08.3.22.15:44: CRAMPED BY CODAS

In "Rebending Tangut Grade I High Vowels", I reconstructed single Grade I vowels before codas instead of bent diphthongs:

-ej (not -eij or -əij or -əɨj)

-ew (not -eiw or -əiw or -əɨw)

-ow (no -ouw)

Shortly after uploading last night's post, I realized that Khmer short vowels also developed into single vowels instead of bent diphthongs before codas:

It's as if Khmer short vowels didn't have enough space to bend before a coda. (Or that they weren't long enough to curve.)

However, Khmer long vowels did bend before codas, whereas I prefer to reconstruct a straight vowel for -eew, the only Tangut Grade I long vowel rhyme with a coda:

The distribution of vowels before codas is very skewed in Grade I:

(-əu, -əi, -əɨ could either be treated as short diphthongs or as short vowel + glide sequences -əw, -əj, -əɰ. Since they are grouped with codaless Grade III -u, -i, -ɨ in Tangraphic Sea, I prefer to interpret them as codaless.)

I suspect that

pre-Tangut lacked -uw and -ij which would have been difficult to distinguish from -uu and -ii

pre-Tangut -u(u)j, -i(i)w, -a(a)G, -ɨ(ɨ)G, -o(o)j, -eej, -oow merged with other rhymes

I don't know what happened to labial vowel + -j sequences (if they even existed at all), but here's what I think might have happened to the other sequences:

(Could *-o(o)j have become -we(e)? Cf. Korean -we < *-oj, still spelled ㅚ -oi.)

If these mergers are correct, there should be word families with alternations such as

-a(a) ~ -ej, -ee, -ow, -oo

-ɨ(ɨ) ~ -ej, -ee, -ew, -eew

-ej ~ -ee

-ow ~ -oo

from earlier

*-a(a) ~ *-aj, *-aaj, *-aw, *-aaw

*-ɨ(ɨ) ~ *-ɨj, *-ɨɨj, *-ɨw, *-ɨɨw

*-ej ~ *-eej

*-ow ~ *-oow

08.3.21.3:45: REBENDING TANGUT GRADE I HIGH VOWELS

Two nights ago, I proposed that Grade I high vowels bent into mid vowel + high vowel diphthongs:

pre-Tangut *u > ou

pre-Tangut *i > ei

pre-Tangut *ɨ > əɨ

This would explain why Tibetan transcriptions for Grade I high vowels sometimes contain mid vowels.

However, Grade I high vowels also occur before the glides -j and -w. Substituting the diphthongs above for the high vowels in Gong Hwang-cherng's Grade I reconstructions would result in unlikely contrasts and awkward vowel-glide sequences: e.g.,

Gong's R1/R4 -u > my -ou : Gong's R56 -ow > my -ouw (!)

To avoid this problem, I have altered the bending pattern of Grade I high vowels

pre-Tangut *high vowel > Tangut schwa + high vowel

(cf. Khmer *ii > əi and *ɨɨ > əɨ - but Khmer *uu > ou, not əu!)

and reinterpreted some of Gong's -VG sequences:

I am not sure whether groups VIII and XI ended in glides or nasalized vowels.

I suspect that groups VIII, IX, and XI have complex origins. Vowels before glides may be from unbent nonhigh vowels

*-aj, *-ej > -ej

*-aw, *-ow > -ow

or bent diphthongs that monophthongized before glides (or nasal codas in VIII?):

*-ɨj/N > *-əɨj/N (*-ɨəj/N?) > -ej (or -eĩ̃ or -ẽ?)

*-iw > *-əiw (*-iəw?) > -ew

*-ɨw > *-əɨw (*-ɨəw?) > -ew

The reconstructions in parentheses incorporate a variation of the achromatic bridge hypothesis to avoid implausible closing diphthongs before glides.

08.3.20.3:16: WHY I BENT R29 AND R91

In Gong Hwang-cherng's Tangut reconstruction, there are two open-vowel rhymes with Grade II -iə:

R29 -iə

R91 -iə̣

The absence of -iə̣ with a tense vowel may be an accidental gap.

Last night, I initially changed Gong's -iə to -ʌ^ɣ with a lowered 'straight' velarized vowel to match my other Grade II vowels. However, R29 was transcribed with Tibetan -i rather than Tibetan -a. Thus I changed R29 to a diphthong -ʌ^ɣɨ (or -ʌ^ɣɯ?) with a high element corresponding to Tibetan -i.

R91, the tense counterpart of R29, would therefore be -ʌ̣^ɣɨ (or -ʌ̣^ɣɯ?).

The Grade I counterparts of R29 and R91 may also have had an unrounded achromatic high vowel corresponding to -i and -u in Tibetan transcription:

The absence of -əə(ɨ)^r with a long retroflex vowel may be an accidental gap.

08.3.19.1:19: A BENT INTERPRETATION OF THE TANGUT GRADE SYSTEM

Gong Hwang-cherng interpreted the three grades of Tangut in terms of medials:

I've recently substituted velarization for his medial -i-:

My last post about bending made me wonder if the Tangut grade system was like Schuessler's three-way distinction in Late Old Chinese:

There is one essential difference: Tangut Grade II did not originate from *-r-. I think that both Tangut and Chinese developed velarized vowels (Grade II in both languages) from different sources. (Pre-Tangut *r conditioned vowel retroflexion in all three Tangut grades, so Tangut Grade II could not be from *-r-.) Perhaps Tangut Grade II could be called 'VS' with V for 'velarized'.

The table below lists what Schuessler-style Tangut vocalism might have been like:

Grade I syllables have low vowels and partly lowered high vowels.

Grade II syllables have (velarized or pharyngealized?) lowered vowels without bending in most cases.

Grade III syllables have high vowels and partly raised nonhigh vowels.

Long, tense, and retroflex vowels have been omitted for simplicity. Bent vowels are in bold.

Reconstructing Grade III without -j- fits the Tibetan transcription evidence better. Gong's Grade III -j- often corresponds to zero in Tibetan transcriptions: e.g.,

Next: Why couldn't I simply reconstruct the Grade II achromatic mid vowel as unbent ʌ^ɣ?

And after that: Summing up the sources of the three grades.

08.3.18.23:59: THE VOICELESS AND THE EMPHATIC

When I first saw Axel Schuessler's Late Old Chinese vowel bending hypothesis back in 2000, I thought, 'that looks like Khmer!'

In both languages, one class of initials (voiceless in Khmer and emphatic in Old Chinese) conditioned the downward bending of nonlow vowels:

(The Khmer notation is based on Diffloth 1992's as presented in Schiller [1996: 471]. The OC vowels are mine, though the bending principles are Schuessler's.)

However, the parallels are almost nonexistent when other initial types are examined. Nonemphatic initials in Old Chinese conditioned the upward bending of nonhigh vowels, whereas Khmer vowels remained unchanged after voiced initials with one exception: aa raised to iə.

(I've long assumed that French romanizations of Khmer with ea - e.g.,

Siem Reap for សៀមរាប SIEM RAAP, now [siəm riəp]

reflected an earlier mid-low diphthong that later became high-mid iə.)

Moreover, Khmer has not yet developed anything like the velarized (or retroflex?) rhymes of Early Middle Chinese:

08.3.17.23:36: DOUBLE GLIDE CODAS?

Does any language allow codas like -wj (as opposed to -ɥ which is both labial and palatal)?

I nearly posted the following table last night:

However, -ʊ^ʕj is not far from -wj, which I have never seen anywhere. So I proposed a different sequence instead.

But then today I realized that in Khmer, *oo had warped to ao, even before -j: e.g., ʔaoj 'give' which is still spelled

ឲ្យ/អោយ ʔOOY

(I've seen the second spelling in recent webpages such as this online French-Khmer dictionary. The first spelling is the one I learned from Huffman's Khmer writing textbook.)

This -aoj is not unlike *-ɑʊ^ʕj. Nonetheless,

1. -aoj seems very unusual.* When I was studying Khmer, I struggled not to pronounce it as two syllables: [ʔa wi]. (It looks like it should be pronounced like Jpn aoi 'blue', but it isn't.)

2. Metathesis is unusual (though not very unusual).

My goal is to reconstruct the most likely course of development. Hence my current solution avoids both an -aoj-like rhyme and metathesis.

*Does any other Mon-Khmer language have -aoj?

ADDENDUM: In Khmer, OO can be written as a combination of E and AA around a consonant letter:

EʔAAY = ʔOOY [ʔaoj]

(This does not mean that Khmer *oo is from an earlier *eaa!)

However, Thai OO has its own vowel symbol โ. Moreover, in Thai, E + AA represents aw:

เอา EʔAA = ʔaw

Is the Thai spelling of aw simplified from Khmer E + AA' with a modified version of the AA' letter: e.g.,

អៅ EʔAA' = ʔAU [ʔaw]

And could Thai โ somehow derive from one of the Khmer independent letters for OO?

โ < ឱ?

(Why does Khmer have two independent letters for OO? See the first spelling of ʔaoj for the other one which resembles Devanagari उ u. Yes, I know Khmer script is not from Devanagari. I'm so sick of the belief that all Indic scripts 'come from' Devanagari.)

Stranger still, the contrast between E + AA and E + AA' in Burmese is tonal:

The front counterpart of Burmese [ɔ] has yet another spelling pattern:

AY = [ɛ] (low tone)

AI = [ɛ] (high tone)

I don't understand why high tone isn't written with visarga (-Ḥ), which is the symbol used elsewhere in the script (implying that high tone was from *breathiness):

E + AA = [ɔ] (low tone)

E + AAḤ = [ɔ] (high tone)

AI = [ɛ] (low tone)

AIḤ = [ɛ] (high tone)

I doubt that the script can be used as evidence for breathiness somehow developing in earlier *o and *ai but not in *au and *aj.

There was, apparently, no clear analogue in Mon [whose script was the basis for the Burmese script] to the opposition between high and low tones and for some six centuries the two were not consistently distinguished in the orthography. In the modern script, the lower-mid vowel signs ... are intrisically high-toned, with additional strokes ('killed-Y' in one case, the killer alone [represented in my transliteration by an apostrophe]- originally a superscript killed-W - in the other) changing them to low.

3.18.3:25: Here's what might have happened:

Indic alphabets have symbols for

the mid vowels EE OO (originally *ai *au)

the diphthong AI AU (originally *aai *aau)

Postinscriptional Burmese had

[e] (< inscriptional IY = [ɪj] or [ɨj])

[ɛ] (< inscriptional AAY = [aaj]

[ɔ] (< inscriptional OOV = [ɔɔw]

The two lower-mid vowels [ɛ] and [ɔ] didn't really match any of the four Indic sounds. So they were written in a variety of ways: AI ~ AY, OO ~ AU (< OO with what was once a superscript 'killed' V [= Wheatley's W]). Eventually, the variants for each vowel were assigned different inherent tone values.

Even if my scenario were correct, I'm still puzzled by the assignment of high tone to the more basic symbols AI and OO. All other long vowel and 'diphthong'** symbols have inherent low tone: AA II UU EE UI.

**UI is [o] in modern Burmese, but in inscriptions (Pulleyblank 1963: 217), it was UIV / IV / EIV = ?*[ɨw], the back counterpart of IY = [ɪj] or [ɨj]. Perhaps the digraph UI without V reflects an intermediate stage ?*[əɰ] without a labial:

3.30.16:23: But according to Matisoff (2003: 179), modern Burmese UI corresponds to UV (his UW) in inscriptions.  Who's right?  I haven't looked at inscriptional Burmese since 1994, when I compiled a basic vocabulary database of Lolo-Burmese languages that I didn't bring to Xin Zexi.

08.3.16.23:59: 東冬鐘江 EASTERN WINTER ON BELL RIVER (PART 6)

I know most (all?) readers want me to get back to Tangut, but I'm still struggling with these Chinese rhymes. The following charts show how they might have developed (ignoring rhymes preceded by *r):

Stage 1: Simple raised/lowered allophony. No bending.

Stage 1': What if OC labial vowel + -ŋ rhymes were achromatic vowel + -wŋ͡m rhymes (cf. modern Vietnamese)?

Stage 1'': Similar to 1', but with back instead of central vowels for nonemphatic rhymes:

The following stages proceed from Stage 1 rather than 1' or 1'' which might describe the phonetics of some OC dialects.

Stage 2: Late Old Chinese. Bending. Start of emphatic *u partly 'pushed down'. Start of nonemphatic *o 'pulled up'.

Stage 3: Early Middle Chinese. Loss of emphasis. Breaking of old simple labial vowels.

The development of the above rhymes is similar to the development of rhymes ending in labial vowels up to stage 2:

Stage 1: Simple raised/lowered allophony. No bending.

(For simplicity, I am excluding OC *-iw and *-iw from 幽 'dark'.)

Stage 2: Late Old Chinese. Bending. Start of emphatic *u partly 'pushed down'. Start of nonemphatic *o 'pulled up'.

However, Stage 3 of OC *-u bears little resemblance to stage 3 of OC *-ung:

OC labial vowel + dental rhymes had yet another pattern of development. All of them became *-w/uVn rhymes in MC:

MC permitted *w before zero and velars but not before *-n. Therefore metathesis (in bold) may have occurred between OC and MC to avoid *-wn clusters:

Perhaps *u and *o partially delabialized to form achromatic bridges between labial vowels and acute codas:

3.17.00:52: The achromatic bridge hypothesis also accounts for the partial delabialization of *u and *o before the acute coda*-j (which may be from an even earlier *-l in some cases):

The loss of *-j after *-ɑ left a gap in the system which was partly filled by *-ɑɕ:

There was no *-ɑj (or creaky-vowelled *-ɑ̰j) in MC because *-ɕ had caused the preceding vowel to become breathy before leniting to *-j.