19.5.9.18:39: TANGUT VOWELS V. 190509

Writing about the Tangut transcription of Sanskrit trailokya got me thinking about the phonetic values of Tangut vowels again. Here's my own take on the four grades influenced by Gong Xun's ideas. Only basic vowels are listed in Tangraphic Sea order, so there are no nasalized, tense, or retroflex vowels. I still have no idea what the distinction that I indicate as -' was.

I write the basic vowel /ə/ as an easy-to-type y in my abstract notation.

The grades:

I. Pharyngealized; lowered and/or backed

Pharyngealization is carried over from Jerry Norman's proposal for the Old Chinese source of Middle Chinese Grade I.

The lowered and backed allophones are similar to Arabic vowel allophones after 'emphatics' as described in Kaye (2009: 565).

Syllables with 'lower' series vowels (*a *e *o) automatically developed pharygealization unless this was blocked by a preceding 'higher' series presyllabic vowel (*ɯ):

*Ca > *Cɑˁ but *CɯCa > *Ca

Conversely, a 'lower' series presyllable vowel (*ʌ) triggered pharyngealization in a following 'higher' series vowel (*ə *i *u):

*CʌCi > *Cɪˁ

Low /a/ cannot be lowered any further, so it is only backed.

Front /e/ is retracted to [ɛˁ]. The underlining indicates retraction. [ɪˁ] without underlining is already backer than front [i], so I do not underline it.

Back /u o/ cannot be backed any further, so they are only lowered.

II. Uvularized; lowered and/or backed

Medial /r/ in pre-Tangut pharyngealized syllables became uvular [ʁ]. This uvular medial was lost, but it colored the following vowel: e.g.,

pre-Tangut *pʰrat >  *pʰʁɑˁt >  pʰɑʶ = 2475 𗧑 1pha2 'to break in two'

Note that Gong Xun reconstructs uvularization in both Grades I and II:

In his system, a medial -ʕ- distinguishes Grade II from Grade I.

Gong has a single unmarked category corresponding to my Grade III and IV. Although it is true that Grades III and IV are in nearly complementary distribution -

Grade III: after v- (a labiovelar glide?), retroflexes, (velarized?) l-

Grade IV: elsewhere

- I still want to work out how they sounded to distinguish between the few minimal pairs that existed.

Syllables with 'higher' series vowels (*ə *i *u) automatically became Grade III or IV dependng on the preceding initial unless there was a preceding 'lower' series presyllabic vowel (*ʌ):

*Ci > *Ci but *CʌCi > *Cɪˁ

Conversely, a 'higher' series presyllable vowel (*ɯ) triggered Grade III or IV in a following 'lower' series vowel (*a *e *o):

*CɯCa > *Cæ

III. Higher and centralized

Grade III was less palatal and more velar than IV. Its palatal vowels /ɰi ɰe/ had velar glides /ɰ/ to distinguish them from the pure palatal vowels [i e] of Grade IV. The sequence /wɰ/ surfaced as [w].

IV. Higher and fronted

Grade IV was more palatal than III. It had front vowels [æ y ø] corresponding to the central or back vowels of other grades.

An exception to that pattern is [ɨ] which, though not front, was still fronter than its back counterparts in other grades.

The Grade IV equivalent of labiovelar [w] in other grades was labiopalatal [ɥ].

Examples

Unattested syllables are in parentheses.

I regard [q] as the Grade I and II allophone of /k/.

Are the gaps in the table random or systematic? Any theory of grades should be able to answer that question.

My hypotheses above regarding the origin of the grades predict that

- lower-vowel syllables should tend to have Grades I and II

- higher-vowel syllables should tend to have Grades III and IV

if *CV monosyllables outnumbered *CV̆CV sesquisyllables.

And above we see

- there is no 1ka3 or 1kwa4

- there is no 1k(w)i1

which fits my predictions.

The absence of 1kwi3 is also not surprising, since ki-syllables should tend to have Grade IV, not Grade III. k- does not belong to the subset of initials associated with Grade III: v-, retroflexes, and l-. There are only three known k-syllables with Grade III, and two of them happen to be in the table: 1kwa3 and 1ki3. The third is 1ka'3 which must have been something like [ka] plus whatever feature was represented by -'.

19.5.8.23:58: PITTAYAPORN'S PROTO-TAI *-ɲ

One of the innovations of Pittayawat Pittayaporn's (2009) PhD dissertation The Phonology of Proto-Tai is his reconstruction of a Proto-Tai final palatal nasal:

Since it has been established that PT [Proto-Tai] allows palatal consonants in the coda [i.e., *-c¹ and *-j], one would also expect to find the palatal nasal occurring in coda position. Although the reconstruction of PT *-c is unequivocal, there is rather little evidence for final *-ɲ. The only potential case I have identified so far is ‘to eat’, which is reflected as /kinA1/ in all SWT [Southwestern Tai] varieties but as /kɯnA1/ in NT dialects [Northern Tai] like Wuming and Yay. We can speculate that the PT form for ‘to eat’ was *kɯɲ A but the vowel was fronted so that the PSWT [Proto-Southwestern-Tai] form for this etyma was *kin A. Therefore, I tentaitively hypothesize that PT had both *-c and *-ɲ.

Let's look at the 'eating' problem from a subgrouping perspective. Unlike Li Fang-Kuei whose classic model of the Tai family had only three branches (Northern, Central, and Southwestern), Pittayaporn (2009: 298) proposed four branches on the basis of clusters of innovations:

A. Most Tai languages

B. Ningming

C. Chongzuo and Shangsi

D. All of Li's Northern Tai languages (such as the displaced Saek in the southeast) plus some of his Central Tai languages

Wikipedia has a clickable version of Pittayaporn's tree.

What is 'to eat' in the four branches?

A. Siamese kin A1

B. Ningming ken A1 (not in Pittayaporn 2009; found in Hudak 2008: 121)

C. Shangsi kɤn A1

D. Yay kɯn A1 but Saek kin A1

There are two types of words for 'to eat': ones with front vowels (A, B, Saek) and ones with back vowels (C, Yay). All end in -n.

Given that -in words are found in both A and D (Saek), let's suppose those branches independently preserve a proto-rhyme *-in.

By analogy, any -in words in Siamese and Saek should respectively end in -en in Ningming, -ɤn in Shangsi, and Yay -ɯn unless complicated by other factors. But is this really the case? Compare the forms for 'to eat' with those for Pittayaporn's *lin A 'water pipe':

A. Siamese lin A2

B. Ningming (no cognate in Pittayaporn or Hudak)

C. Shangsi lin A2 (not ˟lɤn A2)

D. Yay and Saek lin A2 (not Yay ˟lɯn A2)

It is true that in the modern languages, 'to eat' and 'water pipe' belong to different tonal categories (A1 and A2) conditioned by the initials (*voiceless > 1, *voiced > 2). So one could try to salvage the *-in reconstruction of 'to eat' by claiming that *-i- changed before *-n in tone A1 syllables in Ningming, Shangsi, and Yay. But why would, for instance, tone A1 cause *-i- to lower and back to -ɤ- in Shangsi?

Might the original rhyme of 'to eat' be preserved in Shangsi - or Ningming or Yay? No, because the rhymes of 'to eat' in those languages do not otherwise correspond to -in in Siamese and Saek. Here are all the relevant correspondence sets, including those I already mentioned:

Pittayaporn's solution is ingenious:

- It accounts for the front vowel of Siamese and Saek as the result of feature transfer: the palatality of *-ɲ shifted to the vowel *-ɯ-, causing it to independently front to -i- in two distant branches of Tai (assuming the Saek word isn't a loan).

- The shift of *-ɯɲ to -Vn in all branches fits a trend against -Vɲ rhymes in Southeast Asian languages. Khmer does have a high neutral vowel-palatal nasal sequence /ɨɲ/ (e.g., in <beñ> /pɨɲ/, the Penh of Phnom Penh), but it is exceptional. Burmese once had /-aɲ/ as its sole /-ɲ/ rhyme, and Vietnamese only has /-aɲ -eɲ -iɲ/.

There are, however, two problems with his *-ɲ:

Second, there is no external support for *-ɲ either within Kra-Dai or beyond it. Although Norquest (2015) reconstructs *-ɲ in Proto-Hlai, he does not reconstruct *-ɲ in Proto-Qi³ *kʰən (< my pre-Hlai *kən) 'to eat'. Blust's Proto-Austronesian *kaen [kaən] - somehow related to the Proto-Tai and Proto-Qi words - ends in *-n, not *-ɲ. The *k-word for 'to eat' probably goes back to Proto-Kra-Dai and is either inherited or borrowed from some Austronesian-type language⁴. Does Proto-Tai preserve a *-ɲ lost elsewhere?

¹The reconstruction of a Proto-Tai final palatal stop is another innovation of Pittayaporn (2009). Although no attested Tai language has /c/, reconstructing *-c accounts for correspondence set 2 in the following table:

Sets 1-3 are from Pittayaporn (2009: 211-212). Set 4 is based on the forms for 'liver'.

Saek is an aberrant Tai language which "shows many peculiarities that cannot be reconciled within the conventional model of PT [Proto-Tai] phonology" (Pittayaporn 2009: 14).

The Be languages are generally thought to be close relatives of Tai. See Chen (2018: 18) for the placement of Be within four different proposed Kra-Dai language trees. Ostapirat has changed his mind over time; in 2000 he viewed Be as a sister of Tai but in 2015 he viewed Be as a primary branch of Kra-Dai, and as of 2017 he viewed Be as a sister of a Tai-Kam-Sui subgroup.

²Proto-Tai *ˀjen A 'tendon' has a different set of rhyme correspondences that may be conditioned by a palatal initial absent from Proto-Tai *ʰmen C 'porcupine'.

³Proto-Qi is my term for the common ancestor of the Qi subgroup of Hlai. Norquest reconstructs it but has no term for it. Other early Hlai languages had unrelated words for 'to eat'. As only the Proto-Qi word is cognate to the Proto-Tai word, it seems that pre-Hlai must have inherited the word from Proto-Kra-Dai, but only one dialect of Proto-Hlai (i.e., Proto-Qi) retained it whereas other dialects of Proto-Hlai replaced it with innovations of unknown origin: *C-ləːk in Proto-Run and *C-luːɦ elsewhere.

⁴I am deliberately vague here because I do not know if Proto-Kra-Dai is descended from Proto-Austronesian or is a sister to it (i.e., a descendant of Proto-Austro-Dai, if I may modernize Benedict's term 'Austro-Tai'). Or if there is no genetic relationship between Kra-Dai and Austronesian, if Proto-Kra-Dai borrowed from Proto-Austronesian, an ancestor of Proto-Austronesian, or a descendant of Proto-Austronesian.