

As I write my posts on Tangut, I find myself pronouncing David Boxenhorn's alphacodes for tangraphs. It's as if alphacode were a language with 21 consonants and 5 vowels:

Alphacode consonants

Following hyphens and boldface indicate onsets. Preceding hyphens indicate codas. No consonant can be both an onset and a coda.

I pronounce c as [tʃ] (though it could be a palatal stop [c]), x as [x], y as [j], and q as [q], though now I'm inclined to pronounce q as [ʔ] so the 'language' can have 'vowel-initial' syllables (that actually have initial [ʔ]).

Alphacode basic vowels

Alphacode long vowels and diphthongs (vowel/glide sequences?)

All alphacode syllables must be bimoraic. Open syllables must have long vowels or diphthongs.

Diphthongs could also be interpreted as long vowels (without short counterparts) and vowel-glide sequences.

Alphacode syllable structure

Although bimoraic structure is common (e.g., Vietnamese) and consonants that can only occur in one position are common (e.g., Vietnamese again), is there any language which has onsets and codas in complementary distribution?

If alphacode were a real language and its ancestor allowed all 21 consonants as onsets and codas, here's how those 21 could have been reduced to 14 onsets and 7 codas:

*k- > q- [ʔ] (as in Hawaiian)

*x-, *s- > h-

*l- > y- or d-

*r- > z- (as in Vietnamese) or d-

*n- > d-

*m- > b-

All final stops (and affricates) > -k (as in Chaozhou and Fujian)

*-z > -s

*-f, *-v, *-h > -x

*-y, *-w rewritten as -i, -u

-n, -l, -r, -m are unchanged

Did David have any of this in mind when he chose his inventories of onsets and codas?