Here's an excerpt from Dissected Rhymes reproduced in Nevsky (1960 I: 136, as corrected by Gong 1981), with Tangut Telecode numbers and my modifications of Gong's reconstructions:

V: Rhymes IV: TT3417 giee R40 1.39 III: TT3561 khie R37 1.36 II: TT5780 khio R53 2.44 I: TT3389 kio R53 1.51
# TT Name TT Syllable TT Syllable TT Syllable TT Syllable
a 1 1897 bəu 1.1 4704 gəu 1.1 5632 khəu 1.1 0225 khəu 1.1 4139 kəu 1.1
b 3 1967 siu 1.3 4218 giu 1.3 2821 khiu 1.3 O 4648 kiu 1.3
c 9 0560 1.9 3401 1.9 0607 khɪ 1.9 O 2222 1.9
d 11 3933 tshi 1.11 5018 gwi 1.11 0743 khwi 1.11 O 4713 kwi 1.11
e 20 4739 sia 1.20 1367 gia 1.20 O 3059 khia 2.17 0549 kia 1.20

In part 1, I proposed that TT0225 had initial kɦ- unlike TT5632 which had initial kh-. If Tangut had two types of aspirates, the above table could be simplified as

Rhyme IV. g- III. kh- II. ?kɦ- I. k-
a. -əu gəu khəu ?kɦəu kəu
b. -iu giu khiu O kiu
c. khɪ O
d. -i gwi khwi O kwi
e. -ia gia O ?kɦia kia

and there should be no overlap between the fanqie initial spellers of the kh- and kɦ-columns.

However, the initial spellers of TT5632 khəu and TT0225 kɦəu, TT1994 khji R11 1.11 and TT5787 khji R11 2.10, belonged to the same homophone group (LFW V.29) in the edition of Homophones that ignores tones (Li Fanwen 1986: 305, Sofronov 1968 II: 166). That implies they had the same initial, which conflicts with Dissected Rhymes:

Tangraphic Sea Homophones Homophones entries for fanqie initial spellers Dissected Rhymes
TT5632 and TT0225 nonhomophones with the same rhyme nonhomophones (nature of difference not clear) homophones (in spite of tones) nonhomophones with the same rhyme

What about the fanqie initial spellers of the other kh- and kɦ-tangraphs? (Roman numeral-letter combinations represent columns and rows in the first table above: e.g., IIIb = column III, row b.)

IIIb. TT2821 khiu R3 1.3: TT1994, same initial speller as TT5632 khəu R1 1.1 (IIIa)

IIIc. TT0607 khɪ R9 1.9: TT0918 < initial speller TT1170 < initial speller TT0918 (again)

IIId. TT0743 khwi R11 1.11: TT0410 < initial speller TT1994, same initial speller as TT5632 khəu R1 1.1 (IIIa)

IIe. TT3059 khia R20 2.17: no initial speller known, and no members of its homophone group (LFW V.115) have known initial spellers

Like IIIa, IIIb and IIId have an initial speller homophonous with the initial speller for IIa if tone is ignored. This implies that these four tangraphs had the same initial even though they were placed in two different columns in Dissected Rhymes.

The fanqie initial speller of TT3561 khie R37 1.36, representing the initial of column III, is TT2821 (IIIb).

The fanqie of TT5780 khio R53 2.44, representing the initial of column II, is unknown. The initial is Nishida's guess based on the use of the tangraph as the equivalent of 溪 Middle Chinese *khej, the symbol for the MC initial *kh-. The final was converted into my reconstruction from Sofronov's -jo. The basis for Sofronov's reconstruction kjo (with an unaspirated initial!) is unknown. Sofronov (1968 II: 403) listed the homophone group of TT5780 as V.105 (= LFW V.89), but this group does not contain TT5780 (Sofronov 1968 II: 168, Li Fanwen 1986: 318)

I can't think of any scenario that satisfactorily incorporates all of the above facts. All of the following scenarios have problems, and some sound like copouts:

1. Gong is right: the division between columns II and III in Dissected Rhymes is arbitrary.

But why does the distinction between IIa and IIIa happen to coincide with the distinction between those two tangraphs in Tangraphic Sea and both editions of Homophones? Did the authors of all these works agree on an arbitrary distinction without any basis in phonetics?

2. The fanqie in Tangraphic Sea for IIa and/or IIIa is wrong, and the correct initial speller(s) did not have the same initial.

But I'd rather work with what exists rather than claim it's inaccurate because it doesn't match what I want.

3. The group (V.29) containing the initial spellers of IIa and IIIa in one edition of Homophones was really two groups; the circle that would have separated the two groups was accidentally omitted from that edition, though it is present in the other edition

But the circle dividing V.36 and V.37 in one edition of Homophones definitely only indicates a tonal distinction.

My objection for scenario 2 also applies here.

4. IIa and IIIa had the same initial and final but different medials

But this conflicts with their placement in different columns in Dissected Rhymes, since columns have nothing to do with medials.

5. Nishida (1964: 26) proposed that Dissected Rhymes was a Chinese rhyme table in tangraphic script. IIa and IIIa were homophonous tangraphs representing Chinese nonhomophones.

But if IIa and IIIa were truly homophonous, why are they in separate homophone groups in Tangraphic Sea and both editions of Homophones?

Next: Which edition is which? TWO KINDS OF ASPIRATES IN TANGUT? (PART 1)

Although I still don't regard the list of 36 initials in 五聲切韻 Dissected Rhymes of the Five Sounds as a list of 36 Tangut initials, I now wonder if the list is less artificial than I had assumed.

According to the excerpt from Dissected Rhymes cited in Gong's "Voiced Obstruents in the Tangut Language" (1981), the following two tangraphs had the same rhyme preceded by different initials corresponding to Middle Chinese *kh and *g:

TT0225 (Gong: khu) R1 1.1 'blue'

TT5632 (Gong: khu) R1 1.1 'detest, abhor'

Yet Gong and others reconstructed them with the same initial:

Tangut Telecode Sofronov 1968 Li Fanwen 1986 Gong 1997 This site until now
0225 khu khu khu khəu
5632 khu khu khu khəu

One might expect them to belong to the same homophone group in Tangraphic Sea and Homophones. But they don't:

Tangut Telecode Sofronov's homophone group number Li Fanwen's homophone group number
0225 I.9 I.7
5632 I.10 I.8

Sofronov (1968 II: 162) reproduces the different groupings in the two editions.

In the edition of Homophones that ignores tonal distinctions, group I.7 includes both level and rising tone syllables (khəu R1 1.1 and khəu R1 1.2). Why would TT5632 khəu R1 1.1 in I.8 be separated from TT0225 khəu R1 1.1 which was grouped together with TT0916 khəu R1 2.1 in I.7?

Moreover, they don't even share the same fanqie in Tangraphic Sea:

Tangut Telecode Gong Initial speller Gong Final speller Gong
0225 khu R1 1.1 5787 khji R11 2.10 0379 tsu R1 1.1
5632 khu R1 1.1 1994 khji R11 1.11 3149 ʔu R1 1.1

The fanqie of TT3149 is TT4879 + TT0379. Therefore TT0379 and TT3149 presumably represent the same final shared by TT0225 and TT5632.

These facts imply that TT0225 and TT5632 were not, in fact, homophones. Why arbitrarily assign them to different groups with different fanqie if they were truly homophones? Did the authors of both Tangraphic Sea and Homophones hear a distinction that didn't exist? Or did they really have different initials?

Gong pointed out that the Chinese transcription evidence from the Pearl points to a single initial kh- for both TT0225 and TT5632. The data in Sofronov (1968 II: 6) show that

the homophone group including TT0225 was transcribed with Chinese *kh- (枯)

tangraphs of the homophone group including TT5632 were used to transcribe Chinese *kh-syllables (袴苦)

How can this be reconciled with the different-initial hypothesis? I can imagine two possible solutions:

1. The Pearl Tangut dialect had a single initial corresponding to two initials in the dialect(s) of the other sources.

2. The TT5632 group had initial kh- but the TT0225 group had a different initial similar to Chinese *kh-: e.g., kɦ- (which is not the kɦ- I proposed instead of Gong's g- two weeks ago!).

Gong k- kh- g- ŋ-
My 6.13 proposal k- kh- kɦ- ŋ-
This proposal k- kh- kɦ- (< ?*g-) ŋg- ŋ-

The trouble with solution 2 is that the TT0225 group's initial is equated with Middle Chinese *kh- and the TT5632 group's initial is equated with Middle Chinese *g- (> Tangut period northwestern Chinese *kh-), not the other way around. However, this reversal could reflect the merger of the two initials in Tangut period NW Chinese:

Middle Chinese initial name
Middle Chinese *kh- *g-
Tangut period northwestern Chinese *kh-
Tangut initial kɦ- (< ?*g-) kh-

By 1173, the date of the first edition of Dissected Rhymes, the Tangut had no way of knowing what the original distinction between the two initials was, so they may have associated TPNWC primary *kh- with their kɦ- and TPNWC secondary *kh- (< MC *g-) with their kh-.

One might expect a kh-/kɦ- distinction to be somehow reflected in the Tibetan transcriptions examined by Tai Chungpui (2008), but all of Gong's kh- were transcribed as (C)kh(h)- except for TT1705 transcribed as dgheH.

The transcriptions of the very common tangraph TT1738 'center, among, middle' mix kh- and khh-:

Transcription Scribe Frequency
khha B 7
D 5
dkhaH C 4
kha A 1
E 1
Hkha A 1

but there are no other tangraphs with khh-, so I doubt that khh- was an attempt to transcribe a kɦ- distinct from kh-. The transcription Hkh- is also unique to TT1738, so it probably didn't represent kɦ-. It is dangerous to reconstruct an initial based on transcriptions of a single tangraph.

Moreover, I assume that my kɦ- would be a fairly common initial. The ?kɦəu R1 1.1/2.1 homophone group containing TT0225 has six members whereas the khəu R1 1.1 homophone group containing TT5632 has only two. Maybe the Tibetan transcribers simply couldn't hear the difference between kh- and kɦ-.

Next: kɦ-ultivation?

(6.28.10:48: 'cultivate' is the gloss for TT4003, not TT3059, the tangraph I had in mind. 4003 is Li Fanwen 1997's number for TT3059, another possible -tangraph. See part 2.) A BRANNERIZED NOTATION FOR TANGUT?

Although I plan to write a lot about Hainanese in the near future, my interest in Tangut has not gone away. Just now I rediscovered David Prager Branner's 'anti-reconstruction' for Middle Chinese:

Such a system of transcription does not pretend to represent the actual sound of any real language of an earlier time. Rather, it represents the received Chinese scheme of medieval phonology in the abstract. I have coined the name “anti-reconstruction” for this kind of transcription, by analogy with “anti-matter,” to distinguish it from a true reconstruction that purports to recover actual sounds of earlier times. The anti-reconstruction is a tool for the study of Classical Chinese and an adjunct to the study of linguistic history. If it is sufficiently accessible and unthreatening, it can be introduced to students early in their sinological educations, and will serve as an aid to later study of historical linguistics, prosody, and philology.

George Kennedy's anti-reconstruction was in the textbook for my second semester of Classical Chinese. That system provided the foundation for my studies of Chinese historical phonology over the following seventeen years.

I first saw Branner's system eight years ago, and looking at it again, I see some similarities to the systems of Tangut transcription in my "LCD Approach" post. Both Branner and I used numbers to work around the problem of phonetic precision: we know that rhymes 1 and 2 are different, but we don't know exactly how.

However, I think it's too early to successfully transplant Branner's approach to Tangut. The phonological categories of Middle Chinese are well-established in native sources, whereas the Tangut categories are not. Although all Tangutologists agree that there were 97 level tone rhymes and 86 rising tone rhymes, they do not agree on many other categories. The rhyme categories were explicitly listed in Tangraphic Sea, but there is no explicit list of initials* anywhere. Homophones lists nine groups of initials (one per chapter), but the status of the fourth group is controversial: was it truly distinct from the third group, and if it was, how many initials did it contain - one (Huang Zhenhua 1983 and Li Fanwen 1986), two (Li Xinkui 1980), or four (Nishida 1964 and Arakawa 1999)? A neutral Tangut transcription system requires a consensus that does not yet exist.

*I don't know of any Tangutologists who assume that the list of 36 initials represented in 五聲切韻 Dissected Rhymes of the Five Sounds represented 36 Tangut initials rather than 36 Middle Chinese initials. Nishida (1964: 25) reprinted the list with his reconstructions, It is highly unlikely that Tangut and Middle Chinese had exactly the same number of initials. HAINANESE 前音 'FRONT SOUNDS' (PART 1: INVENTORY)

Here's what they sound like to me:

ɓ b m f
ɗ t d n s dz l
k g ŋ
ʔ h (zero)

Other opinions are welcome.

And here's Xu and Yang's romanization for them.

bz b m f
dz d dl n s r l
g gz ng
(zero) k h

Unlike Xu and Yang, I don't think gz is an implosive [ɠ]. If it is, I'm not hearing it.

There may be two more initials, if X&Y's hiV... and hoV... sequences are phonetically [jV...] and [wV...]: e.g.,

hiang ?[jaŋ]

hoang ?[waŋ]

as opposed to

iang ?[ʔjaŋ]

oang ?[ʔwaŋ]

I have already commented on the z-digraphs for implosives and the inconsistent romanization of simple voiced stops (b, dl, gz). Other odd choices (k [h], h [zero], r [dz] were presumably chosen on the basis of the romanizations of the standard Mandarin initials that they sometimes correspond to: e.g.,

Sinograph X&Y Hainanese IPA (my guesses) Standard Mandarin Mandarin IPA
koa ?hwa kuo kʰwo
hoa ?wa hua xwa
roa ?dzwa re ʐɤ

I wonder if some pronounce X&Y's h as [ɦ], though I didn't hear such a fricative in the video. The romanization may reflect variation not represented by the single speaker in the video.

Next: The origins of the 'front sounds'. HAINANESE -OM AND -OP

are the only two rhymes which only appear in 'chorphans' according to Xu and Yang. Chinese languages tend to avoid sequences of labial segments* within a single syllable: -UP, PVP, and most of all, PUP. Yet Hainanese has -om and -op after nearly all initials including some (but not all) labials:

bom (but no fom or mom)

bop and fop (but no mop)

None of these syllables are glossed. Are they loans from a non-Chinese language and/or native words which had not undergone labial dissimilation?

*An avoidance of PUP-sequences is not an inherent characteristic of Chinese-type languages. Vietnamese has PUP-type syllables consisting of nothing but labial segments.

bom 'bomb' (obviously a loanword)

bóp 'squeeze with one's hand'

phốp pháp 'be plump' (sound-symbolic)

móm 'be toothless, mỏm 'promontory', mõm 'muzzle'

móp 'be hollow'

mồm 'mouth'

vòm 'vault'

Siamese has even more PUP syllables:

-um -uum -up -uup -om -oom -op -oop -ɔm -ɔɔm -ɔp -ɔɔp
b- + + + none! none! + + +
p- + + + + + + + +
ph- + + +
f- + + +
m- + + + + + +
w- + +

I compiled the above table using SEAlang's online version of Haas' Thai Dictionary Project data which permits partial match searching: e.g., typing "Com" allows one to find all syllables ending in -om.

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