I can't even remember how many series I've started and never finished this year alone. Yes, I have a short attention span. I also have excessive ambition. I keep picking - or stumbling on - extremely complex topics that I can't tackle in a day. What I think are bite-size pieces just keep growing. Let's see how small I can keep this.

Pittayawat Pittayaporn's (2009: 149) reconstruction of Proto-Tai has only one cluster with an implosive: *ɓl-. That initially struck me as unusual because I am accustomed to Southeast Asian languages with complex onsets but no clusters with implosives as first elements: e.g., Pyu, Mon, and Khmer. But then I remembered that Middle Vietnamese had bl- /ɓl/.

I don't know of any modern Vietnamese dialects that have labial reflexes of Middle Vietnamese /ɓl/, but I know almost nothing about Vietnamese dialects. If not for Middle Vietnamese, I wouldn't be able to reconstruct such a reflex.

On the other hand, there are labial reflexes in modern Tai languages that make the reconstruction of Proto-Tai *ɓl- possible. Below I cite reflexes of Proto-Tai *ɓlɯən A 'moon' mostly from Pittayaporn (2009) and Hudak (2008) plus 扶绥 Fusui and Shan forms from the Austronesian Basic Vocabulary Database):

1. *ɓl-type reflexes

1a. /bl-/: Saek /bliən A1/ 'moon' (only Saek retains *-l-)

1b. /bj-/: (Bao Yen /bjɔːk DL1/ < Proto-Tai *ɓloːk D 'flower')

1c. /mj-/ (mentioned as a reflex in Pittayaporn 2009: 150; unable to find examples)

cf. Vietnamese *ɓ- > m- (but the reverse may have occurred in Pyu!)

2. *ɓ-type reflexes

2a. /b-/: Shangsi /bun A1/

cf. Shangsi /boy A1/ < Proto-Tai *ɓaɰ A 'leaf'

is Shangsi /oy/ [oj] or [oɥ]?

2b. /m-/: Fusui /mɯːn A1/

cf. Fusui /mɯj A1/ < Proto-Tai *ɓaɰ A 'leaf'

(Shan /mɔk DL1/ < Proto-Tai *ɓloːk D 'flower'; see 3c below for the Shan word for 'moon')

3. *ɗ-type reflexes (*ɗ- combines the implosion of *ɓ- with the place of articulation of *-l-)

3a. /ɗ-/: Wuming /ɗɯan A1/

cf. Wuming /ɗoj A1/ < Proto-Tai *ɗɤj A 'good'

3b. /d-/: Thai /dɯən A1/

cf. Thai /diː A1/ < Proto-Tai *ɗɤj A 'good'

3c. /l-/: Shan /lɤn A1/ (but cf. 'flower' in 2b above!)

cf. Shan /li A1/ [liː] < Proto-Tai *ɗɤj A 'good'

3d. /n-/: Po-ai /nɯːn A1/

cf. Po-ai /niː A1/ < Proto-Tai *ɗɤj A 'good'

cf. Vietnamese *ɗ- > n- (no evidence for the reverse in Pyu which had no /ɗ/)

I started writing about Proto-Tai *ɓlɯən A 'moon', but I'm going to move that to a post of its own. This post started on the 17th and has taken me five days to finish. I don't want peripheral material to hold it back any longer. MEITEI PHONOLOGY

Meitei is an isolate within the Sino-Tibetan family. Its speakers constitute the majority of the population in the Indian state of Manipur on the border with Burma.

I originally wanted to write a post with the characters of the Meitei script reorganized for my own convenience in the standard Indic order, but KompoZer doesn't support the Meitei range of Unicode for some reason, so I'm going to write about Meitei phonology instead based on Chelliah (2016).

Consonant phonemes

native initial (ideophones only)
native medial
borrowed initial/medial
borrowed only?




r l



I am not sure whether voiced aspirates appear in native words. In the Wikipedia article on the Meitei native religion of Sanamahism, some names of native deities are spelled with -dhou after u (in the recurring name element Ebudhou),  but others have -thou after -ng. That suggests native voiceless aspirates might have voiced in intervocalic position.

I do not know the origin of voiced obstruents in medial position in native words. Two scenarios with hypothetical examples:

Scenario 1

Initial devoicing but retention of voiced series in medial position:

*gaka > kaka

*kaga > kaga

Scenario 2

Medial voicing of voiceless series; development of new medial voiceless series from something else:

*kaka > kaga

*kaXa > kaka

Although Meitei borrowed voiced aspirates from Indo-Aryan, it did not borrow retroflex consonants.

I can't find a character in the Meitei script for /cʰ/. Are /c/ and /cʰ/ both written with U+ABC6 <c>?

Vowel phonemes

Meitei has a six-vowel system identical to that of Old Chinese and pre-Tangut:


Pyu has a similar seven-vowel system with an additional distinction between front /ä/ and nonfront (back?) /a/.

It would be asking too much for Meitei vowels to precisely line up with those of Old Chinese, pre-Tangut, or Pyu. Nonetheless it is nice to see these mostly straightforward correspondences for the Meitei numerals at Omniglot.

Old Chinese
/p.lä/ < *-e
*C.ŋaʔ *P.ŋa /pə.ŋa/

Needless to say, numerals alone are insufficient evidence for a genetic relationship, as they can be borrowed: e.g., Proto-Tai has *saːm A 'three', *siː B 'four', *haː C 'five', and *krok D 'six' from Chinese (Pittayaporn 2009). (*soːŋ A  'two' is from Late Old Chinese 雙 *ʂɔŋ 'pair'.) What gives away the Chinese origin of the numerals are Chinese-internal innovations: e.g., the irregular lowering of *-u- in 'three' and the loss of *-l- in 'four'. SHARP-HEAVY, ENTERING/DEPARTING, AND B/D

A generic model for tonogenesis in the Sinosphere involves four categories of final consonants.

Category names
Hmong-Mien Kra-Dai
ngang 'even' / huyền 'dark'
平 'level'
sắc 'sharp' / nặng 'heavy'
上 'rising'
hỏi 'ask' / ngã 'fall'
去 'departing'
*nonglottal stops
sắc 'sharp' / nặng 'heavy' 入 'entering'

To demonstrate these categories, I will use hypothetical examples for simplicity.

Vietnamese has six categories. Each proto-category split in two depending on the voicing of the proto-initial consonant, and as I'd expect, all syllables with final stops developed the same tones (sắc/nặng).

*kaʔ > cá (sắc)

́*kak > các (sắc)

*gaʔ > cạ (nặng)

*gak > cạc (nặng)

In White Hmong, syllables with initial *voiced consonants and final *glottal stops developed the same tone as syllables with initial *voiceless (!) consonants and final *nonglottal stops  (Ratliff 2010: 184).

*gaʔ > kas

́*kak > kas

(-s represents a low tone.)

However, in other Sinospheric languages, syllables with final *-h and syllables with final *nonglottal stops may develop similar or identical tones:

1. In standard Mandarin, there is a weak tendency for syllables which once had final *nonglottal stops to have the same high falling tone as syllables which once had final *-h.

*kah > ku (high falling tone)

*kak > (high falling tone)

2. In Cantonese, syllables with final nonglottal stops have noncontour tones like syllables which once had final *-h.

*kah > kuː (mid level tone)

*kak > kɔːk (mid level tone)

3. Gedney (2008) has descriptions of the tone systems of 19 Tai varieties. In all of them, there is at least partial overlap between the tones of the B and D categories: e.g., in Thai, syllables of those categories almost always have the same tones:

*ka(ː)ʔ > kaː (low tone)

́*ka(ː)k > ka(ː)k (low tone)

*ga(ː)ʔ > kʰaː (falling tone)

*gaːk > kʰaːk (falling tone)

but *gak > kʰak (high tone!)

I suspect there was no distinction between */Vʔ/ and */Vːʔ/.

There is a strong tendency for B tones to overlap with D tones with long vowels. Conversely, D tones with short vowels (e.g., *gak in the hypothetical Thai example), tend to go their own way.

Today I realized what might have led to similar tones in the B and D categories (and their equivalents outside Tai). Contrast these two scenarios:

Scenario 1

Stage 1
Stage 2
V + tone 1
+ tone 2
V + tone 3
Vk + tone 2

Scenario 2

Stage 1
Stage 2
*V̰ʔ *Vh
Stage 3
*V̰ *Vh
Stage 4
V + tone 1
+ tone 2
V + tone 3
V(k) + tone 3

Scenario 1 is straightforward; all syllables with *stops develop the same tones. This is what happened in Vietnamese.

Scenario 2 is more complicated.

In stage 2, final *glottal stops condition *creaky voice which is nonphonemic (= predictable and hence nondistinctive).

In stage 3, final glottal stops are lost, and *creaky voice becomes phonemic (= unpredictable and hence distinctive). *V̰ no longer ends in a stop, so it loses its phonetic resemblance to *Vk which still ends in a stop.

In stage 4, modal and creaky-voiced syllables develop tones 1 and 2, whereas syllables ending in obstruents develop tone 3.

Pittayaporn (2009) posits a third scenario for Proto-Tai to which I add a top row (he does not reconstruct segmental sources for Proto-Tai tones).

Scenario 3

Stage 1
Stage 2
*Vʔ *V̰ *k
Stage 3
V + tone 1
+ tone 2
V + tone 3
Vk + tone 3

What I don't understand is why *Vh conditions creakiness (a ʔ-quality)  rather than breathiness (an h-quality).

Pittayaporn's Proto-Tai (stage 2) is somewhat like modern Burmese which distinguishes between creaky vowels and vowels followed by glottal stops:

Stage 1
Stage 2
*V̰ *V̤ *Vk
Stage 3
V + tone 1
  + (tone 2)
V + tone 3
(tone 4)

I could argue that Burmese really only has two tones, low (tone 1) and high (tone 3); it is creakiness and a final glottal stop that distinguish the other two syllable types.

The difference between Burmese and Pittayaporn's Proto-Tai is that creakiness in the former has a more straightforward source (*-ʔ) than in the latter (*-h).

Here is one more scenario that mixes elements of 2 and 3 via a chain shift in Proto-Tai:

Scenario 4

Stage 1
Stage 2
*V̰ *Vʔ *Vk
Stage 3
*V + tone A
*V + tone C
*V̰ *k
Stage 4
V + tone A
+ tone C
V + tone B
Vk + tone D

Stage 1 has no phonemic phonation or tones.

Stage 2 has a distinction between modal and creaky phonation. The latter is from *glottal stop. A new glottal stop from *-h takes the place of that lost glottal stop:

*-Vh > *-Vʔ > *-V̰

Stage 3 continues the chain shift of stage 2:

*-Vʔ > *-V̰ > *-V + tone C

The result is a system resembling that of modern Burmese.

Stage 4 is fully tonal.

Scenario 4 cannot account for Thai in which tone C is still glottalized after (formerly) *voiced initials.

Tangut Yinchuan font copyright © Prof. 景永时 Jing Yongshi
Tangut character image fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2018 Amritavision