is the Tangut pronunciation of

the phrase written right to left at the bottom of the graphic in my last post.

niɨɨ tʊ̣ ɣɑ̣ kew

niɨɨ tʊ̣ giɨɨ kew

The leftmost character is

kew 'year'

from my New Year's greeting

kew siw bɛɛ rəɛʳ!

lit. 'year new happy'

The characters that change from year to year are

Year	Characters	Readings (right to left)	Meanings (right to left)
2011		ɣɑ̣ ləɛw	ten one = eleven
2010		ɣɑ̣	ten
2009		giɨɨ	nine

The two characters other than 'year' that don't change are

niɨɨ tʊ̣ 'two thousand' (right to left)

niɨɨ tʊ̣ ɣɑ̣ ləɛw kew 'two thousand ten one year' = 2011

niɨɨ tʊ̣ ɣɑ̣ kew 'two thousand ten year' = 2010

niɨɨ tʊ̣ giɨɨ kew 'two thousand nine year' = 2009

Can you guess what this means?

... in Tangut:

kew siw bɛɛ rəɛʳ!

lit. 'year new happy' - the reverse of English!

The Tangut year of the (metal) rabbit begins on February 3. The graph for tsɪʳ 'rabbit' (taken from Li Fanwen's Tongyin yanjiu) is at the top of the graphic below:

Can you figure out what the five graphs at the bottom mean?

Hint 1: They are read from right to left.

Hint 2: I've already mentioned one of them earlier in this post.

Hint 3: Last year, I wrote only four out of this year's five characters (from right to left)

niɨɨ tʊ̣ ɣɑ̣ kew

niɨɨ tʊ̣ giɨɨ kew

Previous Tangut reconstructions are full of palatal medial vowels and glides that don't correspond to -y- in Tibetan transcriptions of Tangut or Sanskrit transcribed in Tangut. My latest reconstruction based on David Boxenhorn's 2 x 2 hypothesis of Tangut grades almost entirely avoids those palatals (in bold below):

Note that Gong and Arakawa merge Grades III and IV into a single grade.*

The nonuse of R18 in the Tangut transcription of Sanskrit implies that R18 did not sound like any Sanskrit vowel. My lower low front R18 -æ/-a (the same vowel written with my old letter and the IPA letter) has no Sanskrit equivalent. On the other hand, the absence of R18 could be an artifact of a small number of Tangut transcriptions of Sanskrit.

However, my lower low back R17 -ɑ and upper low front R20 -æ (the most common transcription of Sanskrit a) are less like Sanskrit upper low central short a [ɐ] than upper low central R19 -ɐ. These mismatches are not necessarily counterevidence against the 2 x 2 hypothesis. R19 only occurred after certain initials. Hence imperfect matches (R17 and R20) had to be chosen to transcribe Sanskrit -a after other initials: e.g., Skt mva and ta were transcribed as Tangut bɑ ([mbɑ]?) R17 and tæ R20 since the syllables mvɐ and tɐ did not exist in Tangut. But couldn't the Tangut have created special R19 syllables like bɐ and tɐ for approximating Sanskrit? Were the Tangut actually approximating Sanskrit, or were they trying to pronounce Sanskrit as filtered through Tibetan and/or Chinese?

Next: A more serious problem a-head.

*Arakawa split his Grade III into a IIIa and IIIb that correspond to my III and IV, but his IIIa and IIIb rhymes are identical. Arakawa has a Grade IV in only one rhyme (R21) that is Grade III in Gong's reconstruction and my old and new reconstructions.

10.12.29.23:24: A 2 X 2 = 4 GRADE HYPOTHESIS FOR TANGUT VOWELS

David Boxenhorn proposed the following 2 x 2 scheme for the four grades of Tangut:

Ladefoged and Maddieson (1996: 306) state that "[t]here is considerable similarity" between ±ATR vowels in Akan (Fante is a dialect of Akan) and ±pharyngealized vowels in Even (the pharygealization of ɑ is unwritten):

Pharygealized vowels are in Hongyan Qiang, a relative of Tangut (Evans 2006a and 2006b).

Perhaps Tangut had pharyngealization or velarization instead of ATR. It is impossible to reconstruct such fine details on the basis of the extant evidence. To avoid arbitrarily picking one distinction among these possibilities, I'll simply replace David's ATR with height:

Note that Grade II a here is a low front vowel, not a low central or back vowel.

A table of Tangut vowel components must be complex since Tangut had 105 rhymes that were distinguished without final consonants:

The above two tables exclude long, nasal, tense, and retroflex variants..

The features of this scheme are like those of Pulleyblank's reconstruction of the four grades for Late Middle Chinese, though the actual vowel values are different:

Here's a sample of what Late Middle Chinese might be like if its four grades were like those in my latest Tangut proposal:

I reconstruct high vowels in LMC Grade III and IV -an rhymes to account for -ɨ-, -i-, and -j- in Chinese languages and LMC-based borrowings in Vietnamese, Korean, and Japanese.

I do not reconstruct high vowels for Tangut Grade III and IV a since no Tibetan transcriptions contain -y-.

10.12.29.00:29: AN Æ-WKWARD SITUATION

We can get some idea of the difference between Middle Chinese Grade I and II *a-like vowels by looking at their reflexes in open syllables in modern Chinese languages in Hanyu fangyin zihui (1962):

Grade I

Grade II

Vowels that are in both tables are in bold.

No single Chinese language known to me has the same regular reflex for both Grade I and Grade II: e.g.,

(-a readings of a handful of Grade I words like 他 are irregular archaisms which probably spread through borrowing.)

Grade II reflexes are almost always lower than Grade I reflexes and are never higher.

Grade II reflexes may be palatal whereas Grade I reflexes never are. Conversely, Grade I reflexes may contain u, whereas Grade II reflexes never do.

I conclude that the Grade I *a-vowel was backer and higher whereas the Grade II *a-vowel was fronter and lower.

I generally reconstruct Grade II vowels as lower than their Grade I counterparts in both Middle Chinese and Tangut:

Middle Chinese

(MC Grade I *e merged with Grade IV *ie in the dialect underlying the Yunjing rhyme tables.)

Tangut (excluding long, tense, and retroflex vowels; ignore schwas for height comparisons)

See this post on my use of æ instead of IPA a to represent a low front vowel.

The basic pattern of higher/lower pairs is reminiscent of Fante ±ATR pairs with one exception in bold:

As much as I'd like the reconstructed MC/Tangut a-vowel pairs to match the Fante a-vowel pair, the evidence above points in the opposite direction.

Here's a radical reworking of the Tangut grade system that makes it more like Fante:

Grade I and Grade III/IV are distinguished by ATR, whereas Grade II has some other feature like velarization which would be a trace of the medial *-r- that Gong (1993) proposed as a source of Grade II. Any proposed quality for Grade II would have to be combine with tenseness in the second rhyme cycle and retroflexion in the third rhyme cycle. Does any language have doubly exotic vowels that are simultaneously velarized and tense or retroflex?

10.12.27.21:45: TANGUT VOWEL NOT-Æ-TION

David Boxenhorn asked me how Tangut Grade II æ is lower than Tangut Grade I a. That question could also apply to the reconstruction of Chinese in the previous entry. My answer has two parts.

First, my use of the symbols æ and a does not match IPA:

I don't understand why IPA has no symbol for a low central vowel even though that vowel is extremely common. (89% of languages in UPSID have it!) Since such a vowel is more common than a low front vowel in my reconstruction of Tangut and in general (the latter is only in 6% of languages in UPSID), I prefer to represent the former with the easy-to-type symbol a. It's also possible that the low central vowel of Tangut was actually back *ɑ.

In strict notation, what I wrote as Middle Chinese Grade I *a would be back *ɑ. Schuessler's (2007, 2009) notation for reconstructed Chinese presumably conforms to IPA, whereas my Chinese notation matches my Tangut notation:

Neither my Tangut nor my Chinese reconstruction have a distinction between low central and back vowels, so I use a to represent both.

I can more easily distinguish between the letters æ and a than between *a and *ɑ. Moreover, the right side of the letter *æ reminds me of how that vowel corresponds to e in colloquial Taiwanese and early Chinese borrowings into Vietnamese* and Japanese.

Second, if I try to pronounce the low central vowel of Japanese (or my Tangut reconstruction) and lower my jaw even further, the resulting 'ultralow' vowel is hard to distinguish from my normal low central vowel. But if I front the vowel a bit, the result sounds like English [æ] to me. Fronting makes the ultralow vowel more distinctive.

12.27.22:05: I forgot to mention that I was also influenced by Baxter's (1992) Middle Chinese notation:

*Strictly speaking, Middle Chinese *æ was borrowed as Vietnamese e [ɛ], not Vietnamese ê [e].

10.12.26.23:36: GRADING IN TANGUT AND CHINESE

Yesterday, I mentioned that "Gong established that the Tangut grade system roughly matched the Chinese grade system". I should have written that the idea goes back to Sofronov and Nishida in the 60s. But as far as I know, Gong was the first to write an entire article about the two systems. This much shorter article is a response to David Boxenhorn's question concerning what is known about grades in Chinese.

In the Tangut rhyme dictionary Tangraphic Sea, rhymes are grouped in clusters of three or four. The Tibetan transcriptions of these rhymes indicate that the rhymes of these clusters must have been similar: e.g., the first cluster (rhymes 1-4) was generally transcribed as Tibetan -u. The distinctions between rhymes in a cluster were either absent in the dialect that the Tibetans transcribed or were impossible to write with the Tibetan alphabet which only had five vowel letters. These distinctions are called 'grades' because they correlate with the 'grade' distinctions in Chinese loanwords and transcriptions: e.g.,

My reconstructions of some Tangut rhyme clusters (R = Tangut rhyme number)

(The Grade II -u cluster rhyme is R4 rather than R2. This unusual order may have been influenced by the fourth position of the corresponding rhyme in Chinese rhyme dictionaries.)

Since Tangut was in close contact with Chinese, it would not be surprising if its phonology converged with that of Chinese. This is not to say that the two languages had identical phonologies. Tangut had unique features: e.g., tense and retroflex vowels absent from Chinese.

My interpretation of the four grades of Tangut is based on the four grades of Middle Chinese. In the Yunjing (Mirror of Rhymes), Middle Chinese rhymes are grouped into sets of four by grade. These sets are reminiscent of the clusters of three to four rhymes in Tangraphic Sea.

There are several competing interpretations of Chinese grades. I more or less agree with the interpretation in Schuessler (2007) and (2009). In Schuessler's reconstruction of Chinese, an Old Chinese vowel could be 'bent' in different ways in Middle Chinese depending on factors such as 'emphasis' (pharyngealization) and medials (the presence or absence of (*-r-): e.g.,

(I have chosen *-an rhymes instead of *-a because the bending pattern of *-a is slightly complex. Old Chinese *-a became Middle Chinese *-o, not Middle Chinese *-a.)

I have proposed a similar 'bending' process in Tangut. The Middle Chinese vowels above (*a *æ *ɨa *ia) are identical to the vowels I reconstructed for Tangut R17-R20. (But most vowels in my Tangut reconstruction are not in Middle Chinese.)

Sino-Korean and Sino-Vietnamese borrowings from Middle Chinese hint at the phonetic nature of the four Chinese grades: e.g.,

Grade I is the most straightforward: SK and SV both point to *-an.

Grade II SV forms of the type gian [zaan] < *kjaan may have been Vietnamizations of southeastern Middle Chinese *kjan < *kæn. Early Middle Chinese *æ was borrowed as early SV [ɛ] in open syllables. (But I don't know of any early SV borrowings of EMC *-æn. Such borrowings should theoretically end in [ɛn].)

There is very little evidence for a Grade II *-j- in northwestern Middle Chinese (see Coblin 1991 and 1994), so I don't reconstruct a palatal medial in Tangut Grade II (unlike Gong who reconstructed -i- or Arakawa who reconstructed -y-). Tibetans only rarely transcribed Tangut Grade II with Tibetan -y-.*

Grades III and IV in SK and SV have higher vowels than Grades I and II.

Grade IV has more palatal features in SK and SV than Grade III:

- SK Grade IV always has medial [j], but SK Grade III lacks it after velars

- SV Grade IV has palatalized labials [t th z] < *pj- *phj- *mj- instead of the nonpalatalized labials [ɓ f m] < *p- *ph- *m- found in SV Grade III.

For a fuller examination of the SK and SV evidence for Chinese grades, see Pulleyblank (1984).

One huge problem with using SK and SV in an argument for Tangut grades is the fact that Korean and Vietnamese did not borrow from a northwestern Chinese dialect ancestral to the dialect known to the Tangut. It is possible that the four grades sounded very different in the northwest.

The Kan-on layer of Sino-Japanese was borrowed from northwestern Middle Chinese, but in its present state, it only tells us that Grades III and IV had higher vowels than Grades I and II: e.g.,

I don't have Coblin's (1991) and (1994) works on Tibetan transcriptions of pre-Tangut period northwestern Chinese on hand, but looking at the Tibetan transcriptions of the Thousand Character Classic (Luo 1933 as reproduced in Gong 2002: 324), the following pattern emerges:

All of the above evidence (SK, SV, SJ, Tibetan transcriptions) point to a *-j- in Chinese (and Tangut?) Grade IV, yet there is not one Tibetan transcription of Tangut with -y- for R20 -ia or R24 -iaa. Tai (2008: 209-211) lists only two Tibetan -ya transcriptions for Tangut a-rhymes:

Tib tr. nya for 2491 1na R17 (Grade I!)

Random error?

Did the transcribed dialect have ɲ- corresponding to standard n- in this syllable? The transcribed dialect could have preserved an initial lost in the standard dialect recorded in dictionaries.

Tib tr. nya for 2015 1nɨaa R21 (Grade III)

y might be an attempt to transcribe ɨ, or it could be another example of transcribed dialect ɲ- corresponding to standard n-.

Did the Tangut dialect transcribed by the Tibetans lose -i- or fuse -ia into -æ, or was -i- (or -j- or the like) also absent from the standard dialect? In other words, is my reconstruction of Tangut Grade iV wrong? Even if everything I wrote about Chinese grades is correct, there is no guarantee that the grade system remained basically unchanged in the northwest by the time the Tangut used it as a basis for their own grade system.

*A rare example of a Grade II Tangut syllable transcribed with Tibetan -y- is

Tangraph	Tibetan transcriptions	My reconstruction	Gong's reconstruction	Arakawa's reconstruction
5057 'true'	dge, dghe, Hge, se (sic)	1ɣɛ	1ɣiej	1qye
	dgye
	Hgi

Each of the six different transcriptions of 'true' occurs only once (Tai 2008: 214).

The odd transcription se may be due to confusion with the similar tangraph

4751 1se 'pure'

transcribed in Tibetan as se or gseH.

Arakawa (1999) interpreted the preinitials d- and g- in Tibetan transcriptions as indicators of Tangut level tone.

The correspondence between Arakawa's q- and the g in most of the Tibetan transcriptions reminds me of the shift of q to g in Libyan Arabic.