In "Why No -ọw?", I suggested that Tangut*-Ṿw might have merged with -ụ. If this were true, there should be pairs of related words that fit this pattern:

Word 1: -VwWord 2: -ụ
-iw-(j)ụ (< *-Ṿw)

The above pattern would be parallel to this established pattern of alternation (Gong 1981: 806):

-u ~ -ụ

Perhaps some *-ọw (and *-ẹw?) had become -ọ:

Word 1: -VwWord 2: -ụ
-iw-jụ (< *-ịw)
-ew-jọ (< *-ẹw)
-ow-ọ (< *-ọw)

Cf. Sino-Japanese:

[juu] < *iu

[joo] < *eu

[oo] < *ou

Unfortunately, I know of no word families that fit the above patterns.

Gong (1981) listed the following alternations involving rhymes that he now reconstructs with *-Vw:

34. -ej ~ 56. -ow

42. -iəj ~ 57. -iow

These alternations are "often" found within disyllabic reduplicative words:

Cow-Cej-type words:

dzow 1.56 dzwej 1.33 ?'sin, crime' < Middle Chinese 罪 *dzwəjʔ 'crime'

sow 1.56 swej 1.33 ?'small pieces' < Middle Chinese 碎 *swəjh 'shatter'

Ciow-Ciəj-type word:

nioow 1.57 niəj 1.41 'turbid, unclean' < ?Middle Chinese 泥 *nej 'mud', *nejh 'mired'

Native words of both types also exist.

It seems that Tangut speakers borrowed the Chinese words as CVj and then partly reduplicated them to create new CVw-CVj words. Presumably native CVw-CVj words are also partial reduplications of their second syllables.

The function of this reduplication is unknown. It reminds me of an Old Chinese reduplicative pattern *Ce-Co (Sagart 1999: 137) though the resemblance must be coincidental:

邂 逅 *greʔ-groʔ 'carefree and happy'

Gong noted a third reduplicative pattern in Tangut:

Cjii-Cjoo (rhymes 14 and 55):

gjii 1.14 gioo 1.53 'bite' < Middle Chinese 齧 *ŋet 'bite'.

(native words of this type also exist)

In this pattern, the rounded vowel appears in the second syllable rather than the first. But once again, the back vowel syllable was derived from the front vowel syllable, judging from the Chinese root of gjii-gioo.

These reduplicative patterns do not tell us anything about the phonetics or origins of -(i)ow or -oo.

Gong (1981) did not list any alternations involving -iw and -ew. When I discussed the Tangut words for 'one', I proposed that lew 1.43 < ?*CV-tek could be related to words ending in - 1.29/1.30 and -jịj 1.61. Here are similar sets:

'beat': TT1617 tew 1.43 : TT1562 lhjɨ̣ < ?*CV-C-thjɨ

'new' (identified by Jacques [2003: 13]): TT1465 sjiw 1.46 (< pre-Tangut *-k) : TT1113 sjɨ 2.28 : TT1522 sjwɨ 1.30

This set with -ow is reminiscent of the above sets and of the -Vw ~ -Vj reduplicative pattern:

'pillow' (identified by Jacques [2003: 13]): TT1920 ɣjow 1.56 (< pre-Tangut *-m) : TT4821 ɣjɨj 1.42

The 'beat' set is questionable, as it depends on whether lh- was really from a prefixed, lenited *th- (which in turn could have been from *s-t-?).

The words ending in might have once ended in *-ɨw (< *-ɨk in 'new') before *-w was lost after all achromatic vowels.

However, a similar explanation is difficult to apply to ɣjɨj 1.42 'pillow', unless it once ended in the unlikely final cluster *-jw. WHY NO -ỌW?

In Gong's reconstruction of Tangut, all seven vowels have lax and tense versions:

lax: i, e, ɨ, ə, a, u, o

tense: ị, ẹ, ɨ̣, ə̣, ạ, ụ, ọ

Both lax and tense nonlow, nonback vowels can occur before the glide -j:

lax: -ij, -ej, -ɨj, -əj

tense: -ịj, -ẹj, -ɨ̣j, -ə̣j

However, only three lax vowels can occur before the glide -w:

lax: -iw, -ew, -ow

Why aren't there any rhymes like

tense: -ịw, -ẹw, -ọw?

The tense vowels reflect earlier tense consonants deriving from even earlier clusters:

*CCV > *C̣V > *C̣Ṿ > CṾ

I cannot think of any reason for -w (or any of its sources: e.g., *-k and *-m) to prevent a preceding vowel from being tense.

I could speculate that pre-Tangut tense *-ịw, *-ẹw, *-ọw were so rare that they merged with PT lax *-iw, *-ew, *-ow before the Tangraphic Sea was compiled. However, I cannot think of any reason why there would be few clusters before *-iw, *-ew, *-ow in pre-Tangut.

Perhaps the solution involves another distributional oddity in Tangut: the rarity of nasalized labial vowels.

In Gong's reconstruction, only three syllables end in nasalized labial vowels:

tụ̃ 1.96 'winter' (< Tangut period NW Chn 冬 *tũ 'winter')

tsụ̃ 1.96 (a surname) (< Tangut period NW Chn 宗 *tsũ [a rare surname; usu. 'ancestor'])

xụ̃ 1.96 'red' (< Tangut period NW Chn 紅 *xũ 'red')

All other nasalized vowels are nonlabial: ĩ, ẽ, ã. (There are no nasalized versions of ɨ or ə, implying that PT *-ɨN, *-əN lost their nasality and merged with other rhymes.)

The tenseness of -ụ̃ is suspicious for three reasons:

- All three are loans from Chinese, and Tangut tense vowels were rarely used to approximate Chinese vowels. (See Gong 1981: 776-777 for examples.)

- There is no other tense nasalized vowel.

- Li Fanwen's (1997: 3) table of Gong's rhymes listed lax -ũ. (What does Gong's table on p. 604 of Thurgood and LaPolla [2003] indicate? I don't have access to a library or $340.)

I suspect that -ụ̃ is a typo. Hence I will write this rhyme as -ũ. (Jacques [2007] also wrote this rhyme as -ũ.)

This revision leads to an inventory of four lax nasal vowels: ĩ, ẽ, ã, and the very rare ũ. I assume that these all originated from earlier *-VN rhymes. Surely pre-Tangut must have had many words ending in *-uN and *-oN. I have never heard of a language that would not allow labial vowels before nasal codas.

Could Gong's -ow be reinterpreted as -õ? Like the other nasal vowels (ĩ, ẽ, ã), -õ would have no tense counterpart. -õ may have resulted from the merger of pre-Tangut rhymes like *-uN and *-oN.

However, Gong's -ow also had nonnasal sources: e.g., TT4157 BODY low 2.47 < ?*lowH is probably cognate to Written Tibetan lus 'body', and it is unlikely that *-ow and *-õ would merge as -õ. I suspect the merger went in the other direction: *-ow, *-õ (phonetically [õw]?) > -ow.

But what happened to *-ọw? Did it merge with -ụ?

Moreover, unlike the nasalized vowels, Gong's -ow also has a long counterpart -oow, and it would be awkward to reconstruct a single nasalized long vowel -õõ. I don't know of any external evidence for a nasal in that rhyme. Gong (1981: 776) found only one Tangut -oow loanword from Chinese, TT3504 ɕioow 1.57 ?熟 'ripe'* from very late NW Middle Chinese ?*ɕuk (transcribed in Tibetan as shug) or Tangut period NW Chn ?*ɕu 'ripe'.

I know of no evidence for reconstructing -iw and -ew with nasalized vowels, so the -õ solution cannot be recycled to explain why there is no tense -ịw or -ẹw. Did those rhymes merge into -(j)ụ?

*Glossed by Li Fanwen (1986: 372) as 慈善 'benevolent' and Shi et al. (1983: 488 and 2000: 135) as 識 'recognize'! Which of these glosses, if any, is correct? Nevsky (II: 123) has no gloss at all. LOW-ST AND FOUND: EVIL GRASS

If Tangut do 1.49/2.42 'poison' was borrowed from Chinese, what happened to the native Tangut word for 'poison'?

I think I found it last night:

TT3254 POISON* low 2.47

(the graph consists of GRASS + EVIL)

It has an initial l- like Old Chinese 毒 *luk 'poison' (and unlike the later stop-initial Chinese and Tibeto-Burman forms).

It also has a final -w that might be a trace of an earlier *-k. So was the pre-Tangut word for 'poison' *lok? Was I wrong to propose that pre-Tangut *-k disappeared after both labial vowels (u, o)?

I am not sure, because Tangut -ow often corresponds to rGyalrong and Chinese nasal-final rhymes. Perhaps the pre-Tangut word for 'poison' was *loŋ, with a velar nasal instead of a velar stop.

(For a similar correspondence in the opposite direction, compare Tangut sjiw 1.46 < *sjik 'new' with Old Chinese 新 *sin < *siŋ 'new'.)

*12.14.00:40: Li Fanwen (1986: 450) translated low 2.47 as 毒 'poison', 莠 'weeds'.

Shi et al. (2000: 280) translated it as 毒草 'poisonous weed'.

Perhaps the final nasal of pre-Tangut *loŋ reflects an earlier nasal-initial morpheme:

*lok 'poison' + *NV ?'thing', ?'grass' > *loŋNV > *loŋ

rGyalrong cognates suggest that Tangut -ow may have had nonnasal sources. If so, then low 2.47 may actually be cognate to Old Chinese 莠 'weeds' if it had a liquid initial: *luʔ. The final glottal stop could even corresponding to the Tangut rising tone which I suspect arose from earlier laryngeals.

(Schuessler [2007: 597] and I suspect that 莠 could have had an initial *j- which could link it to Written Tibetan yur-ma 'weeding'. Schuessler [2007: 597] also proposed an even earlier initial *w- which could allow 莠 to be related to *w- of 耘 *wən 'to weed'.) A POISON-OUS PROBLEM

(Last night, I fell asleep shortly before I finished writing the first version of this post, only to wake up when my laptop rebooted itself for some reason. I discovered that the post was completely gone and that I hadn't saved it at all! So I'm rewriting it from scratch with a lot of changes.)

Tangut has two words for 'poison':

TT1100 do 1.49 and TT3225 do 2.42

(One is presumably derived from the other via tonal alternation. See here for a different explanation that I have rejected.*)

Similar words are widespread in Sino-Tibetan:

gDong-brgyad rGyalrong tɤ-ndɤɣ

Somang rGyalrong tə-dók

both from proto-rGyalrong *-ɔk

Ronghong Qiang duə

(I can't find any Mawo or Taoping forms in Sun 1981's glossary)

Written Tibetan dug

Written Burmese tauk

Late Old Chinese *douk > Early Middle Chinese *dɔk > NW Late Middle Chinese *dɦok > very late LMC *tɦuk (cf. the Tibetan transcription thug) > Tangut period NW Chn ?*thu

Schuessler (2007: 216) reconstructed Early Old Chinese*dûk (= *duk in my notation) and Proto-Sino-Tibetan *duk. On the other hand, Sagart (1999: 31) explained why 'poison' should be reconstructed as Early Old Chinese *aluk (=*luk in my notation). I accept Sagart's reconstruction and regard the Tibeto-Burman words for 'poison' as loanwords from Chinese dating after emphatic *l- hardened to *d-.

If Tangut do 1.49/2.42 is a Chinese loanword, why doesn't it have a final -w corresponding to Chinese *-k?

Last night, I thought that do might have been borrowed after pre-Tangut *-k had shifted to -w:

Stage 1Stage 2Stage 3Stage 4
a hypothetical Tangut word*dok*dox*doɣ*doɰdow 1.54/2.47 (seven tangraphs with this reading exists, though I do not know whether any of those readings originated from an earlier *dok, as *-ow may have other sources - see below)
'poison'(not yet borrowed into Tangut)EMC *dɔk or NW LMC *dɦok borrowed into Tangut as *do after it had lost *-k/-x/-ɣ

(The first three stages of *-k are similar to the reflexes of Proto-rGyalrong *-k in Somang [-k], Zbu [-x], and gDong-brgyad [-ɣ].)

Gong (1981: 774) identified

TT4628 READ do 1.49

(even the graph looks like an imitation of Chn 讀 'read')

as a borrowing from Chinese (presumably from LOC/EMC *dok or NW LMC *dɦok). Like do 1.49/2.42 'poison', do 1.49 'read' also has no -w corresponding to *-k.

Tonight, I came up with a second explanation for the -w-lessness of 'poison' and 'read'. Perhaps I was wrong about *-k > -w after all chromatic vowels. Maybe this shift only occurred after palatal vowels (i and e):

Stage 1Stage 2Stage 3Stage 4
*-Vk*-Vx*-Vɣ*-Vɰ*-ANo trace of *-k
*-iw*-ɰ merged with *-w

V = any vowel

A = any achromatic vowel (ɨ, ə, a)

U = any labial vowel (u, o)

I = any palatal vowel (i, e)

In this scenario, the Chinese words for 'read' and 'poison' could have been borrowed at stages 1-3.

This hypothesis can be disproven if there are Tangut words ending in -ow that have cognates or Chinese sources ending in rhymes like *-ok.

Jacques (2003) found that Tangut -ow corresponded to six gDong-brgyad rGyalrong rhymes:

Tangut rhymegDong-brgyad rGyalrong rhymepotential Proto-rGyalrong sources
-orw 1.91-u*-o
-ow 1.54, -jow 1.56-o*-aŋ
-ow 1.54/2.47, -jow 1.56-om*-am
-iow 1.55-um*-om
-jow 2.48–ɤm*-ɐm
-jow 1.56–ɯm*-im, *-um, *-om

None of the gDong-brgyad rhymes end in *-ɣ < PGR *-k.

Gong (1981: 774-776) identified 18 Tangut words ending in -ow (in his current reconstruction) that were borrowed from Chinese. All of their Chinese sources ended in *-ng.

All this seems to indicate that pre-Tangut *-ok did not become -ow, and that Tangut -ow mostly arose from earlier nasal-final rhymes (?*-om, ?*-ong).

Note that Nishida (1964), Sofronov (1968), and Arakawa (1999) sometimes reconstructed final nasalized vowels (in bold) instead of -ow:

Tangut rhymeGongNishidaSofronovArakawa
56. 1.54/2.47-ow-õ-õ-õ
57. 1.55/2.48-iow-jõ-jõ
58. 1.56/2.48/2.49-jow-jwõ-jõ-õõ
59. 1.57-oow-jwo-juo
60. 2.50-joow-juo < *-juoC-jö
97. 1.91/2.82-orwr-ụo < *-ụoC-oor
98. 2.83-jorw(none)-jụo < *-jụoC-wor

Unlike the others, Gong reconstructed only a single nasalized labial-vowel rhyme**: -ũ 1.96, which appears in readings for a mere three tangraphs.

12.13.1:16: I have just found another Tangut word for 'poison' which complicates the picture even further. I don't want to fall asleep at my computer again, so I'll post about it later.

*12.13.1:14: Either do 1.49 or do 2.42 is borrowed from a Chinese word for 'poison': e.g., Late Old Chinese *douk. Could the other word be based on a Chinese word 'to poison' with an *-s suffix?

EOC *luk-s > LOC *douh > EMC *dawh > NW LMC *dɦàw

There are three problems with this hypothesis.

First, 'to poison' is only attested today in southern Chinese languages. It could have also existed in the northwest, but there is no independent evidence for this.

Second, neither do 1.49 nor do 2.42 is a verb.

Third, Middle Chinese *-aw corresponds to Tangut -ew, not -o. One could evade this objection by claiming that it was LOC *douh that was borrowed into pre-Tangut as *do(H), but the first two objections remain intact.

**I assume the tilde in

TT4443 EVIL khiõ 1.50

is a typo, as it is the only instance of -iõ 1.50 which is otherwise reconstructed with an oral vowel as -io. In Gong's reconstruction, nasal vowels are only found in loanwords, and EVIL is presumably native, as I don't know of any similar-sounding Chinese word.

Homophones listed EVIL as a homophone of TT5353 SKILLFUL khio 2.43 with an oral vowel (and a different tone). The oral quality of the vowel can be confirmed from its probable Chinese source: Tangut period NW Chn 巧 *khjaw. NO UVULAR CODAS IN PRE-TANGUT?

Proto-rGyalrong distinguished between *-k and *-q (Jacques 2004: 266). PGR *-k could occur after all eight PGR vowels, but PGR *-q had a more restricted distribution (Jacques 2004: 266):

*-iq(no *-ɯq)(no *-uq)
(no *-eq)*-oq

It is strange that *-q could only occur before a high front vowel (*i). I thought that *-iq might actually have been *-eq. That revision would allow me to claim that *-q could only occur after non-high vowels. However, it would not explain why the descendants of *-eq tend to have high vowels (Jacques 2004: 261):

gDong-brgyad, gSar-rdzong, Da-tshang -ɯɣ

Eastern rGyalrong -ik

but Zbu -eχ

I would expect *i to lower to e (as in Zbu) before *q, not the other way around. So I conclude that *-iq is the correct reconstruction.

Did pre-Tangut have a similar distinction between velar and uvular codas? If PT only had a single *-k corresponding to PGR *-k and *-q, PGR *-q should have the same pattern of correspondences as PGR *-k:

zero after Tangut achromatic vowels and -u (< *-uw < *-uk)

-w after Tangut chromatic vowels other than -u

There was no PGR *-uq, so this pattern could be simplified as

zero after Tangut achromatic vowels

-w after Tangut chromatic vowels

In gDong-brgyad rGyalrong, the five PGR *-q rhymes have collapsed into three:

PGR *-aq, *-ɐq, *-ɔq > DG -aʁ

PGR *-oq > DG -oʁ

PGR *-iq > ?*-ik > DG -ɯɣ

These three rhymes have the following correspondences:

DG -aʁ : Tangut -a 1.17/2.14, -ia 2.15, -ja 1.20/2.17, -jaa 1.21, -aa 1.22, -ạ 1.63/2.56

DG -ʁ < PGR *-q, like DG *-ɣ < PGR *-k, corresponds to zero after Tangut a

DG -oʁ : Tangut -o 2.42 (possibly also -o 1.49), -ior 1.90

DG -ɯɣ : Tangut -e 2.7

in DG kɤ-sɯɣ 'tight': this is from PGR *-iq because the Zbu cognate is kɐ-séχ with a uvular coda

Note that Tangut has -e instead of -ew and -o(r) instead of -ow. This may mean that PT *-q disappeared after all vowels, unlike PT *-k which survived as -w after chromatic vowels other than -u.

(12.11.2:33: There is a 'louse'-y counterexample to this claim which I will investigate later.)

Did Old Chinese also have a final uvular? OC 腦 *n 'brain' corresponds to

GD tɯ-rnoʁ, Zbu tə-rnôχ < PGR *rnoq

Tangut no 2.42 < PT ?*noq-H

(PT *nok-H would have resulted in Tangut now 2.47.)

Proto-Austronesian *punuq (was this word borrowed into Sino-Tibetan?)

This one instance implies that OC *-ʔ partly originated from *-q. (OC *-ʔ is so common that it cannot be from a final uvular in all cases.) One would expect PGR *-q to always correspond to OC *-ʔ, but it also corresponds to OC *-k:

'weave': GD kɤ-taʁ < *-q : OC 織 *tək

'shame': GD tɯ-zraʁ < *-q : OC 色 *srək 'color' (not the best semantic match)

'pea': GD stoʁ < *-q : OC 菽 *stuk 'bean'

'Brain' is an isolated anomaly. Could its possible foreign origin be relevant?


In "W-hy Not?", I described the two paths travelled by pre-Tangut *-k:

pre-Tangut *-Vk > *-Vɣ > *-Vɰ > Tangut -Vw (> -u if V = u)

(V = any chromatic vowel)
pre-Tangut *-Ak > *-Aɣ > *-Aɰ > Tangut -A

(A = the achronmatic vowels ɨ, ə, a)

Pre-Tangut *-p may have travelled along two similar paths:

pre-Tangut *-Vp > *-Vβ > Tangut -Vw (> -u if V = u)

(V = any chromatic vowel)
pre-Tangut *-Ap > *-Aβ > *-Aw > *-Aɰ > Tangut -A

(A = the achronmatic vowels ɨ, ə, a)

In Jacques (2003), gDong-brgyad rGyalrong (DR) -β corresponds to Tangut zero except after Tangut e:

gDong-brgyad rGyalrong rhymeTangut rhymesNotes
-aβ < Proto-rGyalrong *-ap-a 1.17, -ia 1.18, -ja 2.16, -ạ 1.63Tangut achromatic vowels -a and
-ɤβ < Proto-rGyalrong *-ɐp, *-ɯp, *-ɔp (impossible to determine PGR rhymes from DR alone)-jwɨ 1.30, -jɨ̣ 1.69
-ju 2.3, -jụ 1.59Tangut -u < *-uw < *-uβ < *-up?
-wew 1.43Tangut palatal vowel in Sofronov and Gong's reconstructions. Nishida (1964: 54) reconstructed an achromatic schwa instead of e in this rhyme. If Nishida is correct, did a pre-Tangut *e shift to schwa after the *-Aɰ > -A change occurred?

There is only one example of DR -ɤβ corresponding to Tangut -wew 1.43.

I would be more confident about my account of what happened to pre-Tangut *-p if there were Tangut forms ending in-iw (< ?*-ip) and -ow (< ?*-op) that had gDong-brgyad rGyalrong cognates.

DR has a third rhyme ending in -β (-ɯβ), but Jacques (2003) did not propose any Tangut cognates of DR -ɯβ words.

12.10.1:33: I tried to find Tangut cognates for the DR -ɯβ words listed in Jacques (2004: 242), and this is the best I could do:

GlossgDong-brgyad rGyalrongTangutOther cognates
shadowta-ʁjɯβ < PGR *wljip (with an *-l- unlike the others!; cognate to 'be dark' below?)TT4694 rar 2.76 (low vowel does not match)Written Tibetan grib-ma, Written Burmese a rip
be darkkɯ-lɯβ < PGR *lupTT1392 ljwɨ̣j 1.62 (the -j is unexpected)Zbu rGyalrong kə-ldôv
sewkɤ-tʂɯβ < PGR *trupTT1557 r 1.84 (unrounded vowel does not match)Somang rGyalrong ka-tʂóp, WT Hdrub-pa, WB khyup
sleepkɤ-nɯʑɯβ < PGR *jipTT2444 ʔjɨ 2.24Zbu kɐ-rɐjîv, WB ip

Three out of the four Tangut words have problems. None end in -w because they all have achromatic vowels.

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2007 Amritavision