All of my *S and *r-alternation data is from Gong's "Phonological Alternations in Tangut" which is based on a modified version of Sofronov's reconstruction. Since Sofronov reconstructed rhotic rhymes as tense rhymes, Gong conflated the two types of alternations. Do any semantic patterns emerge if the two types are separated? v. = verb with unknown transitivity.

Lax ~ tense (< *S-) alternations:

Gong's example numbers Gloss for lax vowel base Gloss for tense vowel derivative Frequency
105, 111 v.i. v.t. 2
124, 125, 133 adj. v.t. (causative) 3
119, 137 v.t. v.t. (causative) 2
112 v.t. n./v.t. 1
134 v.i./n. n. 1
109 n. v.i. 1
116 n. v.t. 1
117 n. v. 1
107, 108, 113, 118, 126 difference not in terms of transitivity or part of speech 5
106 v.t.: no known difference 1
114 v.i,: no known difference 1
115 n./v.t.: no known difference 1
135 n.: no known difference 1

Lax ~ retroflex (< *r-) alternations:

Gong's example numbers Gloss for lax vowel base Gloss for retroflex vowel derivative Frequency
110 n . v. 1
136 v.i. n. 1
123 v.t. adj./n. 1
132 n./v.i. n. 1
120, 121, 122, 127, 129 difference not in terms of transitivity or part of speech 5
128 v.t.: no known difference 1
130, 131 n.: no known difference 2

The functions of *S- and *r- seem to overlap, with one exception: only *S- can form transitive or causative verbs. This causative *S- is very old and can be traced to Proto-Sino-Tibetan since it is also found in Old Chinese and Tibetan. I could reconstruct it in pre-Tangut as *s- to match the sibilants in those two languages.

I write pre-Tangut *S- with a capital letter as a cover symbol for two or more acute prefixes: e.g., *t- as well as *s-. One or more of the non-causative functions associated with tense vowels could have reflected an acute stop prefix like *t- (5.4.1:20: or a different sibilant prefix like *ɕ-).

Perhaps pre-Tangut *r- also has multiple sources: e.g., a lenited *t- before certain consonants (and/or after a presyllabic vowel?):

*(CV-)t-CV > *(CV-)d-CV > *l-CV > *r-CV > CVʳ

The overlap between functions of tense and retroflex-vowelled derivatives may reflect a common origin (*t-) in some cases. The following table lists potential sources of tension and retroflexion:

Function Early pre-Tangut Late pre-Tangut Tangut
Causative *s1-CV C̣Ṿ CṾ
Noncausative *s2-CV
*(CV-)t-CV *rCVʳ CVʳ
No function; *r is part of root *(CV-)rV *rVʳ
*(CV-)CrV *CVʳ

5.4.00:33: *s1- and *s2- may have had different vowels at an earlier stage. Any pre-Tangut monoconsonantal prefix with multiple functions could have resulted from the merger of very early nonhomophonous prefixes with different vowels (and consonants?): e.g.,

*sa-, *si-, *su-, *ɕa-, *ɕi-, *ɕu- (each with a different meaning) > *s-

I am assuming that different meanings were once correlated with different phonetic forms, though it is possible for two homophonous words with unrelated meanings to be grammaticalized as homophonous affixes: e.g..,

*sa1 > *s1-

*sa2 > *s2-

5.4.1:12: It's not clear if there were any *s-causatives and/or transitives with presyllables in pre-Tangut: i.e., *CV-s-CV. Perhaps such forms developed aspirated initials in Tangut: e.g., Gong's alternation examples #12 and 34 (which I have united into a single set):

TT2366 kwɨə R31 1.30 'to separate' (transitivity uncertain*) < *Cɯ-kwə

TT2364 gwɨə R31 1.30 'to separate' (v.i.) < *Nɯ-kwə

TT4795 khwɨə R31 1.30 'to separate' (v.t.) < *Cɯ-s-kwə

Initial *s-k- would have become k- followed by a tense vowel

*s-kwə > *ḳwə̣ > kwə̣

whereas medial *-s-k- became kh- and the following vowel remained lax.

*Gong glossed both TT2366 kwɨə and TT4795 khwɨə as 'cut off, separate' (alternation example #34). This implies that they are both transitive verbs. If so, then perhaps TT2366 is from *sɯ-kwə which lost its transitive prefix after its high vowel caused the root vowel to bend upwards:


In "No Chɔ̣", I proposed that pre-Tangut had no aspirated s- or r-obstruent clusters at one stage. This constraint against *SCh- and *rCh- was inspired by Middle Korean and Written Tibetan:

Permissible s-obstruent initial clusters in Middle Korean
sp- st- sk-
pst- psk-

(Although Unicode supports han'gUl letter combinations for aspirated s-obstruent initial clusters such as ᄺ sph-, I have never seen these clusters in Middle Korean and suspect that they may be theoretical.)

Permissible s-obstruent initial clusters in Written Tibetan (ignoring medials)
sp- st- sts- sk-
bst- bsts- bsk-
sb- sd- sg-
bsd- bsg-

(Did *sdz- and *bsdz- become other initials?)

Permissible r-obstruent initial clusters in Written Tibetan (ignoring medials)
rt- rts- rk-
brt- brts- brk-
rb- rd- rdz- rj- rg-
brd- brdz- brj- brg-

(Did *rp-, *rc-, and *brc- become other initials?)

It's not impossible for languages to have clusters like sCh- and rCh-: e.g.,

Japhug rGyalrong (Jacques 2004)
sp- st- sc- sk- sq-
sph- sth- sch- skh- sqh-
ɕp- ɕt- ɕtʂ- ɕk- ɕq-
ɕph- ɕth- ɕkh- ɕqh-
rp- rt- rts- rtɕ- rc- rk- rq-
rph- rtsh- rtɕh- rch- rkh- rqh-

Sanskrit also has sph- sth- skh- in addition to sp- st- sk-.

Thus there is typological support for an earlier stage of pre-Tangut which had *s- and *r-obstruent clusters with and without aspiration.

Typology, however, is no substitute for a Tangut-internal test of my hypotheses. Do the morphological alternations listed by Gong in "Phonological Alternations in Tangut" conform to the frequency predictions that my hypotheses entail? Numbers represent tones: 1 = level, 2 = rising. *K- represents a consonant conditioning aspiration (most likely *k- but perhaps also *x-). *S- represents consonants conditioning tense vowels (most likely *s- but perhaps other acutes as well).

Lax ~ tense alternations:

Word family types Possible pre-Tangut alternation Number of affixes Expected frequency (inversely correlated with number of affixes: 1 = high, 2 = medium, 3 = low, 4+=improbable) Numbers of Gong's examples Actual frequency
CV1 ~ CṾ1 *CV ~ *S-CV 0/1 high 105, 107, 111, 114, 117, 126, 133 7
CV1 ~ ChṾ1 *CV ~ *K-S-CV 0/2 medium 0
ChV1 ~ CṾ1 *K-CV ~ *S-CV or
*(K-)ChV ~ *S-ChV
1/1 or (1)/1 medium 112, 116, 118, 135 4
ChV1 ~ ChṾ1 *K-CV ~ *K-S-CV or
*(K)-ChV ~ *K-S-ChV or
*ChV ~ *S-ChV with aspirate restored by analogy
1/2 or (1)/2 low 125 1
CV1 ~ CṾ2 *CV ~ *S-CV-H 0/2 medium 106, 113, 124 3
CV1 ~ ChṾ2 *CV ~ *K-S-CV-H 0/3 low 0
ChV1 ~ CṾ2 *K-CV ~ *S-CV-H 1/2 low 0
ChV1 ~ ChṾ2 *K-CV ~ *K-S-CV-H or
*(K)-ChV ~ *K-S-ChV-H
1/3 or (1)/3 improbable 0
CV2 ~ CṾ1 *CV-H ~ *S-CV 1/1 medium 119, 137 2
CV2 ~ ChṾ1 *CV-H ~ *K-S-CV 1/2 low 0
ChV2 ~ CṾ1 *K-CV-H ~ *S-CV or
*ChV-H ~ *S-ChV
2/1 or 1/1 low or medium 115 1
ChV2 ~ ChṾ1 *K-CV-H ~ *K-S-CV or
*(K)-ChV-H ~ *K-S-ChV
2/2 or 1~2/2 improbable 0
CV2 ~ CṾ2 *CVH ~ *S-CVH 0/1 high 108, 109, 134 3
CV2 ~ ChṾ2 *CVH ~ *K-S-ChVH 0/2 medium 0
ChV2 ~ CṾ2 *(K-)ChVH ~ *S-ChVH or
(1)/1 or 1/1 high if aspiration was not preserved before *-H; otherwise medium 0
ChV2 ~ ChṾ2 *(K-)ChVH ~ *(K-)S-ChVH or
(1)/1~2 or 1/2 possibly high if aspiration was preserved before *-H; otherwise medium (if root is aspirated) or low (if root is unaspirated) 0
Total 21

Lax ~ retroflex alternations:

Word family types Possible pre-Tangut alternation Number of affixes Expected frequency (inversely correlated with number of affixes: 1 = high, 2 = medium, 3 = low, 4+=improbable) Numbers of Gong's examples Actual frequency
CV1 ~ CVʳ1 *CV ~ *r-CV 0/1 high 123, 129, 131 3
CV1 ~ ChVʳ1 *CV ~ *K-r-CV 0/2 medium 0
ChV1 ~ CVʳ1 *K-CV ~ *r-CV or
*(K-)ChV ~ *r-ChV
1/1 or (1)/1 medium 0
ChV1 ~ ChVʳ1 *K-CV ~ *K-r-CV or
*(K)-ChV ~ *K-r-ChV
1/2 or (1)/2 low 0
CV1 ~ CVʳ2 *CV ~ *r-CV-H 0/2 medium 120 1
CV1 ~ ChVʳ2 *CV ~ *K-r-CV-H 0/3 low 136 1
ChV1 ~ CVʳ2 *K-CV ~ *r-CV-H 1/2 low 0
ChV1 ~ ChVʳ2 *K-CV ~ *K-r-CV-H or
*(K)-ChV ~ *K-r-ChV-H
1/3 or (1)/3 improbable 0
CV2 ~ CVʳ1 *CV-H ~ *r-CV 1/1 medium 121 1
CV2 ~ ChVʳ1 *CV-H ~ *K-r-CV 1/2 low 0
ChV2 ~ CVʳ1 *K-CV-H ~ *r-CV or
*ChV-H ~ *r-ChV
2/1 or 1/1 low or medium 128, 130 2
ChV2 ~ ChVʳ1 *K-CV-H ~ *K-r-CV or
*(K)-ChV-H ~ *K-r-ChV
2/2 or 1~2/2 improbable 0
CV2 ~ CVʳ2 *CVH ~ *r-CVH 0/1 high 122, 127, 132 3
CV2 ~ ChVʳ2 *CVH ~ *K-r-ChVH 0/2 medium 0
ChV2 ~ CVʳ2 *(K-)ChVH ~ *r-ChVH or
*K-CVH ~ *r-CVH
(1)/1 or 1/1 high if aspiration was not preserved before *-H; otherwise medium 0
ChV2 ~ ChVʳ2 *(K-)ChVH ~ *(K-)r-ChVH or
*K-CVH ~ *K-r-CVH
(1)/1~2 or 1/2 possibly high if aspiration was preserved before *-H; otherwise medium (if root is aspirated) or low (if root is unaspirated) 0
Total (including #110 below) 12

There is one case involving lenition that doesn't fit neatly into the above scheme since the initials do not simply differ in terms of aspiration:

Gong #110: ɣu R3 1.3 'smoke' ~ kuʳ R81 1.76 'to smoke'

< *Cɯ-ku ~ *r-ku

There are more cases of ChV ~ CṾ/Vʳ (7) than CV ~ ChṾ/Vʳ (1), which is what I would expect if there was deaspiration after *S- and *r-prefixation. I would rather not try to account for the seven ChV ~ CṾ/Vʳ alternations with a *K-prefix before each lax-vowelled form.

The absence of alternations like ChV2 ~ ChṾ/Vʳ2 indicates that I did not need to claim that aspirates were preserved before breathy vowels which later bore the rising tone. If there had been a large number of such alternations and if there was no aspirate preservation, I would have to claim that the aspirates in the nonlax derivatives were either restored by analogy with the base forms or reflected *K-.

Analogical restoration may account for the one case of ChV ~ ChṾ:

Gong #125: khwa R17 1.17 'distant' ~ khwạ R66 1.63 'keep at a distance' (not the expected kwạ R66 1.63)

< *khwa ~ *s-khwa

(*k-s-khwa, which would regularly develop into khwạ, is far less likely)

Tense-vowelled aspirate-initial syllables without lax-vowelled cognates probably have secondary aspiration from *K-: e.g.,

tshị R70 1.67 'way; road' < *K-S-tsi

(but there are no alternations like tsV ~ tshṾ, so maybe *S-tsh- became tsh- and kept its aspiration?)

I would rather not reconstruct a triply prefixed *P-K-S-ku for the four instances of khwụ R61 2.51

TT5105 'bleach'

TT2569 'cutting saw'

TT1673 'slice'

(no TT; Li Fanwen 1997 #5197) 'slice'

(the last three must be cognate)

so I suspect that khw- developed from *P-S-k-:

*P-S-ku > *pxku > *phku > *pkhu > *pkhwu > *ḳhwu > *ḳhwụ > khwụ

This is not quite like Middle Korean psk- which became modern Korean tense ㄲ ḳ-.

Retroflex-vowelled aspirate-initial syllables without lax-vowelled cognates could have

a. secondary aspiration from *K-

b. primary aspiration and root-medial or final *r


khaʳ R85 1.80 'dry' < *khar

cf. Old Chinese 乾 *kar 'dry', Chepang garʔ 'warm oneself in the sun'

(Schuessler [2007: 249] reconstructed Proto-Sino-Tibetan *kar with a final *-r presumably on the basis of Chepang garʔ, the only form he cited with a rhotic coda.)

(Pre-Tangut *khra and *K-r-ka are also possible, but the latter is particularly unlikely.) NO CHƆ̣ OR CHUỌ AT ALL?

In "Mysterious Medicine", I noticed that rhyme 75 could not follow any aspirates. Aspirates are common in Tangut, so I doubted that this gap was accidental. Using Sven Osterkamp's regular expression search tool, I discovered that aspirates were nonexistent or rare before tense and rhotic rhymes in comparison to lax-vowelled rhymes (e.g., R1-R4 and R8-11):

Rhyme cycle Gong's rhyme group Rhyme number Gong My reconstruction ph- th- tsh- tɕh- kh-
Lax I 1 -u -əu 7 12 4 8
2 -ju -ɨu 4
3 -u 4 8 1 7
4 -u 3
II 8 -e -əi 4 2 1
9 -ie 3 1 1 5
10 -ji -ɨi 4
11 -i 10 3 9 9
Total for only 8 out of 60 lax rhymes 110
(rhyme groups II-VIII are out of order)
I 61 -ụ -əụ 3
62 -jụ -ụ 1
VII 63 -iẹj -ɛ̣ 1 1
64 -jịj -iẹ 1 4
VIII 76 (out of order!) -iə̣j -ɛ̣̃
65 -jɨ̣j -iẹ̃ 2
V 66 -ạ -ạ 1 1 2
67 -jạ -ɨạ 2
II 68 -ẹ -əị
69 -iẹ -ɪ̣ 1
70 -jị -ị 1 1 1
VI 71 -ə̣ -ə̣ 2 2
72 -jɨ̣ -ɨə̣̣̣ 2 2 1
X 73 -ọ -ọ 1 4 2
̃74 -iọ -ɔ̣
75 -jọ -uọ
Total for all tense rhymes 39
(short vowel rhymes 80-87, 90-98 first, with long vowel rhymes 88-89 in between, followed by the remaining long vowel rhymes 99-103)
VII (out of order!) ̃̉77 -ejʳ -eʳ
78 -iejʳ -ɛʳ
79 -jijʳ -ieʳ
I 80 -uʳ -əuʳ
81 -juʳ -uʳ 1
II 82 -eʳ -əiʳ
83 -ieʳ -ɪʳ
84 -jiʳ -iʳ
99 -eeʳ -əiiʳ
101 (out of order!) -jiiʳ -iiʳ
IV 85 -aʳ -aʳ 3
86 -iaʳ -æʳ 2
87 -jaʳ -ɨaʳ 4 1 4
88 (out of order!) -aaʳ -aaʳ 3
89 (out of order!) -jaaʳ -ɨaaʳ 1
VI 90 -əʳ -əʳ 2 2
91 -iəʳ -ʌʳ 1
92 -jɨʳ -ɨəʳ 1 4
100 -jɨɨʳ -ɨəəʳ 1
IX 93 -eʳw -eʳw
94 -jiʳw -iʳw
X 95 -oʳ -oʳ
96 -ioʳ/-joʳ -ɔʳ/-uoʳ 1
102 -ooʳ -ooʳ
103 -jooʳ -uooʳ 1
XI 97 -oʳw -õʳ
98 -joʳw -uõʳ 1
Total for all retroflex rhymes 33

Why are aspirates less frequent before tense and retroflex rhymes? Perhaps aspirate root initials deaspirated before acute and rhotic prefixes: e.g.,

*S-ChV > *SCV > *C̣V > *C̣Ṿ > CṾ

*r-ChV > *rCV > *rCVʳ > CVʳ

However, aspirate root initials may have remained intact before retroflex rhymes that developed from rhotic medials and codas:

*ChrV > *ChrVʳ > ChVʳ

*ChVr > *ChVʳr > *ChVʳ

I doubt there is a single explanation for all 39 cases of aspirates before tense vowels.

Some aspirates may have developed from a velar or glottal prefix added after the reduction of *S-clusters:
(*K-S-C(h)V > *K-SCV > *K-C̣V > *K-C̣Ṿ >) *K-CṾ̣ > *hCṾ > ChṾ

TT0971 BIG thạ R66 2.56 is probably a loan from Tangut period northwestern Chinese 大 *tha 'big'. A Tangut prefix could account for the tense vowel, and the aspiration might have been restored by Tangut speakers who knew the Chinese word:

*tha > *s-tha > *sta > *ṭa > *ṭạ > *tạ > thạ

Could a few other aspirate-tense syllables be borrowings with restored aspiration? Such a convoluted scenario probably wouldn't occur more than once.

4.30.1:45: Could breathiness have blocked root-initial deaspiration?

*S-ChV > CṾ (level tone)


*S-ChVH > *ChṾʰ > ChṾ (rising tone)

Perhaps level tone aspirate-tense syllables only have secondary aspiration from *K- whereas rising tone aspirate-tense syllables could either have primary (i.e., root-initial) or secondary aspiration:

*K-S-ChV > ChṾ (level tone)

*(K-)S-ChVH > ChṾ (rising tone)

Although rising tone aspirate-tense syllables may have had an extra source, they are still outnumbered by level tone syllables

Level tone ChṾ: 22

Rising tone ChṾ: 16

Unknown tone ChṾ: 1

But this is not surprising, since the level tone is more common. MYSTERIOUS MEDICINE: MORE ON RHYME 76

I'm grateful that Mahaadaatṛ sent me a copy of Kychanov and Sofronov 1963, which contains explanations for the reconstruction of each Tangut rhyme, much like Nishida 1964 and Li Fanwen 1986. (I'd love to see a rhyme-by-rhyme explanation from Gong, who's outlined the principles of his reconstruction but has not published the detailed reasoning behind it.)

I was baffled by K&S' explanation for R76 for three reasons:

1. They said the phonetic tables indicated no medial, yet they reconstructed R76 with medial -j-.

(Have the tables been reprinted anywhere? I've never seen more than descriptions and bits of them. I have no way to know if my reconstruction fits the tables or not.)

2. They stated that R76 (their -jo) was in complementary distribution with the rising tone part of R75 which they also reconstructed as -jo. This is true, but I wonder if it is accidental, since the initials of R76 (p- and ʔ-) do not comprise an obvious phonetic class distinct from the initials of R75 rising tone syllables:

Homophones chapter I II III IV V VI VII VIII IX
voiceless obstruents k s
no aspirates!
voiced obstruents d z
sonorants m w n l(h)

If X occurs only after A and Y occurs only after B, that doesn't necessarily mean that X and Y are phonemically identical.

Moreover, if the final of R76 was a special variety of R75 that only occurred after p- and ʔ-, why wasn't TT5503 BAKE R75 1.72 with initial p- classified under R76? (There are no R75 syllables with the Homophones chapter VIII initial ʔ- and either tone.)

3. Until now, I thought the only Chinese transcription for an R76 syllable was the Grade II sinograph 擺 ?*pɛ, but they only mention a totally different transcriptive sinograph: 藥 'medicine', which may have been pronounced *jo during the Tangut period. Its pre-Tangut Tibetan transcription was yag and in modern northwestern dialects, it is yu (Xining), yo (Xi'an, Pingliang, and older Lanzhou), ye (newer Lanzhou), and (Dunhuang). 藥 ?*jo 'medicine' is the basis for their reconstruction -jo. Obviously, doesn't sound like -jo; both have mid vowels and palatality (in the vowel or in the glide), but that's all they have in common.

I have not been able to confirm 'medicine' as a transcription of TT4031 GULLY, the only R76 syllable with initial ʔ-. ('Medicine' could not possibly represent any of the four R76 syllables with initial p-.) However, I have found a syllable transcribed with 'medicine' whose tangraph resembles TT4031 GULLY without EARTH on the left:

TT4197 DITCH/GULLY R3 1.3 (Gong/this site: ɣju)

TT4031 GULLY R76 2.65 (Gong: ʔiə̣j; this site: ʔɛ̣̃)

Did K&S misread TT4197 as TT4031? WAS RHYME 76 ROUNDED?

I briefly considered reconstructing Tangut rhyme 76 as a back tense nasal vowel to balance the front tense nasal vowel iẹ̃ of rhyme 65.

Currently I have mechanically converted Gong's Grade II R76 -iə̣j into -ɛ̃.̣ (In my reconstruction, Grade II vowels are low to lower mid. Gong assigned R76 to his rhyme group VIII which I reconstruct with nasal vowels. Rhyme group VIII has five members: three in the first cycle of lax-vowelled rhymes [R41-43] and two in the second cycle of tense-vowelled rhymes [R65 and R76].)

R76 is strange because it is a front vowel between an oral back diphthong (R75 -uọ) and an oral front vowel (R77 -er). So I wondered if R76 was to R73-75 what R56-58 were to R54-55:

Oral vowel rhymes Nasal vowel rhymes
Lax vowel rhymes R54 -oo, R55 -ɔɔ, -uo [uɔ]? R56 -õ, R57 -ɔ̃, R58 -uõ
Tense vowel rhymes R73 -ọ, R74 -ɔ̣, R75 -uọ R76 ?-ɔ̣̃ (instead of -ɛ̣̃)

Although reconstructing a rounded value for R76 would account for its placement after the rounded vowel rhyme 75, it conflicts with the only transcription that I know of: the Grade II sinograph 擺. The rhyme category of 擺 is in modern northwestern dialects (e.g., 擺 is in Xi'an) and this Chinese rhyme was transcribed by Tibetans as -e, -eHe, -eHi, -yi after labials in the pre-Tangut period. Thus R76 was presumably something like -ɛ.

An e-type value is not surprising because R76 is followed by a trio of retroflex* e-type rhymes:

Rhyme number 76 77 78 79
Tibetan transcriptions n/a -e, r-e n/a -e, r-e
Kychanov and Sofronov 1963 -jo -ɜ̄ -jɜ
Nishida 1964 -jẹ -ẹ -jʊr -jẹ/-ʊr**
Hashimoto 1965 -jəwN -jə -jə/-ewN***
Sofronov 1968 -ại -ại -ại -jẹ
Huang 1983 n/a -oĩ n/a -ï̃
Li Fanwen 1986 -oi -ui/-ẽ -jụe/-jue/-jẹi/-jei**** -jẹ
Gong 1997 -iə̣j -erj -ierj -jirj
Arakawa 1999 -ẹ -jẹ -ẹʔ -jẹʔ
This site -ɛ̣̃ -er -ɛr -ier

(Thanks to a reader - whom I'd like to call Mahaadaatṛ - for sending me a copy of Kychanov and Sofronov ̣1963, which I haven't seen in about a decade!)

I've been trying to figure out ways to explain why R76 -ɛ̣̃ was between R75 -uọ and R77 -er. The solutions fall into two categories:

- A rhotic solution

There are only four R76 syllables (pɛ̃ x 3, ʔɛ̃ x 1) which would have originated from pre-Tangut *SʌCVN syllables. Last night, I wondered if these syllables were pronounced with *-n or *-r in different pre-Tangut dialects. (Similar liquid/nasal variation occurred in Old Chinese.) In *-n dialects, the rhyme would have become *-ɛ̣̃, but in *-r dialects, it would have lost its tension and become r (if *S-r type syllables became -Vr rather than -Ṿ syllables in Tangut). The unique placement of R76 may indicate that it was tense (*-ɛ̣̃) like R75 in some dialects and rhotic (-ɛr) like R77 in others.

- o/e solutions

R76 had both o- and e-like qualities which uniquely suited it for placement between the last tense o-rhyme and the first retroflex e-rhyme. Li Fanwen's -oi has an o followed by an i that is palatal like e. Perhaps this -oi was pronounced as (w)e or ø depending on dialects. (Middle Korean ㅚ oj is now pronounced as (w)e or ø in modern Korean.) However, in my adaptation of Gong's reconstruction, this rhyme would have to be tense, and I doubt that Tangut had a tense -ọj or -ø̣ that had no lax counterpart.

Kychanov and Sofronov's reconstruction is the mirror image of Li's with the o- and i-type elements in the opposite order. It's based on a surprising Chinese transcription which I'll discuss next time.

*The extant Tibetan transcriptions for R77 and R79 have a (pre)initial r- found in other retroflex rhymes:

Rhyme number Tibetan transcriptions
77 re, rwe, Hrgwe, Hhgwe (sic!)
79 me, rme

Initial r- cannot occur with nonretroflex rhymes except in the R43 syllable riẽ 2.37.

Pre-Tangut must have had many syllables like *r(C)VN with rhotic (pre)initials and nasal codas. Since Tangut presumably did not have any nasal retroflex rhymes, most of those syllables developed oral retroflex rhymes:

*rVN > rVr

*rCVN > CVr

But I don't know why *Cɯ-reN became riẽ R43 2.37 instead of rier R79 2.68.

4.28.00:04: There is no Tibetan or Chinese transcriptional evidence for R78, so the various reconstructions are guesses based on how rhymes pattern elsewhere. For example, Gong assumed that rhymes generally appear in sets of three or four with different medials:

1. no medial

2. medial -i-

3 (and sometimes 4). medial -j-

I have reinterpreted Gong's sets in terms of vowel height:

1. mid

2. low

3 (and sometimes 4). high

See "3 x 3: The Tangut Vowel System?" for details.

As the middle member of a trio, R78 has a medial -i- in Gong's reconstruction and a low(er mid) vowel ɛr in my reconstruction.

**Nishida regarded R76 2.65 and R79 1.74 as his rhyme 75 (-jẹ) and reconstructed R79 2.68 as his rhyme 77 (-ʊr).

***Hashimoto reconstructed R79 1.74 as -jə and R79 1.75 as -ewN because he regarded them as different rhymes (75 and 76 in his numbering system).

****Li Fanwen (1986) reconstructed this rhyme in four different ways:

-jue (p. 178, 419)

-jụe (p. 189)

-jei (p. 468)

-jẹi (p. 468)

One or more of these four may be slips of the pen (since Li 1986 was reproduced from handwriting).

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2008 Amritavision