David Boxenhorn got me thinking about how Tangut dialectal variation might be reflected in the variation among Tibetan transcriptions of Tangut.

Tai's (2008) PhD dissertation

classifies the Tibetan transcriptions into six types of handwriting, according to the origin of fragments, the content of fragments, and the writing style of Tibetan transcriptions. These handwritings are labeled from A to F. It demonstrates that, for the first time, the practice of transcription varied among different handwritings. It also points out that the inconsistency of transcription formats noticed by previous scholars are in fact due to different transcription practices among different handwritings, as manifested in the use of prescripts.

Here are the Tibetan transcriptions for Tangut Grade I rhymes 44, 34, and 38 classified by handwriting plus Tibetan transcriptions found by Nishida but not listed in Tai (2008):

Rhyme Gong 1997 This site earlier This site now Tibetan transcriptions
A B C D Nishida
R44 -ew -ew -ɑɛw -i (9) -e (1) -o (1) -iH (9) -i (6) -e (1) -iH (4) -i (3) -iH (6) -i (3) -a (1)
R34 -ej -e -ɑɛ -e (6) -i (4) -eH (3) -e (2) -i (1) -i (5) -iH (3) none
R38 -eej -ee -ɑɛɛ -eH (1) -e (1) -eH (1) -e (1) -eH (1) -eH (1) -e (1) -uH, -o after m-

General remarks

None of these three rhymes are represented in the E transcriptions. The F transcriptions are difficult to read and were excluded from Tai's analyses.

Foreign vowel or vowel-glide sequences are usually transcribed in Tibetan as VHV. (Exceptions: Sanskrit ai and au are be transliterated with special letters.) Although VHV is common in Tibetan transcriptions of Chinese, it is absent from Tibetan transcriptions of Tangut: e.g, there are no transcriptions like -eHi corresponding to Gong's reconstructed Tangut -ej, etc. It is not clear what the absence of diphthongs in the Tibetan transcriptions means. Did the transcribed dialects lose final glides? Or did Tangut diphthongs contain elements that were difficult for Tibetans to perceive and/or write: e.g., Tibetan has no schwa, so the əɛ I proposed earlier for Grade I e would be transcribed as -e.

The function of final -H in Tibetan transcriptions is unclear. It could signify a final consonant or be a diacritic indicating an unusual reading of the preceding vowel. I don't think it represents vowel length which is normally indicated by a subscript H-like symbol in Tibetan transliteration of Sanskrit. In Written Tibetan, final -H is a mater lectionis with a sound value of zero: C1C2H is read as C1C2a.

R44 is mostly transcribed in Tibetan as i(H), even though Chinese transcriptions in the Pearl have a schwa. Did -ew become -iɰ or even just -i in the dialect(s) underlying the Tibetan transcriptions? But -i(ɰ) would not account for the non-i transcriptions. Maybe the rhyme was something like -ɘɰ or -ɪɰ with a vowel lower than i.

R38 and R34 should have similar transcriptions if Gong's reconstruction or my reconstructions are correct, but they don't. R34 has i-transcriptions whereas all R38 transcriptions contain e. Was R34 a higher vowel than R38 in the dialect(s) underlying the Tibetan transcriptions?

1.9.0:18: The C transcriptions never have non-i vowels for R44 and R34. Does this reflect a greater consistency on the part of the transcriber or a difference in dialect: i.e., the dialect underlying the C transcriptions had higher vowels for R44 and R34 than other dialects?

The -uH and -o transcriptions listed in Nishida (1964: 53) for R38 are baffling. Do they represent nonstandard dialectal readings not listed in Tangut dictionaries? I cannot imagine anyone hearing an e-like vowel (as indicated by the A, B, and D transcriptions) and writing it as u or o. KIM-RENAUD'S MIDDLE KOREAN VOWELS (PART 1)

I reconstruct the vowels of Middle Korean as

front central back unrounded back rounded
neutral/'yin' i ə ɯ u
a ʌ o

The vowels in the top row are 'yin' (Korean ŭm) except for neutral i.

The vowels in the bottom row are 'yang' (Korean yang).

This is a slightly modified version of my reconstruction in Miyake (2003: 116).

I was quite surprised to see Young-Key Kim-Renaud's (2010) new reconstruction. Differences from mine are in bold:

neutral 'yin' 'yang'
high i ɨ u
ə o
ɑ ɔ

Although I have seen her values for nonlow vowels (i, ɨ, ə, u, o) in other reconstructions before, I was surprised by her reconstructions of the low vowels (ɑ, ɔ) which seem to be reversed.

Like many linguists, Kim-Renaud believes that a Great Korean Vowel Shift (GKVS) transformed the Middle Korean vowel system - whatever it might have been - into the modern Korean vowel system, whereas I think the middle Korean vowel system survived intact with the exception of ㆍʌ* which only survived as [ɔ] in the Cheju dialect.

front central back unrounded back rounded
neutral/'yin' i ə ɯ u
a (lost!) o

As far as I know, no modern Korean dialect has a vowel system like any of the systems reconstructed by GKVS proponents: e.g., no dialect has an unrounded ㅜ, a high ㅗ, etc. Thus Korean dialects must be descended from dialects that had all undergone the GKVS. If there was a GKVS sometime after the creation of the hangul alphabet for Middle Korean in the 15th century, is it really possible that there were no outlier dialects that had escaped the GKVS within the past six centuries? The English Great Vowel Shift was

not entirely uniform, and differences in degree of vowel shifting can sometimes be detected in regional dialects both in written and spoken English.

Are there no archaisms in Korean dialects preserving pre-GKVS vocalism? Isn't it easier to account for the relative uniformity of vowels in Korean dialects by not positing an exceptionless GKVS? For instance, if ㅏ is a in all modern dialects, why not reconstruct it as a in Middle Korean? Kim-Renaud's reconstruction is the only one I have seen in which ㅏ shifted from ɔ to a, "forcing the low central vowel [her ㅓ ɑ] to raise [to the position of her ㅡ ə], and thus starting a chain reaction."

Kim-Renaud's reconstruction represents "the basic vowels of the Korean alphabet in 1446" which had not yet been through the GKVS. Sino-Korean was borrowed from Middle Chinese approximately seven centuries before 1446. For details, see Miyake (2003: 109-117). Assuming that Kim-Renaud's reconstruction is correct, I would expect

Middle Chinese -like vowels to be borrowed as pre-GKVS ㅓ ɑ which became post-GKVS ㅓ ə

Middle Chinese -like vowels to be borrowed as pre-GKVS ㅏ ɔ which became post-GKVS ㅏ a

There is no absolute consensus on the reconstruction of Middle Chinese, whose dialects must have been quite diverse. Nonetheless, there is widespread agreement on many points: e.g., that the rhyme category 歌 was something like *-a or *-ɑ.

Middle Chinese 歌 *kɑ 'song' was borrowed into Japanese as ka and into Vietnamese as ca. . I would expect 歌 to be borrowed into pre-GKVS Korean as which would became post-GKVS 거 kə. But in fact the modern Sino-Korean reading for 歌 is 가 ka, implying a pre-GKVS kɔ. Although 歌 was read as *kɔ in very late Middle Chinese and still has similar readings in modern Chinese languages (e.g., Cantonese kɔ), Sino-Korean predates the > shift in Chinese. It is simpler to project modern ㅏ a all the way back to 8th century Old Korean:

Middle Chinese 歌 *kɑ 'song' > 8th c. Kor *ka > 15th c. Kor 가 ka > 21st c. Kor 가 ka

(rather than 21st c. Kor 거 which Kim-Renaud's reconstruction would predict)

cf. Sino-Japanese ka, Sino-Vietnamese ca [kaa]

Assuming that a has remained unchanged through Korean language history simplifies Koreo-Japonic comparisons. Middle Korean a corresponds to Old Japanese and Proto-Japonic *a (Frellesvig & Whitman 2008: 35) in words that are either pre-GKVS loans or inherited from a common pre-GKVS ancestor: e.g.,

Middle Korean 다히- tahi- 'make a fire' : Proto-Japonic *tak- 'burn'

Middle Korean 살 sar 'arrow' : Old Japanese sa 'id.' (not reconstructible in PJ)

Middle Korean 잣 tsas 'castle' : Old Japanese sasi 'id.' (not reconstructible in PJ)

(The last two if not all three are loans. If 'make a fire' is a loan, it is a very old.)

To accomodate Kim-Renaud's reconstruction, this equation would have to be rewritten as Middle Korean ɔ : OJ/PJ *a. One could posit a shift of Proto-Koreo-Japonic to Proto-Japonic *a, but what about loans into Western Old Japanese like 'arrow' and 'castle' which would postdate such a Proto-Japonic shift? Why wouldn't pre-GKVS ɔ ever be borrowed into Japanese as o?

If Korean ㅏ a was always a, then Kim-Renaud's derivation of ㅓ ə from ɑ cannot be correct. If her derivation were correct, I would expect Sino-Korean ㅓ ə < pre-GKVS ɑ to correspond to low a in Sino-Japanese and Sino-Vietnamese. But in fact ㅓ corresponds to nonlow vowels in SJ and SV: e.g.,

Midlde Chinese 魚 *ŋɨə 'fish' > 8th c. Kor *ŋə > 15th c Kor 어 (ŋ)ə > 21st c. Kor 어 ə

(rather than 21st c. Kor 아 a which Kim-Renaud's reconstruction would predict)

cf. Sino-Japanese gyo < *ŋgyə, Sino-Vietnamese ngư [ŋɨɨ]

Korean ㅓ ə corresponds to OJ/PJ *a as well as (Frellesvig & Whitman 2008: 35; F&W use the letter o instead of ə). Although the first correspondence fits Kim-Renaud's reconstruction -

post-GKVS ㅓ ə (< pre-GKVS ㅏ *ɑ) : OJ/PJ *a

- the second one does not:
post-GKVS ㅓ ə (< pre-GKVS ㅏ *ɑ) : OJ/PJ (not *a!)

I suspect that (a) one of these two correspondences is invalid or (b) one correspondence reflects borrowing (from an early Koreanic language with *ə/a merger?).

Next: Was arae a (ㆍ) really arae o?

*In this series, I will only be concerned about the simple vowels of Middle Korean and their modern Korean reflexes, not complex vowels like Middle Korean ㅔ əy which became modern Korean e. AN IMP-LƏƐW-SIBLE RECONSTRUCTION (PART 2)

In my Tangut reconstruction from the last three years, the rhyme of 'one' (1lew R44) has the same vowel (e) as Grade I -e R34. R44 and R34 still share a vowel in my latest reconstruction. Gong and I regarded R38 as the long version of R34 whereas Sofronov and Arakawa reconstructed R38 as R34 pluas a final consonant:

Rhyme Nishida 1964 Sofronov 1968 Li Fanwen 1986 Gong 1997 Arakawa 1999 This site earlier This site recently This site very recently Tibetan transcriptions Used to transcribe Sanskrit
R44 -əw -eɯ -əu -ew -eu -ew -əɛw -ɑɛw -i, -iH, -e, -o, -a never
R34 -ai -ej -e -e -əɛ -ɑɛ -i, -iH, -e, -eH
R38 -e -aiC -e -eej -e' -ee -əɛɛ -ɑɛɛ -e, -eH; Nishida: also -uH, -o after m-

In contrast with Tangut, both Written Tibetan (hereafter, 'Tibetan') and Sanskrit have very limited vowel systems:

Written Tibetan vowels



Sanskrit vowels (excluding syllabic liquids)

ai a au


The similar Tibetan transcriptions for R44 and R34 may corroborate the assumption that those two Tangut rhymes have the same vowel. But none of the reconstructions of R44 has an i like their Tibetan transcriptions (though perhaps Tibetan i could be an attempt to transcribe Sofronov's ɯ) Moreover, i is absent from the Tibetan transcriptions of R38.

Nishida (1964: 53) found uH and o after m- in Tibetan transcriptions of R38 which are absent from Tai (2008).

Sanskrit has no -ew-like rhyme, so it's not surprising that R44 was never used in Tangut transcriptions of Sanskrit.

However, Sanskrit does have ee (normally romanized as e since its length is predictable). If R34 or R38 were a simple e(e), one or both rhymes should have been used to transcribe Sanskrit e. But the absence of such transcriptions implies that both rhymes were not -e(e). The variation in Tibetan transcriptions implies that both rhymes had no Tibetan equivalent. Hence I feel justified in reconstructing them (and other rhymes like R44 sharing their vowels) with vowels other than simple e(e):

R34 -ɑɛ (or -ae in simpler notation)

R38 -ɑɛɛ (or -aee in simpler notation)

R44 -ɑɛw (or -aew in simpler notation)

My short diphthong ae was inspired by Khmer ae (as in Khmer Khmae < *khmɛɛr 'Khmer').

My long diphthong aee is like the aee of Avestan corresponding to Sanskrit ee: e.g.,

Av haeena 'army' : Skt seenaa 'id.' AN IMP-LƏƐW-SIBLE RECONSTRUCTION

I am skeptical of reconstructions that look like random collections of IPA symbols. Phonemic inventories of real languages tend not to look chaotic.

My latest reconstruction of Tangut vowels has a structure, but it still bothers me because it contains some diphthongs that I've never seen in any other language: e.g., I now reconstruct 'one' as 1ləɛw (as in my last post) instead of 1lew. 'One' is a Grade I syllable with e as its base vowel:

Base vowel i, e ɨ, a u, o
Grade I: -high, -palatal ə + lowered lowered
Grade II: -high, +palatal lowered ɪ + lowered ɪ + lowered > lowered and fronted
Grade III: +high; -palatal ɨ + raised raised
Grade IV: +high; +palatal raised i + raised i + raised > raised and fronted

Grade I front vowels are lowered and partly depalatalized:

i > əɪ

e > əɛ

These diphthongs could be rewrrtten as əi and əe.

Does any languages have such schwa + V diphthongs? UPSID has only two such diphthongs, əi and əu. UPSID is not a complete inventory of all the sounds of the world's languages, so that does not mean other əV-diphthongs are impossible. Nonetheless, I am hesitate to reconstruct diphthongs that might not be attested even if I can pronounce them.

UPSID does have the diphthongs ae and ao. I have considered reconstructing such diphthongs in Grade I in the past. I could reconstruct a height difference in Grade I e so that the diphthong's two halves and ɛ) are at different heights like its Grade III counterpart ɨe.

Base vowel u i a ə e o
Grade I: -high, -palatal ʊ əɪ ɑ ə ɑɛ ɔ
Grade II: -high, +palatal ʏ < ɪʊ ɪ a < ɪɑ ɪə ɛ œ < ɪɔ
Grade III: +high; -palatal u ɨi ɐ ɨ ɨe o
Grade IV: +high; +palatal y < iu i æ < e ø < io

ɑɛ could be rewritten as ae. In Khmer, the 'first register' reflex of earlier is ae. The development of Khmer vowels has inspired my Tangut reconstructions for the past three years. The [±high] feature in my latest system and 'register' in Khmer are associated with similar vowels (in Pinnow's 1980 notation with slight modifications):

Earlier Khmer i e ɛ ɨ ə a u o ɔ
*voiceless initial > 1st register: lower vowels əj e ae ə a o ao ɑ
*voiced initial > 2nd register: higher vowels i ɪ ɛ ɨ ə u ʊ ɔ

Although initial voicing was the conditioning factor for vowel shifts in Khmer, it had no influence on the development of Tangut vowels. I think Tangut (and Chinese) grades were conditioned by the presence or absence of 'emphasis' (which in turn had been conditioned by presyllables, uvular initials, and medials).

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2011 Amritavision