I should have put this information into the previous post.

and dʒ- R100 1.92 have the initial speller

TT5075 dʒɨiʳ R84 2.72 'butter'

which is also in Mixed Categories (19B42).

The initial speller of 'butter' is the MC tangraph

TT4369 dʒɨụ R62 2.52 'bait'

which has an MC initial speller

TT2195 dʒɨị R70 1.67 'fete; feast; banquet'
cognate to TT2253 dʒ- R100 1.92 'id.'

whose initial speller is 'bait'

dʒ-, like dz-, is apparently exclusive to Mixed Categories. (I would appreciate it if anyone could find counterexamples.) Why?

The initial speller of the other Mixed Categories R100 syllable

TT3136 lh- R100 1.29 'lose'

is the MC tangraph

TT2453 lhɨʌ̣ R72 1.69 'bow'

whose initial speller is the MC tangraph

TT4995 lhiụ R62 1.59 'belt; girdle'

whose initial speller is the MC tangraph

TT4810 lhị R70 2.60 'below'

whose initial speller is unknown. lh- is another initial that is apparently exclusive to Mixed Categories. Why? lh- does not form a natural class with the other MC-only (?) initials dz- and dʒ- unless it too were a voiced affricate dɮ-. But the transcriptions of lh- generally lack d (Tai Chung Pui 2008: 201):

Nishida, Sofronov, Gong Tai Tibetan transcription
lh- ld- ld- (1)
lh- lh- (13)
b-lh- (1)
sl- (2)
zl- (2)

Suppose that lh- might have been a voiceless lateral fricative [ɬ] in Tangut period Tibetan. (Gong [2003: 605] reconstructed Tangut lh- as [ɬ] .) If Tangut period Tibetan had no symbol(s) for a voiced lateral fricative [ɮ] or affricate [dɮ], lh- could have been the best available choice for a Tangut [dɮ] in spite of its voicelessness. s- and z- could have been attempts to indicate the fricative component of [dɮ].

Tai reconstructed ld- as the initial of

TT2453 lhɨʌ̣ R72 1.69 'bow'

Tibetan transcription lda

which is also the initial speller of

TT4995 lhiụ R62 1.59 'belt; girdle'

its neighbor in Mixed Categories.

Tai's ld- may also correspond to l- in other reconstructions of non-Mixed Categories tangraph readings: e.g.,

TT0889 laa R22 1.22 'pollute'

Tibetan transcription gldang

I've been bothered by Tai's ld-cluster for months because it is sui generis in his system. Instead of reconstructing an anomalous lC-cluster, I could reconstruct two affricates, one voiced and one voiceless:

Liquid fanqie chain numbers Tibetan transcriptions Nishida Sofronov Gong Tai This site
3-6 ld- (15), zl- (4), b-ld- (2), c- (1), gz- (1) ɫ- ld- l- ld- dɮ-
7 ld- (1) lh- tɬ-

I don't know why chain 7 has been reconstructed with lh-. Perhaps it did have the same initial as chains 3-6. If I merge those chains, I could assign tɬ- to chains 8-10 and/or 18:

Liquid fanqie chain numbers Nishida Sofronov Gong Tai This site (proposal 1) This site (proposal 2)
1-2 l-
3-6 ɫ- ld- l- ld- dɮ- ɮ-
7 lh- (Gong 2003: [ɬ])
8-10 lh- (t)ɬ- ɬ-
11-12 r-
13, 16, 19 ʁ(z)- z̀- z- ʁz- ɮ- z-
14-15, 17 ňž- ź- ʁź- ʐ- ʐ-
18 l- l- lh-/ld- (t)ɬ-/dɮ- ɬ-/ɮ-
20 ʁz- ld- dɮ- ɮ-

Nishida and Tai reconstruct ʁ-clusters that I would rather avoid. I think it is strange for a language to have ʁz- but not z- (and in Tai's case, ʁz- and ʁź- but not ʁ-, z-, or ź-).

My proposal 1 consists purely of laterals and rhotics (if one counts the retroflex fricative ʐ- which could have sounded like a Mandarin r- as opposed to Tangut r- which may have been a tap).

My proposal 2 lacks lateral affricates and carries over z- from Gong's reconstruction even though it is neither lateral nor rhotic. This sound was transcribed in Tibetan as

gz- (35)

z- (9)

rdz- (2) (before the retroflex rhymes R84 -iʳ and R85 -aʳ)

H-z- (1)

without any l and with r only twice. (The r- probably indicates retroflexion in the rhyme rather than any rhotic quality in the initial.) Its Chinese transcription was the fanqie initial speller 移 with a diacritic indicating that its initial was not ientical to Chinese *j-. On the one hand, I might think this represented another palatal like ʑ-, but the Tangut initial was never transcribed as Tibetan zh- [ʑ]. On the other hand, if the Tangut initial really were z- (a sound probably absent in Tangut period NW Chinese), why wasn't it transcribed as Chinese *s- with a diacritic? Maybe the initial inventories of the dialects underlying the Tibetan and Chinese transcriptions were slightly different:

Liquid fanqie chain numbers Tangut dialect transcribed in Tibetan Tibetan transcription Tangut dialect transcribed in Chinese Chinese transcription
13, 16, 19 z- z- ʑ- *j-
14, 15, 17 ʐ- gzh- ʐ- *ʒ- R100 IN MIXED CATEGORIES OF THE TANGRAPHIC SEA

For some reason not all R100 1.92 tangraphs are in the level tone volume of Tangraphic Sea. I still don't know why certain tangraphs were placed in Mixed Categories which was organized by tones and the nine initial classes also found in Homophones:

Mixed Categories: level tone

I. Labials (missing; in Mixed Categories of the Precious Rhymes of the Tangraphic Sea)

II. Labiodentals (missing; in MCPRTS)

III. Dentals (also missing in MCPRTS)

IV. 'Retroflexes'

V. Velars

VI. Alveolars

VII. Alveopalatals

VIII. Glottals

IX. Liquids, 'tongue and tooth sounds'

Mixed Categories: rising tone

I-IV (missing; in MCPRTS)

V-IX (same as level tone)

Here are the R100 MCTS tangraphs and the tangraphs preceding and following them for context. I've used Sofronov's (1968 II: 55-59) MCTS fanqie index to find them. Let me know if I've missed any. (Sofronov missed the homophone group including 'burn' right before 'give up'.)

Level tone VII (MCTS07B31-B41):

Tangraph Tangut Telecode Reconstruction Rhyme Tone.rhyme Gloss Final speller Notes
0574 dʒɨʌ̣ R72 1.69 burn TT4421 ʃɨʌ̣ R72 1.69
3767 dʒ- R100 1.92 give up; abandon TT5524 tʃ- R100 2.85 (rising tone!)
2253 banquet; feast; fete cognate to TT4517 dzɨi R10 1.10 'eat'?; shift of initial class conditioned by prefix?
1436 dʒɨw R46 1.45 pursue TT2188 tʃhɨw R46 1.45

Level tone IX (MCTS11B11-B21):

Tangraph Tangut Telecode Reconstruction Rhyme Tone.rhyme Gloss Final speller Notes
4761 lhwɨɛ R36 1.35 oblique; tilted; slanting TT4619 ʃwɨɛ R36 1.35 cognate to Old Chinese 邪 *slja 'oblique'
3136 lh- R100 1.92 lose TT1533 tʃ- R100 1.92 (1.92 homophone group 2) cognate to Old Chinese 失 *hlit 'lose'
2453 lhɨʌ̣ R72 1.69 bow TT4421 ʃɨʌ̣ R72 1.69 cf. Matisoff's (2003: 642) Proto-Tibeto-Burman *ləj 'bow'

The same tangraphs are in Precious Rhymes of the Tangraphic Sea in the same order (3.3.2209-2302, 3.4.2108-2110).

I can't find any R100 2.85 tangraphs in Sofronov's fanqie index or the Mixed Categories section of Precious Rhymes of the Tangraphic Sea. Is the absence of a final from Mixed Categories significant? HOMOPHONE GROUPS 7 AND 8 OF R100 1.92 IN TANGRAPHIC SEA

are reconstructed identically by Sofronov (1968) and Gong (in Li Fanwen 1997), even though they have different fanqie:vh

Group 7:

kiʳ R84 1.79 + lhw- R100 1.92 (homophone group 9)

Sofronov: kwjẹ

Gong: kjwɨɨʳ

Group 8:

kiʳ R84 1.79 + pw- R100 1.92 (homophone group 6)

Sofronov: kwjẹ

Gong: kjwɨɨʳ

and are nonhomophones according to Homophones. The fanqie have the same initial speller, so presumably the difference must be in the finals.

Group 7 is larger than group 8, so I would expect group 7 to have a less exotic phonetic value than group 8:

Group 7 (three tangraphs):

Group 7.1: TT3423 kw- R100 1.92 'second half of diʌʌ kw- (Xiongnu)'

with Li Fanwen radical 221 'hand' on the right (why?)

Group 7.2: TT3425 kw- R100 1.92 'second half of diʌʌ kw-, a kind of bird'

with Li Fanwen radical 196 'bird' on the right

These first two tangraphs share a left-hand phonetic element that is also in the left halves of the disyllabic word for 'line; ranks':

TT3424 khiʌʌ R33 2.29

with Li Fanwen radical 213 (meaning unknown) on the right

and a ə-rhyme implying that R100 group 7 also had a ə-rhyme

TT3422 xõ R56 2.47

the left side is not phonetic, but a carryover from the first graph;

the right side is Li Fanwen radical 373 'bird' (why?)

TT3423 and TT3425 also both happen to be second halves of homophonous disyllabic words written with first tangraphs sharing a phonetic element TT0653 diʌʌ R33 2.29 'with':

diʌʌ kw- 'Xiongnu'

diʌʌ kw- 'a kind of bird'

with 'person' on the left of both tangraphs!

I suspect that the bird name is an extended use of the Tangut name for the Xiongnu which doesn't sound anything like 匈奴 Late Old Chinese ?*xuoŋna or its descendants (e.g., Middle Chinese *xuoŋno or Tangut period NW Chn *xjonu). Could the name be borrowed from a Xiongnu autonym distinct from the presumably Xiongnu source of ?*xuoŋna? Or did the Tangut coin their own word for the Xiongnu?

The last member of group 7

Group 7.3 TT1726 kw- R100 1.92 'steal'

has already been discussed in this post.

Group 8 contains only one tangraph:

TT3170 kw- R100 1.92 'coarse; coarse cloth'

with Li Fanwen radical 267 'bad' on the left

and Li Fanwen radical 067 'jewel' < right of TT5138 noo R54 2.45 'brown' on the right

I still suspect that groups 7 and 8 belonged to different high-vowel grades, but I have no way of determining which group was Grade III and which group was Grade IV. It's less likely that they had the same grade with different vowels: i.e., one was kwɪɪʳ and the other was kwiəəʳ. WAS R100 REALLY TWO RHYMES?

In recent posts, I've tried to determine whether R100 was a Grade II i-rhyme or a Grade III/IV ə-rhyme. Could it have been both?

In "A Hypothesis of Three Grades and Vowel Length Distinction in Tangut" (1994), Gong proposed that R33 and R55 were each combinations of two rhymes in one. He later (1997, 2003) reconstructed R33 solely as -jəə but split R96 into two:

Rhyme number Gong Arakawa My first reconstruction My second reconstruction
R33 -iəə (II; 1994 only?) -jI' (II only) -ʌʌ (II) -ʌʌ (II)
-jəə (III) -ɨəə (III) -ɨʌʌ (III)
-iəə (IV) -iʌʌ (IV)
R55 -ioo (II) -jo' (II only) -ɔɔ (II) -ɔɔ (II)
-joo (III) -ɨoo (III) -uɔɔ (III)
-ioo (IV) -yɔɔ (IV)
R96 -ioʳ (II) -joʳ (II only) -ɔɔʳ (II) -ɔɔʳ (II)
-joʳ (III) -ɨooʳ (III) -uɔɔʳ (III)
-iooʳ (IV) -yɔɔʳ (IV)

If Gong's rhyme-splitting proposal is correct, my first reconstruction must be wrong because it is unlikely that the Tangut would regard rhymes with vowels of different heights as members of the same rhyme category. However, I have doubts about all of the above proposed phonetic values because none of the Tibetan transcriptions of these rhymes have a -y- corresponding to the palatal glides or vowels in any of the first three reconstructions. My second reconstruction allows intergrade rhyming and avoids -i- before labial vowels. I will elaborate on this second reconstruction in ???.

Gong presents three types of evidence for splitting rhymes:

1. Alternations with distinct Grade II and Grade III (= my Grade III/IV) rhymes

2. Separate treatment of subtypes in rhyme tables (He's seen much more of them than I have! Have they ever been published in their entirety?)

3. Ordering in Tangraphic Sea: Grade II syllables precede Grade III syllables

I don't think alternations are sufficient evidence by themselves since it's possible that Grades II and III were merged in R33, R55, and R96: e.g. (in Gong's 1994 reconstruction),

ee-forms oo-forms (-J- = neutralization of -i- and -j-)
giee R13 'foolish' (Grade II) gJoo R55 'foolish' < *gioo
gjee R14 'bite' (Grade III) gJoo R55 'bite' < *gjoo

instead of

ee-forms oo-forms
giee R13 'foolish' (Grade II) gioo R55a 'foolish' (Grade II)
gjee R14 'bite' (Grade III) gjoo R55b 'bite' (Grade III)

But the other types of evidence indicate that no merger of grades had taken place. How could the compilers of the tables and Tangraphic Sea separate the grades if they had merged? Why would the R55 words for 'foolish' and 'bite' be listed as nonhomophones in Tangraphic Sea with distinct fanqie if they were both Grade II gioo or Grade III gjoo?

All of Gong's splits are within rhyme groups:

R33a/b - Grade II and III within group VI (long subtype)

R55a/b - Grade II and III within group X (long subtype)

R96a/b - Grade II and III within group X (long retroflex subtype)

However, I doubt that a rhyme category would be split across rhyme groups:

R100a - Grade II within group II (long retroflex subtype): -ɪɪʳ?

R100b - Grade III/IV within group VI (long retroflex subtype): -ɨ/iʌʌʳ?

Still, it would be interesting if Tangraphic Sea implied two subtypes of R100. Here is a list of all the fanqie for the homophone groups in TS' R100 1.92 section (1.91B51-92A62). The function of the circles dividing some groups is unknown.

1. kiuʳ R81 1.76 + l- R100 1.92 (homophone group 4)


2. tʃiʳ R84 1.79 + l- R100 1.92 (homophone group 4)

3. tsiʳ R84 1.79 + l- R100 1.92 (homophone group 4)


4. lị R70 2.60 + k- R100 1.92 (homophone group 1)


5. zị R70 1.67 + tsiʌ R31 1.30 (not an R100 final speller!)

6. pị R70 1.67 + kw- R100 1.92 (homophone group 8)


7. kiʳ R84 1.79 + lhw- R100 1.92 (homophone group 9)


8. kiʳ R84 1.79 + pw- R100 1.92 (homophone group 6)

9. ɣiu R3 1.3 (error for lhiạ R67 1.64?) + kw- R100 1.92 (homophone group 7)

Groups 1-4 form a set, as their final spellers spell each other.

Alveopalatals and alveolars are usually in different grades, yet groups 2 and 3 with these initials are placed between circles.

The nonretroflex final speller of group 5 tells us that at least one variety of R100 ended in a retroflex version of Grade IV R30 -iʌ. Sofronov reconstructs this group with a palatal fricative initial, but that is improbable since Grade IV is associated with alveolar rather than (alveo)palatal obstruents.

Groups 6 and 8 form a set. Gong reconstructed both groups 7 and 8 as kjwɨɨʳ, even though they have distinct fanqie and group 7 forms a set with group 9. Groups 7 and 8 are also distinct in Homophones. Since both groups 7 and 8 have initial kw-, the difference between them must be in their finals. I would like groups 6/8 and 7/9 to belong to different grades. However, the initials p- (group 6) and lh- (group 9) are both associated with Grade IV, implying that the kw-syllables in groups 7 and 8 are also Grade IV.


Guillaume Jacques (2006) found that R100 corresponds to gDong-brgyad rGyalrong -e, -ɯt, and -ɯ.

DG -e in kɯ-βde 'four' (cf. TT3509 lɪɪʳ or liəəʳ? R100 1.92) is from proto-rGyalrong *-ej and is cognate to Written Tibetan bzhi < *bli, Written Burmese leh, and Old Chinese *s-hli-s.

DG -ɯt in tɯ-ɕkrɯt 'bile' (cf. TT3611 kɪɪʳ or kiəəʳ? R100 2.85 'gall bladder') is from proto-rGyalrong *-it and is cognate to Written Tibetan mkhris-pa and Written Burmese saɲɲh-khre.

These correspondences suggest that R100 had a palatal vowel.

DG kɤ-mɯ-rkɯ 'steal' is cognate to

TT0311 kiiʳ R101 2.86 'steal'


TT1726 kwɪɪʳ or kwiəəʳ? R100 1.92 'steal'

written with a 'mirror-image' tangraph

If R100 were the Grade II counterpart of Grade III/IV R101, I could reconstruct these words in pre-Tangut as

*rkii-H > kiiʳ R101 2.86 'rob, steal' (Grade IV)

*X-P-rkii > kwɪɪʳ R100 1.92 'steal, pilfer' (Grade II)

(*X = the unknown conditioning factor for Grade II; it may not have been in antepreinitial position)

But I don't know of any other cases of a word family containing both Grade II and non-Grade II members. This suggests that the conditioning factor for Grade II may have been a root segment (e.g., an emphatic consonant?) absent in non-Grade II roots.

If R100 were a Grade III/IV rhyme, I could posit an -iiʳ :-iəəʳ alternation parallel to

R10 -ɨi : R30 -ɨə (both Grade III)

R11 -i : R31 -iə (both Grade IV)

(but are there any cases of R84 -iʳ : R92 -iəʳ or R70 -ị : R72 -iə̣?)

The pre-Tangut root would still be *rkii in this scenario.

There is one huge problem. One might think that DG is from proto-rGyalrong *-i, but in fact it is from proto-rGyalrong *-u as well as *-ɯ (Jacques 2004: 239). I presume that the pG word for 'steal' was *rku since Written Tibetan rku and Old Chinese 寇 *khos both have labial vowels. (Written Burmese kuih may have been *[khɯ] but ui is ultimately of labial origin.) There is also a Tangut cognate with a labial vowel (Jacques 2004: 235):

TT1728 kiuʳ R81 2.70 'steal'

Although it is tempting to derive this from pre-Tangut *rki-w, most cognates have rounded root vowels. Are the R100 and R101 words for 'steal' derived from a root *rku via ablaut or umlaut?

*rku-H > *rkɨu-H > kiuʳ R81 2.70 (with breaking of *u in Grade IV; cf. the breaking of Old Chinese nonemphatic *u to Late Middle Chinese *-iw)

*rkuu-J-H > *rkyy-H > kiiʳ R101 2.86 'steal' (*J = unknown conditioning factor for umlaut)

*P-rkuu-J-ə > *P-rkyyə > kwiəəʳ R100 1.92 'steal'

Next: The split rhyme solution.

10.30.1:59: I forgot to mention that the rGyalrongic language Zbu has kɐ-mə-rkəʔ and nɐ-mə-rkʰiʔ for 'steal'. Schwa is an expected reflex of proto-rGyalrong *u, but i is not. I doubt this i has anything to do with the front vowels of the R100 and R101 words for 'steal'. THE CHINESE TRANSCRIPTIONS OF GRADE II -I AND GRADE III/IV -K RHYMES

One might expect Tangut grades to correspond precisely to Chinese grades, but this is not the case. (Since the reconstruction of Tangut period northwestern Chinese is uncertain, I indicate Chinese rhyme classes in terms of my Middle Chinese reconstructions with two exceptions: medial *-w- has been ignored and MC *e has been rewritten as ie to reflect its diphthongization in late MC.)

Tangut grade Tangut rhyme My reconstruction Grades of Chinese transcriptions
I: nonhigh back vowels II: nonhigh front vowels III: high nonpalatal vowels IV: high palatal vowels
II or III/IV? R100 -ɪɪʳ or -ɨ/iəəʳ? -ək -ɨt
II R9 -æk, -ɛk
R13 -ɪɪ -ək (none!)
R69 -ɪ̣ -æk, -ɛk -iej
R83 -ɪʳ -æk
III R30 -ɨə -ək, -əŋ -ɨit, -ɨek, -ɨej, -ut -it, -ip, -iej
IV R31 -iə -əŋ, -ək, -ət -ɨ, -ɨet, -ut -i, -it, -in, -ip, -iek
III/IV R33 -ɨ/iəə -ək, -ət -æk -ɨin, -ɨet -it, -ip, -iek
R72 -ɨ/iə̣ -əŋ, -ək, -ət -ɨak, -ɨi, -ɨit
R92 -ɨ/iəʳ -ak, -ək, -ət -æŋ -ɨj, -ɨat, -ɨit, -ɨip, -ut -ie

The MC codas were presumably all lost by the Tangut period: e.g., MC -əŋ, -ək, -ət may have merged into *-ə in Tangut period northwestern Chinese.

There are no Chinese Grade II transcriptions of R100. The nonpalatal transcriptions of R100 are similar to those of the rhymes, suggesting that R100 should be -ɨ/iəəʳ, in spite of its placement between the -i rhymes R99 -əəiʳ and R101 -iiʳ. On the other hand, Grade II R13 was also transcribed with Grade I -ək, and ə and ɨ are nonlow, nonback, n to ɪ. Moreover, most of the transcribed Tangut syllables have initials (ts-, l-, ʒ- or z-) that usually* never occur in Chinese Grade II, so it was not possible to transcribe them with Chinese Grade II rhymes.

*There are some Grade II MC syllables with l-, but there are none with MC ts-, ʑ-, ʐ-, or z-. THE TIBETAN TRANSCRIPTIONS OF GRADE II -I AND GRADE III/IV RHYMES

Do the Tibetan transcriptions of R100 resemble those of Grade II rhymes or Grade III/IV -ɨ/iəə rhymes? The data from Tai Chung Pui (2008) indicate that R100 was generally transcribed with -i(H) but not u, e, and o which were more often used to write Grade III/IV -ɨ/iəə rhymes. This suggests that R100 was the Grade II counterpart of its Tangraphic Sea neighbors R99 -əiʳ (Grade I) and R101 -iiʳ (Grade III).

Grade Rhyme My reconstruction -a r- -a -ang -aH -ar -i r- -i -yi -iH r- -iH -ir -ii (sic) -u r- -u -uH -ue (sic) -e r- -e -ye -o r- -o
II or III/IV? R100 -ɪɪʳ or -ɨ/iəəʳ? 2 6 3
Percentages (out of 11) 18.2% 81.2% 0%
II R9 9 5
R13 -ɪɪ (1)
R69 -ɪ̣ 1 1 1
R83 -ɪʳ 1 4 1 1
rhyme totals 4 20 1 1 1
Percentages (out of 27) 14.8% 74.1% 3.7% 3.7% 3.7%
III R30 -ɨə 3 14 4
IV R31 -iə 4 1 88 1 1 9 2 9 3 1 1 1
III/IV R33 -ɨ/iəə 4 1 16 14 1
R72 -ɨ/iə̣ 6 2 1 17 11 2 1 1 1
R92 -ɨ/iəʳ 2 4 14 1 2
-ɨ/iə rhyme totals 24 198 13 6 2
Percentages (out of 243) 9.9% 81.5% 5.3% 2.5% 0.8%

r- indicates a preinitial in the transcription.

The only known transcription of R13 is in Nevsky (1926) and is in parentheses.

The near-total absence of -e from the transcriptions of R9/13/69/83 casts doubt on Gong's reconstructions with -ie(e) corresponding to my -ɪ(ɪ).

The total absence of -y- [j] from the transcriptions of R9/13/69/83 casts doubt on Arakawa's reconstruction of medial -j- corresponding to Gong's -i- and my zero medial.

Four unusual disyllabic transcriptions have been omitted from the above table:


thi shi (once) for thiəə 'this'; shi seems to represent its Chinese translation equivalent 是

se zi (once) for siəə 'die'; se or zi may represent its Chinese translation equivalent 死


lda thi (once) for lɨə̣ 'earth'; thi seems to represent its Chinese translation equivalent 地

me lda (once) for lɨə̣ 'treasure'; I don't know of any Chinese or Tibetan word for 'treasure' like me

The translation equivalents may have been alternate readings of tangraphs.

Next: The Chinese transcriptions of R100. THE INITIALS OF GRADE II -I RHYMES

If R100 really is a Grade II retroflex long -i-type rhyme, it should have initials similar to other Grade II -i-type rhymes. Although R100 lacks chapter III (dental) initials which are very rare before Grade II initials, it can combine with the chapter VI initial ts- which never occurs before Grade II initials:

Homophones chapter I II III IV V VI VII VIII IX
R9 p, ph, b w k, kh, g tʃ, tʃh, dʒ, ʃ ɣ lh, ʒ
R13 -ɪɪ b kh, g ʃ
R69 -ɪ̣ p, b, m k, kh ɣ l, lh
R84 -ɪʳ k ɣ ʒ
R100 ?-ɪɪʳ p k ts tʃ, dʒ ʔ l, lh, z or ʒ

All four ts- R100 syllables with the fanqie

tsiʳ + l- R100

cannot be regarded as exceptional foreign transcriptions:

TT3163 ts- R100 1.92 (first half of disyllabic word ts- gia 'stammer')

TT3176 ts- R100 1.92 'chrysalis'

TT3176 ts- R100 1.92 'five, fifth, fifth son' (not the normal word for 'five')

TT0807 ts- R100 1.92 'lychee'

TT0807 obviously has TT3176 as phonetic beneath Li Fanwen radical 102 'wood'. TT3176 is also phonetic in its fanqie's initial speller

TT0150 tsiʳ R84 1.79 'wife'*

The choice of TT3176 as phonetic in TT0150 might make more sense if TT3176 had an i-rhyme (-ɪɪʳ) rather than a ə-rhyme (-iəəʳ). Are there other instances of ə-phonetics in i-tangraphs?

So far, there are three arguments for reinterpreting R100 as a Grade II -i rhyme and one against:


1. The placement of R100 between Grade I R99 and Grade III R101

2. External cognates with -i: 'water' and 'four'

Against: The un-Grade II initial ts-

Next: Was R100 transcribed like other Grade II i-rhymes?

*The resemblance between tsiʳ and Middle Chinese 妻 *tshej 'wife' is probably coincidental since the initials and vowels do not match. The retroflexion in the Tangut word is not a fatal problem since it could reflect a Tangut prefix added to a borrowing. R91 = R100?

While looking up reconstructions for

TT4282 ʒɨəəʳ or ziəəʳ R100 2.85

'water', I noticed that Arakawa (1999 and 2001) reconstructed both R91 and R100 as -jIʳ. (I don't know whether the capital I represents [ɪ] or [ɨ].) He regarded R91 -jIʳ but not R100 -jIʳ as a member of his rhyme group XVII (equivalent to part of Gong's group VI):

R90 -Iʳ (Grade I)

R91 -jIʳ (Grade II)

R92 -IIʳ (Grade III)

He grouped R100 -jIʳ together with R101 -jeʳ2 (what does the 2 represent?) yet placed other -jI and -je rhymes into separate groups. I don't understand why a single rhyme would be simultaneously assigned to two different rhyme groups.

If R91 and R100 really were homophonous, I would expect them to have initials in complementary distribution. Yet they share three initials (in bold):

Homophones chapter I II III IV V VI VII VIII IX
R91 -jIʳ b k, ɣ , tʃh, , ʃ
R100 -jIʳ p k ts , ʔ l, lh, z or ʒ

(Is the absence of dentals before both rhymes simply the product of chance, or did dentals affricate before those rhymes?)

I doubt that the compilers of both Tangraphic Sea and Homophones would arbitrarily separate tangraphs such as

TT2565 k- + R91 1.85 'thick; saw' (latter a loan from Chn 鋸; both have 'metal' radicals)

TT0619 k- + R100 1.92 'worker'

if they were homophones. Therefore I would reconstruct them with distinct rhymes:

TT2565 kʌʳ R91 1.85 'thick; saw'

TT0619 kiəəʳ R100 1.92 'worker'

There is very little evidence for reconstructing R91:

R91 is between a Grade I and a Grade III rhyme in Arakawa's rhyme group XVII (retroflex ə-rhymes in my reconstruction), so it's probably Grade II and should therefore be reconstructed with a low retroflex achromatic vowel ʌ.

TT2565 kʌʳ R91 1.85 'saw' may be a loanword from Middle Chinese 鋸 *kɨə (also the source of Vietnamese cưa [kɨə] 'saw'). The retroflexion may be a trace of a native pre-Tangut *r-prefix.

R91 was used to transcribe the rhyme of 參 which ended in *-əm in Middle Chinese. Gong reconstructed its Tangut period rhyme as *-ã with a low vowel unlike the high vowel of Arakawa's R91 -jIʳ. This is the only instance of R91 in Chinese transcription, suggesting that it was unlike anything in Tangut period NW Chinese.

R91 was never used to transcribe Sanskrit.

No Tibetan transcriptions of R91 are known.

There is more evidence for R100:
R100 was transcribed in Tibetan as -i(H) and -a, corresponding to the -i/ɨ- and -əəʳ of my reconstruction. The Tibetan script has no way to write -ia and has no letter for schwa. One could also view -i(H) and -a as evidence for a Tangut vowel like ɪ (= Arakawa's -I?) which is i-like but lower.

R100 was transcribed with Middle Chinese *-ək and *-ət sinographs. Gong reconstructed their rhyme as *-ə in the Tangut period.

R100 was never used to transcribe Sanskrit.

It is odd that Gong reconstructed R100 as a member of his rhyme group VI with ə and ɨ even though it is positioned between two rhyme group II rhymes (R99 -eeʳ and R101 -jiiʳ in his reconstruction). Perhaps R100 was the Grade II counterpart of R99 and R101:

R99 -əəʳi (Grade I; Gong's -eeʳ)

R100 -ɪɪ (Grade II; Gong's -jɨɨʳ)

R101 -iiʳ (Grade III; Gong's -jiiʳ)

The cognates of the R100 word

TT3509 'four', transcribed as

ldi(H) and zlaH in Tibetan

*lə in Tangut period NW Chinese

end in front vowels:

gDong-brgyad rGyalrong kɯ βde < Proto-rGyalrong *pə-tlej (Jacques 2004: 314)

Written Tibetan bzhi < *bli

Written Burmese leḥ

Old Chinese *s-hli-s (Gong proposed *ljid > *sjid with an odd sound change of *l- > *s-; Shensi Middle Chinese *sit may have *-t < *-s rather than from *-t-s)

Hence I think R100 originated from a pre-Tangut front vowel and I would reconstruct 'four' as

Tangut lɪɪʳ < pre-Tangut ?*rʌ-lii

(I have tentatively resurrected the hypothesis of *Cʌ- as a source of Grade II, though I questioned it in two posts last month. One could rewrite the pre-Tangut form as *X-r-lii with *X- representing the unknown conditioning factor for Grade II.)

I can then reconstruct 'water' as

Tangut ʒɪɪʳ < pre-Tangut ?*rʌ-tiiH or ?*X-r-tiiH

(-H = source of rising tone)
with affrication before *i followed by medial lenition:
*-ʒ- < *-dʒ- < *-tʃ- < *-t-

If R100 really is Grade II, it's unlikely that 'water' had z-, an initial that otherwise never appears before Grade II rhymes.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2008 Amritavision