Why does all of 2bie 'high' plus the center and right of 1riəʳ 'bone' (from my last entry) equal ...


... the character for the second syllable of

Tangut 1giuʳ 1kwæ 'kidney'?

That analysis from the Tangraphic Sea makes no semantic or phonetic sense.

The Tangraphic Sea analysis for the first half makes more sense. 1giuʳ is written as the right of 1tʂʰɨi 'meat' (indicating an organ) plus the left of 1nie 'red' (the color of a kidney):


I don't know of any external cognates for 1giuʳ 1kwæ 'kidney'. Could the word be substratal?

12.22.23:31: Without more data, I cannot narrow down the possible reconstructions of 'kidney' in pre-Tangut:

1giuʳ < *Tgu, *gur, *TNKu, *NKur?

1kwæ < *krwa, *qwa, *Pkra, *Pqa?

No, wait, those syllables resemble Sino-Tibetan forms such as

Jingpho n-hkyun 'kidney' (more cognates here)

Written Tibetan mkhal-ma 'kidney' (more cognates here)

Is the Tangut word for 'kidney' from a redundant compound like *T-N-kjun-P-qal? Medial *-j- and the codas *-n and *-l could not be reconstructed solely on the basis of Tangut. *ju and *u (when not preceded by a low-vowel presyllable) merged into ɨu or iu generally depending on the preceding consonant. *-n disappeared without a trace after *-u, and *-l disappeared everywhere.

Old Chinese 腎 *ginʔ 'kidney' may be from an earlier *N-kinʔ. Is its vowel from Proto-Sino-Tibetan *-ju- (via an intermediate stage *-y-)? STRESS AND SCHWA IN TANGUT: A BRIGHTENED BONE AND A THIEVING TRIO

When looking for more examples of the 'brightening' of pre-Tangut back rounded vowels in Jacques' (2006) Japhug-Tangut comparisons, I found two word families:


Japhug ɕɤ-rɯ < *-u

Tangut 2rieʳ < *CI-ro-H and 1riəʳ < *Cɯ-rə

also cf. Written Tibetan རུས་ rus

'to steal'

Japhug kɤ-mɯ-rkɯ < *-u

Tangut 2kuʳ < *rku-H, 1kwiəʳ < *Pɯ-rkə, 2kiʳ < *CI-rku-H or *rI-ku-H

also cf. Written Tibetan རྐུ་ rku and Old Chinese 寇 *kʰos

What is the origin of the alternations between pre-Tangut and other vowels?

And why does Tangut have schwa in

1ŋwə 'five'

instead of an -a as in

Written Tibetan ལྔ་ lnga

Written Burmese  ငါး <ŋaḥ>

Old Chinese 五 aʔ

If there were a 'five' word family in Tangut, at least one member would have -a.

Perhaps these schwas are reductions of earlier unstressed vowels that later became stressed:

'bone': *Cɯ́-ro > *Cɯ́-rə > *Cɯ-rə́ > *Cɯ-rɨə́ > 1riəʳ

*C- might have been a fricative like Japhug ɕ-.

'to steal': *Pɯ́-rkə > *Pɯ́-rkə > *Pɯ-rkə́ > *Pɯ-rkɨə́  > *Pɯ-kɨə́ʳ  > *Pkɨəʳ > 1kwiəʳ

'five': *Pʌ́-ŋa > *Pʌ́-ŋə > *Pʌ-ŋə́ > *Pŋə > 1ŋwə

On the other hand, originally stressed vowels were not reduced to schwa:

'bone': *CI-ró-H > *CI-rǿ-H > *CI-ré-H > *CI-rié-H > 2rieʳ (not *2riəʳ)

'to steal': *CI-rkú-H > *CI-rký-H > *CI-rkí-H > 2kiʳ

or *rI-kú-H > *rI-ký-H > *rI-kí-H > *rki-H > 2kiʳ

I could even try to reconcile different initials for 'five' -

Hakha pa-ŋâa

Ao phaŋa < *pr- < *pl-?

Atong banga [baŋa]? < *bl- < *pl-?

Lavrung mŋá < *pŋ-?

But what about forms like Meluri manga in which m- is not next to -ŋ? Could the first -a- be epenthetic rather than original?

Written Tibetan lnga

- by reconstructing *pláŋa at the Proto-Sino-Tibetan level. However, the southern dialect of Old Chinese that was the source of Proto-Tai *haːC 'five' may have had a prefix with a voiceless nonlabial initial like *s- or *k- that fused with *-ŋ-:

*H-ŋaʔ > *ŋ̊aʔ > *haʔ.

I do not know whether such a prefix can be constructed in the ancestor of all Old Chinese dialects.

12.21:23:36: If Sino-Tibetan is related to Austronesian as Sagart proposed, could my Proto-Sino-Tibetan *pláŋa and Proto-Austronesian *lima 'id.' share an ancestor?

*lipáŋa >

*lpáŋa > PST *pláŋa

*lipaŋá > *lipŋá > PAN *lima

I could even try to somehow link Shorto's (2006) Proto-Mon-Khmer *pɗam (> Khmer pram) 'five' and even Proto-Indo-European (!) **pénkʷe 'five', but I'd rather stop. This demonstrates how imagination can make long-distance comparisons seem plausible. PROTO-SINO-TIBETAN *T-K(J)AJ 'CRAB'?

Last night, I proposed that the retroflex vowel of Tangut

2kiʳw (Tibetan transcription: rki) 'elbow'

was from pre-Tangut *T- which could have been a *t- corresponding to the (*)t- of cognates like Somang tə-krú and Old Chinese 肘 *t(k)ruʔ 'id.' (The retroflexion could not be from pre-Tangut medial *-r- which did not condition retroflexion in Tangut.)

I have also proposed *T- as a source of medial *-r- in Old Chinese (OC) because I find it hard to believe that Old Chinese originally had medial *-r- in a large number of words. I suspect some of those *-r- are secondary. For instance, I think the *-r- of OC 蟹 'crab', reconstructed by Zhengzhang (2003) as reːʔ and by Baxter and Sagart (2011) as *m-kˁreʔ, originated from a *t- corresponding to the t- in other Sino-Tibetan words for 'crab':

Padam ta-ke

Khami tə-ai

Lepcha tă-hi

Note, however, that not all Sino-Tibetan words for 'crab' have t-prefixes: e.g.,

Mindat ng-ai 

Hakha tsǎŋ-ʔâay

Black Lahu á-ci-ku

According to Matisoff (2003: 139), Benedict projected a dental prefix back into Proto-Tibeto-Burman *d-k(y)aːy 'crab'. I don't believe in a Tibeto-Burman subgroup in Sino-Tibetan, so maybe *t-k(j)aj can be reconstructed at the Proto-Sino-Tibetan (PST) level. (I prefer *t- to *d- since no daughter language has d- in 'crab'. I am reluctant to reconstruct a long vowel solely on the basis of long vowels in Chin forms like Hakha tsǎŋ-ʔâay.)

Here is how PST *t-k(j)aj 'crab' could have become Baxter and Sagart's (2011) Old Chinese *m-kˁreʔ:

1. Suffixation: *t-k(j)aj-ʔ

The final glottal stop could be a trace of some other stop *-C or an unstressed *-CV syllable.

I am hesitant to reconstruct *-ʔ at the PST level since it is apparently only in Chinese.

2. Monophthongization: *t-keʔ

The correspondence between OC 八 *pret 'eight' and Written Tibetan brgyad 'id.' could imply that OC *e is partly from *ja.

Pulleyblank (1991) reconstructed OC *-aj instead of *-e as in other recent reconstructions.

3. Preinitial lenition: *r-keʔ

The shift of *tC- to *rC- is parallel to the shift of *p(V)r- to *wVr- in some rGyalrongic varieties: compare OC 八 *pret 'eight' to

甘孜新龙甲拉西 Ganzi Xinlong rGya rwa gshis ˋɟe (why a superscript b?)

壤塘蒲西斯遥吾 Rangtang Puxi siyaowucun βjot

金川观音桥 Jinchuan Guanyinqiao Wobzi vjɑ́

阿坝柯河 Aba Kehe wə'rɟet

which exemplify different points in a continuum of lenition.

4. Metathesis: *kr

5. Prefixation: *m-kreʔ

*m- is an Old Chinese prefix for the "names of small animals" (Sagart 1999: 85).

No other Sino-Tibetan forms in the STEDT database have m-, so I assume it is a Chinese innovation.

6. Emphasis: *m-kˁreʔ [mˁqˁʀˁeˁʔ]? (the phonetic interpretation is my own)

Old Chinese syllables with lower vowels (*e *a *o) became emphatic (pharyngealized).

Twelve more steps are needed between OC *m-kˁreʔ and modern standard Mandarin xiè:

7. Assimilation and fusion: *ŋkˁreʔ > *ŋgˁreʔ > *reʔ [ɢˁʀˁeˁʔ]

8. Lowering after [ʀˁ]: *rɛʔ [ɢˁʀˁɛˁʔ]

9. [ʀˁ]-loss: *ɛʔ [ɢˁɛˁʔ]

10. Loss of emphasis: *gɛʔ [ɢɛʔ]

[ɢ] is an allophone of Late Old Chinese */g/ before nonhigh vowels.

11. Voiced uvular lenition: *ɣɛʔ [ʁɛʔ]

[ʁ] is an allophone of Early Middle Chinese */g/ before nonhigh vowels.

12. Glottal stops condition creaky phonation: *ɣɛˀ

13. Breaking: æjˀ

14. Initial devoicing and voiced aspiration: *æjˀ

15. Assimilation (creaky to breathy after voiced aspiration): *æjʰ

16. Tonogenesis: breathy phonation to 'grave' tone: *xʱæ̀j

The 'grave' tone is now a falling tone in modern standard Mandarin, but there is no guarantee that it had that shape at this point, so I use a term that merely refers to a grave accent, not to the contour of the tone.

17. Loss of voiced aspiration: *xæ̀j

18. Breaking before velars: *xj

This restores the original *ja lost in step 2 above.

19. Monophthongization: *xjè

History repeats itself; cf. step 2 above.

20. Palatalization: xiè [ɕje˥˩]

Cantonese haai5 'crab' is the result of somewhat different developments after step 12:

13'. Velar to glottal: *ɦɛˀ

14'. Breaking: æjˀ

Same as step 13 in Mandarin.

15'. Tonogenesis: creaky phonation to 'acute' tone: ǽj

The 'acute' tone is now a rising tone in modern standard Cantonese, but there is no guarantee that it had that shape at this point, so I use a term that merely refers to an acute accent, not to the contour of the tone.

16'. Devoicing and tonogenesis: from voiced initial + 'acute' tone to voiceless initial + lower 'acute' tone: *hæ̋j

A lower allophone of the 'acute' tone after voiced aspirates became phonemic once voiced initial obstruents devoiced.

I use a double acute accent to indicate the lower 'acute' tone.

17'. Centralizing and lengthening: haai5 [haːj˩˧]

Lengthening could also have occurred at some earlier stage.

Could Korean 게 ke < kəj 'crab' be an old loan from Chinese? (Its Sino-Korean equivalent is 蟹 hae < haj, borrowed from an ancestor of Mandarin at step 11: *ɣæjˀ.) Or is it somehow related to Japanese kani 'crab'?

12.20.23:21: Vovin (2010: 150) wrote,

Since the Middle Korean sequence [C]Vni within the same morpheme is practically unattested, a reconstruction of MK :key [kəj˩˥] as PK [Proto-Korean] *keni [kəŋi] seems to be viable, but we must keep in mind that the Ceycwuto dialect has the form kəŋi ‘crab’, and other forms with -ŋ- sporadically appear in other dialects (Choy 1978: 946-947). This medial -ŋ- is unlikely to be from PK *-nk-, but the reconstruction of the Proto-Korean form is not quite clear. Tentatively, I accept this etymology [relating it to Japanese kani], but it still may be rejected in the future once we have a better understanding of the Proto-Korean reconstruction.

If Japanese borrowed 'crab' from Korean, I would expect *koni < *kəni or perhaps even *kogi given Unger's (2008) proposal to derive Japanese g from *ŋ. Maybe the word was borrowed from *kaŋi in some extinct Koreanic language on the peninsula. The borrowing may have taken place before Proto-Japonic left the peninsula because Japanese kani also has cognates in Ryukyuan: e.g., Okinawan gani, Nakijin gai, and Yamatoma Amami and Hirara Miyako -gaN.

12.21.1:05: Those forms are from the Ryukyuan Phonetic Database which apparently does not contain Amami and Miyako forms for 'crab' in isolation. See Thorpe (1983: 274) for more Ryukyuan forms: e.g., Yamatoma kaN. (Unfortunately, Thorpe 1983 has no forms for Hirara.) There is a Ryukyuan tendency to voice the initial consonants of animal names (Curry 2004: 74). Could that voicing be a trace of an earlier nasal prefix?

12.21.1:30: The medial -ŋ- in Korean dialects rules out a Chinese origin for 게 ke < kəj 'crab' since there is no medial nasal in early Old Chinese *t-keʔ or in its descendants. THE ELBOW ENIGMAf

If I had no other evidence, I would derive Tangut

2kiʳw (Tibetan transcription: rki) 'elbow'

from pre-Tangut *TkiwH or *TkikH. However, other Sino-Tibetan words for 'elbow' indicate an earlier *-u: e.g.,

Written Tibetan gru-mo

Somang tə-krú

and perhaps Old Chinese 肘 *t(k)ruʔ (there is no internal evidence for *k, though Zhengzhang 2003 and Baxter and Sagart 2011 both reconstruct *tk-)

Since I recently proposed that

*CI-Cu > *CI-Cy > Ci

I would like to reconstruct pre-Tangut *TI-kruH (phonetically [ti-kruʔ]?):

*T- became *r- which then conditioned the retroflexion of the following vowel:

*TI-krV > *Tk(r)V > *rk(r)V > *rk(r)Vʳ > *k(r)Vʳ

*-I- conditioned the fronting of *-u-

medial *-r- disappeared

*-H conditioned the second (i.e., rising) tone

However, that would not account for Tangut -w which should go back to *-w or *-k. Nagano and Prins' rGyalrongic Languages Database lists forms ending in -k: e.g.,

宝兴硗碛 Baoxing Qiaoqi Yuojie nə'krik̚

甘孜丹巴半扇门 Ganzi Danba Banshanmen tə'kriʔk (was pre-Tangut *-kH phonetically [ʔk]?)

The final fricative of 观音桥 Guanyinqiao Lavrung lkhɑ́ɣ could be from *-k.

Perhaps the glottal stops of forms like 宝兴硗碛勒乐村朵果组  Baoxing Qiaoqi Leilecun Duoguozu 44krɨʔ44 are also from *-k.

Was a suffix *-k an innovation shared by Tangut and rGyalrongic?

Is the final -t of 金川二楷 Jinchuan Erkai ɬqʰet̚  a different suffix? Here's a wild idea: what if 'elbow' originally ended in *-kt which simplified in different ways?

Tangut: *-kt > *-kk > *-ʔk > *2-k (tonogenesis) > *2-ɣ > *2-ɰ > 2-w

Jinchuan Erkai *-kt > -t

Other rGyalrongic varieties: *-kt > -k (> or -ʔ)

Still other rGyalrongic varieties have no codas in 'elbow': e.g., Japhug tɯ-zgrɯ (Jacques 2006: 17). AN OATH TO GRAZE SIX GASES UNDER A JUNIPER TREE

Last night, I wrote that I didn't have examples of

*CI-Cu > *CI-Cy

*CE-Cu > *CE-Cy

*CE-Co > *CE-Cø

in pre-Tangut. Now I do.

Today I realized that

1tʂʰɨiw < *KI-tryk < *KI-truk 'six'

was an example of the first change. Its *u fronted to *y  after *I, lost its labiality as i, and gained -ɨ- before a retroflex initial. The sequence *K-tr- is like that of forms such as Somang kə'tɽok̚. One can find many more rGyalrongic forms by looking up 'six' in the left-hand menu of this page of Nagano and Prins' rGyalrongic Languages Database.

Other Sino-Tibetan forms with u in 'six' are Old Chinese 六 *ruk and Written Tibetan drug.

Another example of the first change is

1ʐɨiw < *CI-šyk < *CI-šuk 'juniper'

cognate to Zbu xɕôx and Written Tibetan shug-pa 'id.'

2ŋwəi < *PE-ŋyH < *PE-ŋuH 'oath' (cf. Japhug kɯ-jŋu, Somang kə-jŋó ká pa 'id.' from Jacques 2004: 304 and 2006: 21)

is an example of the second change. Its *u fronted to *y  after *E, lost its labiality as i, and partly lowered to after a nonhigh presyllabic vowel.

1ɬew < *KI-løk < *KI-lok 'to graze' (cf. Somang ka-lôk, Zbu kə-ltʰôx 'id.' from Jacques 2004: 260)

1lwew < *PI-løk < *PI-lok 'gas' (cf. Somang ta-jlôp, Zbu tɐ-rjôx 'vapor' from Jacques 2004: 323)

are examples of the third change. *o fronted to after *I, and lost its labiality as e. Did *-p dissimilate to *-k after the labial vowel *o in Tangut 'gas' and Zbu 'vapor'? (If so, Zbu *-k then lenited to *-x.)

Four out of these five examples end in -w < *-k. 'Oath' is the sole exception. Is the Tangut word for 'oath' an unrelated lookalike of the rGyalrong forms? Or are more examples ending in other codas waiting to be found?

12.18.3:00: None of the rGyalrong forms have a front vowel in their presyllables (and the presyllable of the Zbu word for 'juniper' was reduced to a preinitial x-). Did pre-Tangut preserve presyllabic vowel distinctions lost in rGyalrong? Part of the mismatch in presyllable vowel quality is due to different prefixes: pre-Tangut had *P-prefixes in 'oath' and 'gas', but their rGyalrong cognates have k- and t-prefixes. HOW PRESYLLABIC VOWELS 'BRIGHTENED' TANGUT IN THE YEAR OF THE HIGH HORSE

Tangut has some words with -ie where o might be expected (see 5 below). I think these -ie are from an earlier *-ø which in turn was from an *o that was fronted after a palatal vowel in a presyllable. That's just one part of a larger scenario which I'll summarize below.

1. A very early ancestor of Tangut (Proto-Sino-Tibetan?) had disyllabic words with the basic shape CVCV́(C). CV- was often an unstressed prefix added to a stressed monosyllabic root CV́(C).

At this point, both the first and second V could be one of six vowels:

front/palatal central/neutral back/labial
high *i *ə *u
low *e *a *o

'High' and 'low' are relative phonological terms, not phonetic terms.

2. The first syllables were reduced to presyllables when their unstressed vowels were reduced to four:

unstressed front back
high *I (< *i) (< *ə, *u)
low *E (< *e) (< *o, *a)

The symbols for the front presyllabic vowels indicate their effects on *a (see 3a below).

One could also write these four unstressed vowels with breves to distinguish them from similar stressed vowels: *ĭ *ĕ *ə̆ /*ŭ *ă/*ŏ.

I will write hyphens after presyllables. These hyphens often correspond to morphemic boundaries.

3. Schwa shifted from the high to the low class. (Maybe it was originally *[ɨ].)

The presyllabic vowels conditioned changes in the following stressed vowel:

a. front unstressed vowels caused *a to raise to different degrees:

*CI-Ca > *CI-Cia > *CI-Ci

*CE-Ca > *CE-Cea > *CE-Ce

I don't know what they did to *ə.

b. front unstressed vowels caused back/labial vowels to front:

*TI-kok > *TI-køk 'year'

*mI-roH > *mI-røH 'high'

*mI-ro > *mI-rø 'horse'


*CI-Cu > *CI-Cy

*CE-Cu > *CE-Cy

*CE-Co > *CE-Cø

thought I don't have examples yet.

These front rounded vowels then lost their labiality. *e broke to *ie after *I.

*TI-køk > *TI-kek > *TI-kiek 'year'

*mI-røH > *mI-reH > *mI-rieH 'high'

*mI-rø > *mI-re > *mI-rie 'horse'


*CI-Cy > *CI-Ci

*CE-Cy > *CE-Ci

*CE-Cø > *CE-Ce

c. back high unstressed vowels caused stressed low vowels to break into rising diphthongs:

*Cɯ-Ce > *Cɯ-Cie

*Cɯ-Cə > *Cɯ-Cɨə

*Cɯ-Ca > *Cɯ-Cɨa

*Cɯ-Co > *Cɯ-Cuo

They had no effect on high stressed vowels:

*Cɯ-Ci > *Cɯ-Ci

*Cɯ-Cu > *Cɯ-Cu

d. back low unstressed vowels caused high stressed vowels to break into falling diphthongs:

*Cʌ-Ci > *Cʌ-Cei > *Cʌ-Cəi

*Cʌ-Cu > *Cʌ-Cou > *Cʌ-Cəu

They had no effect on low stressed vowels:

*Cʌ-Ce > *Cʌ-Ce

*Cʌ-Cə > *Cʌ-Cə

*Cʌ-Ca > *Cʌ-Ca

*Cʌ-Co > *Cʌ-Co

4. The presyllables either fused with the initials of stressed syllables or were lost:


*mI-rieH > *mrieH > *mbrieH > *brieH 'high'


*TI-kiek > *kiek 'year'

*mI-rie > *rie 'horse'

5. After various other changes (*-r- > zero, *-H > 2-, *-Ø > 1-, *-k > -w, *rV > rVʳ):

2bie 'high', 1kiew 'year', 1rieʳ 'horse'

Many more vocalic changes have been excluded for simplicity: e.g, the development of Grade II vowels after uvulars and *Cr-clusters, the nasalization of vowels, etc.

Cognates in other languages preserve rounded vowels: e.g.,

'high' (see more):

Longxi Qiang

Mianchi Qiang bʐú < *br-

'year' (see more):

Namuyi kuəʴ55  (the retroflexion could be a trace of the *TI-presyllable which might have had initial *r-)

Ergong ko

'horse' (see more):

Longxi Qiang ʁò

Maerkang rGyalrong mbro

o-type vowels in 'high' and 'horse' are from an even earlier *-aŋ: cf. Written Burmese မြင့် <mraŋ'> 'high' and မြင်း <mraŋḥ> 'horse'.

12.17.0:52: It is tempting to project the shift of *-aŋ to *-o back to Proto-Qiangic, but a similar change occurred toward the end of the first millennium AD in northwestern Chinese, the next-door neighbor of Tangut. There is no way that Proto-Qiangic could only be a thousand years old. Perhaps the shift diffused through the region over a long period of time. TANGUT IN YU'S (2012) "PROTO-ERSUIC"

Last week I got a copy of Dominic Yu's 2012 dissertation on Proto-Ersuic. It contains three sections referring to Tangut.

1. Did Proto-Ersuic and Tangut raise *-a to -i at the same time, or did 'brightening' occur independently in each language? Yu favors the former scenario on page 48:

One way of estimating the time depth [of Proto-Ersuic] is to look at Tangut, which like Proto-Ersuic had undergone the brightening change of PTB [Proto-Tibeto-Burman; I would change that toie Proto-Sino-Tibetan] *-a > -i. Since Tangut is documented since the eleventh century, Proto-Ersuic should also date to that time, assuming the brightening change was historically the same change (either a shared innovaton in a common ancestor or an areal change that spread through the region).

2. How is Tangut related to Proto-Ersuic? Here I simplify two Qiangic subgrouping proposals from pages 210 and 212:

Sun's (2001) subgrouping

Northern Southern
rGyalrongic Qiang proper Tangut Zhaba Guiqiong Ersuic

Jacques and Michaud's (2011: 6) subgrouping

Na-Qiangic Lolo-Burmese
Naish Ersuish Qiangic Loloish Burmish
Tangut Zhaba Qiang proper rGyalrongic Burmese

As for Guiqiong, Jacques and Michaud wrote on page 5,

Further research will also be necessary to clarify the relationship of Guiqiong and Tujia to the Burmo-Qiangic group as defined here.

Tangut is Qiangic according to both proposals, but Ersuic (equivalent to Jacques and Michaud's Ersuish) is a sister rather than a daughter of Qiangic in the latter proposal.

3. Yu's (2012: 216) Proto-Ersuish *Cba and *kʰu, both 'year', resemble Jacques and Michaud's Proto-Tangut *C-pja (in 'this year, next year, last year') and *kjuk (with numerals). The latter forms correspond to my pre-Tangut *CI-Pa and *CI-kok, ancestral to Tangut

1vɨi and 1kiew

I'll discuss the vowels of my pre-Tangut forms in detail in my next post.

Tangut fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2013 Amritavision