This time, I include Gong's forms for comparison. Although I am not sure if 'Proto-Tibeto-Burman' ever existed*, I wanted to see what a *-j-less reinterpretation of Gong's PTB would look like.

p. 466: Correspondences between Written Burmese, Tangut, and Proto-Tibeto-Burman

Pre-Tangut represents what I can reconstruct on the basis of Tangut-internal alternations with minimal external assistance. Thus pre-Tangut lacks segments such as *-t which can only be reconstructed on the basis of external comparisons.

Pre-Tangut *-ɤ- and *-ɯ- represent mid and high vowels of lost presyllables that conditioned downward and upward vowel bending. I assume that variation in vowel height in Tangut is a Tangut-internal innovation and does not reflect a PTB (or Proto-Sino-Tibetan) distinction.

I don't reconstruct sequences like *-uw and *-ij because they are difficult to distinguish from u(u) and i(i). (Can Russians distinguish -ий and -ии by ear without context?)

Notes on the various correspondences:

A1: It is unlikely that PTB *-uw, a sequence of labials, would become Tangut -ə unless *-uw dissimilated from a preceding labial initial. Dissimilation can account for Gong's examples (17-22), but not for his (23) which has a dental initial:

nəʳ2 R90 2.76 'finger' < Gong's PTB *-uw, my pre-Tangut *r-nə < PTB *nəw

A1 and A3: PTB *-w became the achromatic glide *-ɰ after achromatic vowels (*ə, *a) before disappearing. I know of no Tangut-internal evidence for *-w in achromatic open-vowel rhymes, so I cannot reconstruct it at the pre-Tangut stage.

B2 and B3: Presumably Gong derives Tangut -ɨj from his PTB *-ij via dissimilation.

B2-B4: I don't know of any Tangut-internal evidence for *-j in these nonpalatal rhymes, so I cannot reconstruct it at the pre-Tangut stage.

C1-C3: Gong projected a three-way distinction back to the PTB level, but I think that a single original rhyme *-a split into three depending on vowels in early Tangut presyllables and/or preinitials:

*-(a)-a > -a

*-i-a > -i and/or acute preinitial: *T-a

*-u-a > -u and/or labial preinitial: *P-a

Later Tangut presyllabic vowels and preinitials affected *a differently:

*(-ɤ)-a > -a [ɐ] (Grade I/mid series)

*-ʌ-a > -ɑ (Grade II/low series)

*-ɯ-a > ɨa, ia (Grade III and IV/high series)

*T-a > -ạ (tense vowel)

*P-a > -wa (labial medial)

C3: It is unlikely that Gong's nonlabial PTB *-jə would become labial -ju after dental n- and alveolar dz-. I think the labiality of Tangut -u is from a presyllable or preinitial appended to a root with a nonlabial vowel:

nu R3 1.3 'ear' < *Cu-na or *P-na

dzu R2 1.2 'to love' < *Cu-dza or *P-dza

(It is tempting to equate *P- with the m- in Written Tibetan mdzaH- 'to love'.)

D1-D2: Chinese cognates indicate that pre-Tangut *-e may have originated from an earlier *-V + acute consonant in some cases:

nie R37 1.36 'near' < *Cɯ-ne : 邇 OC *nelʔ

miee R40 1.39 'tail' < Cɯ-mee : 尾 OC *məjʔ

(But I would expect OC *-əj to correspond to Tangut -ɨ [correspondence B2]!)

I will present internal evidence for pre-Tangut *-aj as a source of Tangut e next time.

*In other words, I am not sure if Sino-Tibetan really can be divided into Sinitic and Tibeto-Burman (i.e., everything but Sinitic). I prefer Guillaume Jacques' more neutral viewpoint in which Sinitic is just one of many ST branches.

The ever-generous Mahadaatṛ continues to send me Tangutological gems that I can only belatedly digest. One of them is a 2007 article by Gong that I had never heard of, "The Position of Tangut in the Comparative Study of Sino-Tibetan Languages".

In this post, I will 'translate' Gong's tables using my reconstructions of Old Chinese (OC) and Tangut to see if his observations can more or less apply to a different framework.

p. 448: Gong's OC and Tangut have different vowel systems.

Gong's OC vowels

Gong's Tangut vowels (ignoring length, tension, and retroflexion)

Gong's Tangut ɨ and ə are in complementary distribution and could be regarded as a single nonlow central vowel phoneme.

My OC and pre-Tangut reconstructions have the same six-vowel system (ignoring length, tension, retroflexion, and unstressed presyllabic vowels):

However, as I will demonstrate below, I do not necessarily assume that OC and pre-Tangut vowels have one-to-one correspondences.

p. 450: Gong's Proto-Sino-Tibetan (PST) *-j-

Gong believed that OC and Tangut preserved a PST *-j- that was (generally) lost in Written Tibetan (WT) and Written Burmese (WB):

Instead of a *-j-, I reconstruct nonemphasis in OC and high vowel (diphthongs) in Tangut:

Note that Tangut vowels do not necessarily match OC vowels. Currently I have no Tangut-internal evidence to suggest that the pre-Tangut vowels for these words were different from the Tangut vowels.

I believe there was a 'primary yod' (using Bodman's term) in Proto-Sino-Tibetan whose reflexes were distinct from those of Gong's *-j-:

Gong reconstructed OC 'eight' as *priat without *-j-. In Gong's system, Tibetan lost PST *-j- but developed a new -y- in 'eight', whereas I think WT brgyad 'eight' preserves PST *-j- and Gong's *-j- words ('firewood', etc.) never had a medial for Tibetan or Burmese to lose.

p. 452: Correspondences of Benedict's Proto-Tibeto-Burman (PTB) *-ik/*-iŋ

I have ignored the retroflexion in the Tangut rhymes of 'joint' and 'leopard' since it is secondary:

'joint': tseʳw R93 1.87 < *r-tsek

'leopard': zeʳw R93 2.78 < *rɤ-tsek (with *-ts-lenition; cf. gDong-brgyad rGyalrong kɯ-rtsɤɣ < Proto-rGyalrong *-ek)

Gong listed no PTB rhyme for 'new'; I have replaced this with *-ik on the basis of *sik in Matisoff (2003: 660).

Perhaps eʳ in 'joint' and 'leopard' was from an *i lowered to assimilate to a preceding mid vowel:

In this scenario, both words were originally homophonous until the vowel of the presyllable of 'joint' was lost. This syncope postdated vowel lowering and predated the lenition conditioned by presyllabic vowels. Retroflexion could have occurred at any point prior to the loss of the *r(ə)- that conditioned it.

'Joint' underwent a similar lowering in Old Chinese:

*Cʌ-tsik > *Cʌ-tsit > *Cʌ-tsit > *tsit [tsˁɪˁtˁ] > late OC *tset [tsɛt]

pp. 455-456: Correspondences of the 侯 'marquis' and 幽 'secluded' rhymes of Old Chinese

My PTB reconstructions are offhand guesses that require much more thought.

Tangut ɨ(ɨ) and i in -iw could be from a *ə(ə) which raised to assimilate to a preceding high vowel *ɯ:

(There is no way to tell whether the high vowel *ɯ in 'six' was part of the first and/or the second prefix. The vowel [if any] in one prefix could have assimilated to the *ɯ in the other prefix.)

*ɨ in 'six' may have fronted to assimilate to the preceding acute consonants (*ktr- after syncope) or to dissimilate from the following grave consonant *-kw.

Final *-w assimilated to the preceding achromatic vowel , becoming a *ɰ that was later lost.

While looking through Nishida 1964 for examples of TT1465 siw R47 1.46 'new' in his typeset edition of the Pearl, I found three different reconstructions of TT3448 tɕhiw R47 1.46 'six':

p. 193: Pearl 06.4.5: tʃhu (-u is Nishida's R1!)

p. 206: Pearl 20.1.7: tʃhjuɦ (-uɦ is Nishida's R3!)

p. 207: Pearl 20.2.5, p. 218 32.3.11, p. 223 37.1.1: tʃhju (-ju is Nishida's R2!)

Strangely, his reconstructions of the readings for tangraphs in Homophones chapter VII (alveopalatals) don't list 'six' at all (p. 126).

Moreover, his reconstruction of R47 is -jew (p. 55), which doesn't match any of the three rhymes in the above reconstructions of 'six'. If I had not seen the back of his book, I would have guessed that 'six' would be tʃhjew in his system. But his 1966 dictionary lists his reconstruction as tʃhju (p. 435), the most common of his three reconstructions. I presume the other two are typos, since alveopalatals such as tʃh- cannot combine with R1 and R3 (pp. 42-43). (Li Fanwen [1986: 195] lists tɕ- + R3 as a possible syllable, but this is an error for tɕ- + R2.)

Ironically, even though 'new' does not appear in Nishida's version of the Pearl, Nishida included its transcription 悉 in his list of R47 transcriptions (p. 55), but he excluded 抽, the transcription of 'six' which occurs four times in his Pearl edition (see above). ('New' can be found at 26.2.7 on p. 262 of Kwanten's 1982 photographic reproduction.) Nishida also left out chiH, the Tibetan transcription of 'six', though he have been aware of it from Nevsky (1960 II: 16)..

None of this is intended to fault Nishida. I am amazed at how he was able to reconstruct Tangut without referring to the Tangraphic Sea and its fanqie. tʃhju is not an unreasonable reconstruction of 'six', given no other information than the transcription 抽 (*tʃhɨw in my reconstruction) and the placement of 'six' in the 'isolates' section of Homophones chapter VII, indicating that it had an alveopalatal initial and no homophones.

Nishida's 1964-1966 book A Study of Hsi-Hsia Language was originally written in 1962 as a PhD dissertation, and he did not see the Tangraphic Sea until 1963. It is unfortunate that as far as I know, Nishida has not published a full-scale revision of his reconstruction in accordance with the fanqie in the Tangraphic Sea.

At least he was able to refer to the rhyme categories which were listed for some tangraphs in Nevsky's posthumous 1960 Tangut dictionary. Nevsky's entry for 'six' (1960 II: 16) omitted the rhyme category, so Nishida had no way of knowing that 'six' belonged to R47. This explains why he excluded the transcriptive information for 'six' from his section on R47, though I still don't know why he didn't include 'six' in his section on Homophones chapter VII.

I can empathize with Nishida because I too feel like I'm working in the dark. My modifications of Gong's reconstruction do not take into account the phonetic table fragments which I have never seen (with the exceptions of excerpts published in Nevsky and Nishida's works). I cannot imagine trying to reconstruct Middle Chinese without the guidance of the Yunjing tables (though the relationship of the Yunjing 'language', if such a thing existed, to MC dialects remains unclear). An optimal Tangut reconstruction would account for variation in Tangut across space and time as recorded in the phonetic tables, dictionaries, and transcriptions from different periods.

Guillaume Jacques' groundbreaking paper "Essai de comparaison des rimes du tangoute et du rGyalrong" lists the following correspondences for the two rhymes that I've been focussing on:

It is dangerous to draw firm conclusions from only three words ('louse', 'six', 'new'), but so far the external evidence does not back up my hypothesis that Grade IV R47 was more palatal than Grade III R46. 'New' more or less conforms to my expectations (the nonhigh vowel in rGyalrong is unexpected) whereas I would expect 'louse' to be R47 and 'six' to be R46, not the other way around.

08.5.13.3:03: R46 TRANSCRIPTIONS IN COMPASSION AND FILIAL PIETY

Guillaume Jacques' 2007 translation of the Tangut text New Collection of Records of Compassion and Filial Piety has a very convenient index of all of its tangraphs with glosses. He identified seven transcriptive R46 tangraphs for syllables in Chinese names. Unfortunately, there were no transcriptive R47 tangraphs in the Records.

The TPNWC reconstructions are very tentative and do not include tones.

Although there are a couple of sinographs that did not appear in last night's post (攸 and 臼), the basic pattern that I described there also applies to this data:

- Tangut Grade III R46 corresponds to Chinese Grade III.

- The exceptions involve 'best fit' substitutions: there was no Grade IV jiw R47 that would match TPNWC *jew and *jiw, so Grade III jɨw R46 tangraphs were used to represent them.

08.5.12.1:52: GRADING THE TRANSCRIPTIONS OF R46 AND R47

I've been talking about R47 all this time without looking at the transcriptive evidence in depth. The table below compares the transcriptive evidence for R46 and R47 from Sofronov (1968 II: 32-33)*. Sinographs that represent Tangut syllables ending in both rhymes are in bold.

I have no explanation for the final nasal in one Tibetan transcription.

The subscript _合 'closed' indicates a Tangut medial -w-.

I know of no cases of R47 tangraphs used to transcribe Chinese syllables. However, there are many instances of R46 tangraphs representing Chinese syllables: e.g., in the Forest of Categories (Gong, "類林西夏文譯本漢夏對音字研究"):

In both of the above tables, there is a strong correlation between R46 and Chinese Grade III. At first glance, the correspondences between R46 and Grade IV on the one hand and R47 and Grade III on the other seem to conflict with my reconstructions of those two rhymes. However, these unexpected correspondences were necessitated by gaps in the syllabic inventories of the two languages:

- There was no Tangut syllable jiw R47, so jɨw R46 was the closest available equivalent of Chinese Grade IV *jew (堯姚陶繇) and *jiw (由幽) in spite of its nonpalatal vowel ɨ.

(One might conclude that R46 2.40 had a higher vowel than R46 1.45 on the basis of these six transcriptions of Grade IV, but both level and rising tone tangraphs were used to transcribe Chinese *-jew and *-iw syllables.)

- There were no Chinese syllables with *tʃh- and *ʒ- followed by the Grade IV final *-iw, so Grade III 抽 *tʃhɨw and 柔 *ʒɨw were used to transcribe the Tangut Grade IV R47 syllables tɕhiw and ʑiw.

The 'outlier' transcriptions

Grade III TT2500 (a surname) lɨw R46 1.45 for Grade I 婁 *ləw (I follow Gong in treating 婁 as Grade I, though 婁 also had a Grade III reading *lɨw)

Grade I 苟 *kəw for Grade III TT3548 YEAR kɨw R46 1.45 (Pearl 9.4.1, 11.1.8, 11.1.10, 35.1.4)

confirm that R46 was less palatal than R47. If R46 had a palatal vowel and/or glide (e.g., as in Gong's reconstruction -jiw), it would make little sense to associate it with Chinese *-əw.

*I wish I had a good index for the Pearl. Kwanten's (1982) index only covers Chinese transcriptions of Tangut and has many errors. Nishida, Sofronov, Kwanten, and Li Fanwen's lists of Chinese transcriptions do not match: e.g., Nishida (1964: 55) did not list 苟 as a transcription of an R46 tangraph even though it is the transcription of TT3548 YEAR (Pearl 9.4.1, 11.1.8, 11.1.10, 35.1.4).

**Kychanov and Sofronov (1963: 96) list -o but not -au as a Tibetan transcription of R46. I don't know whether they have found an -o transcription absent from other lists or have read -au as -o.

08.5.11.2:11: SEVEN MOUTHS AND TEN WATERS

(This post is ultimately relevant to Tangut, though patience is needed to see the connection.)

The title refers to two Sino-Korean readings that have bothered me for some time:

叱 즐 cUl 'scold' : Middle Chinese *tɕhit

written as 口 'mouth' (semantic) + 七 'seven' (phonetic)

normally read 질 cil; some sources do not list cUl at all

Tong-A hyOndE hwaryong okphyOn (1987 ed.) regarded cil as 'vulgar' and cUl as the correct (though unused) reading

but the (artificial?) prescriptive reading in Tongguk cOng'un is chirq [tɕhirʔ]

Yu Changgyun (1994: 75) regarded 叱 as a phonogram for Old Korean *sUr (not *sir!) as well as *s

汁 즙 cUp 'juice' : Middle Chinese *tɕip

written as 氵 'water' (semantic) + 十 'ten' (phonetic)

Tong-A hyOndE hwaryong okphyOn (1987 ed.) regarded cUp as 'vulgar' and 집 cip as the correct (though unused) reading

There are other readings with U instead of an expected i: e.g.,

膝 슬 sUl 'knee' : Middle Chinese *sit

which is supposed to be homophonous with

悉 실 sil 'all' : Middle Chinese *sit, used to transcribe Tangut siw R47 1.46 'new'

輯 즙 cUp 'gather' : Middle Chinese *dzip

which can also be homophonous with

集 집 cip 'gather' : Middle Chinese *dzip

Tong-A hyOndE hwaryong okphyOn (1987 ed.) regarded cip as 'vulgar' and cUp as the correct (though unused) reading

Some of these readings (e.g., cUl 'scold' and cUp 'gather') may not exist outside dictionaries but others are attested in real words: e.g.,

汁液 즙액 cUbEk 'juice' < cUp + Ek

膝蓋骨 슬개골 sUlgEgol 'kneecap'

(though I don't know how old these words are - they may be modern coinages incorporating readings created by dictionary compilers)

In any case, I am more interested in the class of readings with SK U instead of i rather than in particular readings. All of these readings have the shape {s, c}U{l, p}. There is no shift of i to U in Korean syllables of that shape. Therefore the U [ɨ] must reflect a nonpalatal, nonlabial vowel in the source dialect of Chinese.

I hypothesize that there was a process of *i-retraction in closed syllables that occurred in different ways in different Late Middle Chinese dialects.

1. *i-retraction and lowering in southern Late Middle Chinese

MC *i regularly corresponds to Cantonese a [ɐ] and Sino-Vietnamese â [ə] in closed syllables:

叱 Ct chik (irregular coda), SV sất (irregular initial), xất

汁 Ct cap, SV trấp (irregular initial), chấp

膝 Ct sat, SV tất

輯 Ct cap, SV tập

This retraction occurred after all initials including grave initials:

民 Ct man, SV dân < MC *mjin

SV d- is the regular reflex of MC *m before *i:

dân < *jən < *mjən < *mjɨn < *mjin

敏 Ct man, SV mẫn < MC *mɰin

吉 Ct kat, SV cát [kaat] (irregular further lowering and lengthening) < MC *kjit

僅 Ct kan, SV cẩn < MC *kɰin

2. *i-retraction in a northeastern Late Middle Chinese dialect

MC *i retracted to *ɨ after sibilants in some closed syllables in one of the source dialects of Sino-Korean:

叱 SK cUl < NELMC *tɕhɨr < MC *tɕhit

汁 SK cUp < NELMC *tɕɨp < MC *tɕip

膝 SK sUl < NELMC *sɨr < MC *sit

輯 SK cUp < NELMC *dzɨp < MC *dzip

Palatal sibilants may have retracted to postalveolars: e.g., *(t)ɕ > *(t)ʃ.

Alternate readings with i (cil, cip) were borrowed from a nonretracting dialect.

*i-retraction did not occur after grave initials other than nonpalatalized velars:

民 SK min < MC *mjin

敏 SK min < MC *mɰin

吉 SK kil < NELMC *kir < MC *kjit

but

僅 SK kUn < NELMC *kɨn < MC *kɰin

3. *i-retraction in another northeastern Late Middle Chinese dialect

MC *i retracted to *ɨ after postalveolar sibilants in closed syllables in the ancestor of standard Mandarin:

叱 Md chi [tʂhr̩] < NELMC2 *tʃhɨr < MC *tɕhit

汁 Md zhi [tʂr̩] < NELMC2 *tʃɨp < MC *tɕip

but

膝 Md xi [ɕi] < NELMC2 *sir < MC *sit

輯 Md ji [tɕi] < NELMC2 *dzip < MC *dzip

If *i-retraction occurred after grave initials, the resulting *ɨ must have fronted back to i, since there is no trace of a nonfront vowel:

敏 Md min < MC *mɰin

民 Md min < MC *mjin

僅 Md jin [tɕin] < MC *kɰin

吉 Md ji [tɕi] < MC *kjit

4. *i-retraction in the Yunjing Late Middle Chinese dialect?

The identity of the Yunjing dialect is unknown. Nonetheless, the categories of the Yunjing (e.g., the 'grades') have been used as a framework for Middle Chinese reconstruction. Various types of evidence surveyed by Pulleyblank (1984) indicate that Grade III was less palatal than Grade IV. Therefore I wonder if *i-retraction was a feature of Grade III but not Grade IV in the Yunjing dialect:

叱 YJLMC ?*tʃhɨt < MC *tɕhit (Grade III)

汁 YJLMC ?*tʃɨp < MC *tɕip (Grade III)

敏 YJLMC ?*mɨn < MC *mɰin (Grade III)

僅 YJLMC ?*kɨn < MC *kɰin (Grade III)

膝 YJLMC *sit < MC *sit (Grade IV)

輯 YJLMC *dzip < MC *dzip (Grade IV)

民 YJLMC min < MC *mjin (Grade IV)

吉 YJLMC *kit < MC *kjit (Grade IV)

Consequences for Tangut reconstruction

The Tangut were familiar with the Chinese rhyme table tradition (if not the Yunjing itself) and the arrangement of the Tangraphic Sea rhymes was partly based on a Tangutization of the Chinese grade system: e.g., the lax part of Gong's rhyme group IX consisted of four rhymes in ascending grade order:

(I have previously interpreted Arakawa's -eu: as -eeu, but now I assume he meant short e followed by long u. I don't know why Arakawa reconstructed his Tangut Grade III as long. Chinese Grade III is not correlated with length in any modern Chinese language.)

It's not clear how Gong and Arakawa differentiate R46 from R47 since there are minimal pairs with the same initials preceding both rhymes. However, if Yunjing Grades III and IV were differentiated in terms of palatality, and if Tangut Grades III and IV had a similar distinction, then I can reconstruct R46 as -ɨw and R47 as -iw:

Next: If R47 was Tangut Grade IV, was it transcribed with sinographs for Middle Chinese Grade IV syllables?