I compiled the table below using data from Shintani (1991: 2, 6-7):

Correspondence Common Tai Zhuang Be Baoding Hlai Jiamao Hlai
ɓ1 ɓ p f, ɓ ɓ ph
ɓ2 ɓ ɓ
ɓ3 ɓ p
ɓ4 v f
ɓ5 f ?
ɓ6 v
ɓ7 x kh
ɗ1 ɗ t h ɗ t, ɗ
ɗ2 ɓ (sic!) l t
ɗ3 ɗ ?
ɗ4 v
ɗ5 ʔj
ɗ6 j ts, z

I suspect the basic pattern is exemplified by correspodences ɓ4 and ɗ3:

Common Tai Zhuang Be Baoding Hlai Jiamao Hlai
ɓ ɓ v ɓ (nonimplosive voiceless labials)
ɗ ɗ l ɗ (nonimplosive voiceless dental[s?])

Comments on nonbasic correspondences:

ɓ1 and ɓ2 involve Chinese loanwords with *p- (板 'board' and 扁 'flat'). Common Tai has ɓ-, perhaps because of a glottal prefix: *ʔ-p-. Zhuang may have reborrowed the words from Chinese, replacing the loans inherited from Common Tai with ɓ-.

Be has f- and ɓ- instead of the expected v- in ɓ1-ɓ3. f- could reflect a voiceless prefix added to v-. ɓ- in Be (ɓ2-ɓ3) and Jiamao (ɓ2) could reflect borrowings from other Hlai languages or from Hainanese which shifted *p- to ɓ-.

Jiamao ph- in ɓ1 cannot be explained in terms of Chinese (unless early Hainanese had an unexpected aspirate in this word before *ph- became *f-), so it may reflect a Jiamao-internal prefix that conditioned aspiration: e.g., *k-p- > ph-. Another possibility is that Hlai languages preserve distinctions lost in Common Tai: e.g., *ʔf- remained a fricative in Hlai but merged with a stop ɓ- in Common Tai.

Baoding v- instead of ɓ- in ɓ6 looks like the Be initial and could be a borrowing from Be (but is that geographically plausible?) or the result of Baoding-internal lenition after a prefix: *CV-ɓ- > v-.

Baoding f- in ɓ5 could have developed from this secondary v- plus a voiceless prefix: *k-v- > f-.

I'm not sure whether the Be and Baoding words for 'light (not heavy)' in ɓ7 are really cognate with the Common Tai and Zhuang words, but if they are, perhaps their common ancestor had a cluster initial like *ɓk-. Cf. ɗ2. The Proto-Tai form is *ɓəu in Li Fang-kuei's system, but the Proto-Kra-Dai initial could have had a more complex initial.

ɗ1 involves a loanword from Chinese with earlier *d- (銅 'copper'). The implosion could have been conditioned by a glottal prefix: *ʔ-d- > ɗ- Zhuang t- may reflect a reborrowing from a Chinese form with a devoiced initial. Jiamao t- is from *ɗ-, and the Jiamao ɗ-form could be a borrowing from another Hlai language or Hainanese which shifted *d- to ɗ- (via a *t-stage). Be h- might be a borrowing from a Hainanese form with h- < *th- < *d- (see here).

ɗ2 has a Proto-Tai initial cluster *ɓl- or *ɓr- in Li Fang-kuei's system, and such a cluster could be projected up to the Proto-Kra-Dai level to account for the labial in Zhuang on the one hand and the alveolars in the others.

The Baoding nonimplosive initials in ɗ4-ɗ6 and Be j- in ɗ6 could reflect lost prefixes, cluster initials, or other distinctions that were lost in Common Tai and Zhuang (or Kam-Tai as a whole?). Baoding v- in ɗ4 is really strange since it's not even alveolar.

I'd love to see Matisoff's (1988) "Proto-Hlai initials and tones: a first approximation" and Ostapirat's (2004) "Proto-Hlai Sound System and Lexicons" in the volume honoring Hwang Cherng-gong's 70th birthday. (All roads do eventually lead back to Tangut!) I am guessing in near-total darkness here, and I wouldn't be surprised if I am mostly wrong. IMPLOSIVES ON HAINAN: INNOVATIONS AND RETENTIONS

According to Tadahiko LA* Shintani's "Preglottalized Consonants in the Languages of Hainan Island, China", all of the languages of Hainan have the initials ɓ and ɗ. These initials may either be innovations or retentions:

Languages in which ɓ and ɗ are innovations (i.e., not carried over from the mainland)

1. Hainanese

Hainanese ɓ and ɗ are from *p and *t which remain intact in most mainland Chinese languages. (Exceptions are *p which shifted to f in many languages and Toisanese *t which became zero.)

Hainanese has developed a new t from *(t)s, but has yet to develop a new initial p. (Final p did not become an implosive.)

2. Be

Be ɓ and ɗ generally do not correspond to mainland Tai or Hlai ɓ and ɗ because they are from earlier *p and *t. Earlier ɓ and ɗ have become v and l in Be. (v is close to my guess of w on Sunday.) I suspect that the old implosives were lost before Be came to Hainan, and that Be developed new ones under the influence of its neighbors. It is less likely that Be came to Hainan with implosives, lost them to become the 'odd man out', and then regained them. Be words with ɓ and ɗ corresponding to Tai and Hlai ɓ and ɗ may be borrowings from Hlai on Hainan. (They could not have been borrowed on the mainland, because their initial and would have been shifted to v and l.)

Unlike Hainanese, Be retained original *s. At least some instances of Be t are in loans from Hainanese dating after the *(t)s > t shift in Hainanese. Shintani does not mention any native sources of Be t or the existence of a new Be p.

3. Mun (a Mien language)

Mun shifted *p and *t to ɓ and ɗ and developed a new t from *s (via θ). As recently as 1926, Mun still had no implosives and *t had become θ.

Language in which ɓ and ɗ are retentions (i.e., carried over from the mainland)

Huihui (an Austronesian language)

Both Huihui and mainland Chamic languages have ɓ and ɗ. Such sounds cannot be constructed in early stages of Austronesian. Perhaps early Chamic developed them under the influence of neighboring languages like Vietnamese and the ancestors of the Huihui took these implosives with them to Hainan.

Languages in which ɓ and ɗ are a mix of innovations and retentions

Jiamao Hlai

Jiamao ɓ and ɗ do not always correspond to ɓ and ɗ in other Hlai languages and mainland Tai. The correspondence patterns are complex and may be due to a mix of Jiamao innovations and intra-Hlai borrowing.

Other Hlai languages

Non-Jiamao Hlai ɓ and ɗ do not necessarily correspond to mainland Tai ɓ and ɗ. Perhaps some non-Jiamao implosives are innovations.

Next: Tai-Be-Hlai implosive correspondences.

*I wonder what LA stands for. L obviously cannot represent a Japanese name. HAINANESE 前音 'FRONT SOUNDS' (PART 5: AFFRICATE RETENTION)

In part 2, I wrote that *ts became Hainanese t. This is not quite true. *ts (partly from even earlier *dz, *ʈ, *ɖ) palatalized before front vowels and merged with *tɕ:

*ts- > tɕio

*ts- > tɕiŋ

*dz- > *ts- > tɕi

*dz- > *ts- > tɕe

*ʈ- > *ts- > tɕiaŋ

*ɖ- > *ts- > tɕik

*tɕ > tɕiu

*tɕ > tɕin

Such a palatalization never occurred in the source of Sino-Vietnamese, which has t < *(t)s corresponding to Chinese *ts even before front vowels: e.g., 精 tinh < *(t)siŋ (but Hainanese tɕiŋ).

One might expect *dʑ to devoice to just as *dz devoiced to *ts, but *dʑ merged with *s (via > *ɕ) and became t (see part 2):

*dʑ- > *ʑ- > *ɕ- > *s- > ti


*ʑ- > *ɕ- > *s- > tot

*ɕ- > *s- > tu
西 *s- > tai

I can't explain why the Middle Chinese homophones 集 and 辑 have different initials in Hainanese:

*dz- > *ts- > tɕip

but 辑 is tip, as if it were from *sip (< ?*zip) rather than *dzip

I also have no idea how to account for instances of before a in Hainanese in the syllables tɕa, tɕai, tɕak. All other syllables with have front vowels after their onsets: tɕia, tɕi, tɕiu, tɕe, tɕio, etc.

The three exceptional syllables tɕa, tɕai, tɕak are 'chorphans' (character orphans) without sinographs, implying that they may represent morphemes without Middle Chinese sources (e.g., loans from other languages on Hainan?).

tɕa and tɕak contrast with tɕia and tɕiak. tɕai has no counterpart tɕiai since Hainanese lacks the rhyme -iai. HAINANESE 前音 'FRONT SOUNDS' (PART 4: YOD FORTITION)

In Vietnamese, *j hardened to *dʲ (romanized by 17th century missionaries as d, a spelling that persists to this day) which then weakened to [z] and perhaps even back to [j] depending on dialect. (It's also possible that modern [j] is an archaism in some cases, but it cannot be an archaism when it goes back to *C-t- and *C-d-.)

Hainanese *j hardened in a slightly different manner, becoming dz with an allophone before i. (Oddly, /s/ has no palatal allophone before i.) Voiced affricates are absent from modern Vietnamese, though *dʑ existed in the 17th century (when it was spelled as gi; this sound is now [z] or [j] in modern dialects).

Vietnamese *j is partly from earlier *mj, but Hainanese *j is partly from nonlabial nasal sources as well as *j:

dzia < *j- < *ɲ- < *ŋj-

cf. Sino-Vietnamese nhã [ɲa], Cantonese ŋa, Mandarin ya

dzoa < *j- < *ɲ-

cf. Sino-Vietnamese nhiệt [ɲiət] and Cantonese jit
尿 dzio < *j- < *ɲ- <*nj-

cf. Sino-Vietnamese niệu, Cantonese niw, Mandarin niao

dzi < *j-

cf. Sino-Vietnamese [zɨ] < *jɨ, Cantonese jy, Mandarin yu

Earlier *mj became Hainanese m and did not merge with *j which later became dz:
mit < *mj- (not dzit)

cf. Sino-Vietnamese diệt [ziət] < *j- < *mj- but Cantonese mit, Mandarin mie

Unlike Vietnamese *j, Hainanese *j can also come from *w (via *wj?):

dzon < *j- < *wj- < *w- (?*wion < ?*win)

cf. Sino-Vietnamese vận [vən], Cantonese wɐn

SV was borrowed from a dialect like the ancestor of Cantonese in which *i lowered to a -like vowel; no such lowering occurred in the ancestor of Hainanese

dzioŋ < *j- < *wj- < *w-

cf. Sino-Vietnamese vinh, Cantonese wiŋ


Aspirated stops generally shifted to fricatives in both Vietnamese and Hainanese:

Original stop





ph [f]

th [th]

kh [x]




I assume these shifts occurred independently.

De Rhodes' mid-17th century romanization and description of Vietnamese phonetics indicate that Vietnamese aspirated stops had not yet become fricatives. Some 20th century Vietnamese dialects still preserved *ph as a stop, and [ph] and [f] coexisted in early 20th century Saigon, though only [f] survived there a few decades later (Gregerson 1969: 149). Gregerson (1969: 165) reported an affricate pronunciation [kx] of kh in unspecified "[m]odern dialects". I know of no Vietnamese dialect with a fricative pronunciation of th. If I take all of this into account, I can only conceive of one highly improbable scenario linking the two shifts:

- contact between Vietnamese and Hainanese occurred sometime after the mid-17th century

- *ph and *kh lenited to *f and *x in one language that influenced the other (with the exception of some Vietnamese dialects)

- later, Hainanese lenited *th to *h (via *θ?) and *x merged with *h

Aspirate to fricative shifts are not unusual. Greek and Avestan underwent such shifts independently. The shift of to /h/ occurred in Irish. (Irish /h/ from is spelled th.) If two nonadjacent languages both underwent a nonexotic change, there is no need to believe that contact was involved.

My guess is that

- most Vietnamese dialects shifted grave aspirated stops to fricatives after the mid-17th century, leaving *th unchanged.

- Hainanese shifted all three aspirated stops including *th (stage 1) to fricatives at some unknown point in the past (stage 2). backed to *h (stage 3), which then merged with *x (stage 4).

Stage 1




Stage 2



Stage 3




Stage 4



Earlier Hainanese *h became zero (before and *x became *h?):

*θ, *x > *h > Ø


*thi > *θi > hi ?'sky'

*khu > *xu > hu ?'go'

*hu > u ?'exist'

This change did not occur in Vietnamese: the Sino-Vietnamese reading of 有 'exist' is hữu. (Vietnamese has no true zero initial. Orthographic zero is really initial glottal stop: e.g., ữu is [ʔɨw], not [ɨw].)

Next: Fortition of *j in Vietnamese and Hainanese. HAINANESE 前音 'FRONT SOUNDS' (PART 2: THE ALVEOLAR CHAIN SHIFT)

The following chain shift occurred in Vietnamese:

*ch > *(t)s > *t > [ɗ] (spelled đ)

(It's not clear whether Vietnamese ever had a *ts distinct from *s.)

*t became an implosive, leaving an empty t-slot to be filled by *(t)s, and so on. These changes can be inferred from external comparisons (Gage 1985: 503, 509):

đến [ɗen] 'come': cf. Khmer taən 'get up'

tóc 'hair': cf. Khmer sɑk 'hair'

xéo [sɛw] 'slanting': cf. Khmer chaew 'crooked'

the Vietnamese spelling x reflects the closest Portuguese equivalent to 17th century Vietnamese *[ɕ] which later became [s]. 17th century Vietnamese had an empty s-slot.

A similar chain shift occurred in Hainanese:

*tsh > *(t)s > *t > ɗ

(7.16.2:00: See part 5 for instances of *ts that become *tɕ.)

Other Chinese languages such as Mandarin maintain the unshifted original consonants:

ɗo ?'knife': cf. Md dao [taw]

ta 'three': cf. Md san

to ?'do': cf. Md zuo [tswo]

so ?'mistake': cf. Md cuo [tshwo]

Hainanese often bears a superficial similarity to Sino-Vietnamese: e.g,
ɗo ?'knife': cf. SV đao [ɗaaw]

ta 'three': cf. SV tam

to ?'do': cf. SV tố

but 错 so < *tsh- ?'mistake' corresponds to SV thác < *(t)shak because early Vietnamese *(t)sh merged with *th instead of shifting to *s

(It's not clear whether early Vietnamese had *tsh or an aspirated *sh like Burmese.)

I presume that the shifts underlying this similarity must have taken place independently since I know of no evidence for early contact between Vietnamese and any of the ethnic groups of Hainan.

Did Be - also on Hainan - undergo a similar shift? The only Be words I have ever seen are in Ostapirat (2000), and none have initials going back to *s or *ch.

I can't quickly find any evidence for such a shift in Hlai using the limited comparative material I have on hand.

If Vietnamese did influence Hainanese, there should be more points of similarity. But most sound changes in Hainanese do not resemble those of Vietnamese, as I'll explain in future installments of this series. VOICING THE VOICELESS: IMPLOSIVES

Last night, I asked,

Are there any other languages in which voiceless obstruent onsets have voiced before vowels?

Today I realized that I could have answered my own question. In a number of Southeast Asian languages, initial *p- and *t- have become voiced implosives [ɓ] and [ɗ]: e.g.,

Vietnamese (comparisons from Gage 1985)

bắn [ɓan] 'shoot'; cf. Khmer paɲ

đến [ɗen] 'come' : cf. Khmer taən 'get up'

Hainanese (Xu and Yang)

ɓi ?'side' < Middle Chinese *pen 

ɗo ?'knife' < Middle Chinese *taw

(I need to pick up where I left off with Hainanese.)

Be (Ostapirat 2000: 57), a Kam-Tai language spoken on Hainan

ɓoi 'go' < *p-; cf. Siamese pai

dam (typo for ɗam?) 'low' < *t-; cf. Siamese tam

(Earlier *ɗ- seems to have become Be l-. Ostapirat has no examples of Be reflexes of earlier *ɓ-. Did *ɓ- also become a sonorant [e.g., w-]?)

Siamese has b- and d- (presumably from earlier *ɓ- and *ɗ-) which correspond to Indic p- and t-/ʈ- in loanwords (Gedney 1947: 80):

buuchaa 'to worship' < Skt or Pali puujaa

daaraa 'star' < Skt or Pali taaraa

diikaa 'petition' < Skt or Pali ʈiikaa

Khmer has ɓ- and ɗ- which correspond to Indic p- and t-/ʈ- in loanwords:

ɓootɕiə 'sacrifice' < Skt or Pali puujaa

ɓɑɑreʔ (also paʔreʔ!) 'around' < Skt or Pali pari-

ɗaaraa (also taaraa!) 'star' < Skt or Pali taaraa

ɗəjkaa 'petition' < Skt or Pali ʈiikaa

Many Indic loanwords in Siamese probably came into the language through Khmer.

Note that both Siamese and Khmer also have p- and t- corresponding to p- and t-/ʈ-. Gedney (1947: 81) proposed that Siamese words with initial p- and t- for Indic p- and t- were borrowed after a shift that voiced the initials of earlier loans.

There is no evidence for implosives in Indic and tonal and vocalic evidence in Siamese and Khmer respectively point toward voiceless sources of ɓ- and ɗ-: e.g., *ʔp- and *ʔt-. Presumably Indic p- and t- could sound like *ʔp- and *ʔt- to early Khmer (and Siamese?) ears.

*ʔp- and *ʔt- sound like Korean tense pp- and tt-. There are rare cases in which Japanese p- t- k- were borrowed as Korean tense pp- tt- kk- rather than lax p- t- k- (Ito, Kang, and Kenstowicz 2006: 77). (I know of no examples of Japanese ch- [tɕ] borrowed as Korean tense cc-.)

Ito et al. (2006: 92) also list examples of French p- t- k- borrowed as Korean tense pp- tt- kk-:

ppari (more commonly phari) < Paris

ttullujU (more commonly thullujU) < Toulouse

kkannU (more commonly khan; also khannU) < Cannes

(Is -nnU influenced by Jpn kannu and/or the French spelling with -nn-?)

Note, however, that there is no evidence for a *ʔk- in Siamese or Khmer comparable to the kk- of Korean. Indic k always corresponds to Siamese and Khmer k. If *ʔk- ever existed in Siamese and/or Khmer, it merged with *ʔ- (as Pinnow 1980: 130 suggested for Khmer) or *k- instead of becoming ɠ- or g-. ɠ is rarer than ɓ, ɗ, and even palatal ʄ in the UPSID sample:

ɠ ʄ ɗ ɓ
% of the 451 languages in the UPSID database 1.1% 2.2% 5.1% 10.9%

Siamese j- is partly from *ʔj- (Li Fang-kuei 1977: 181) which could be partly from an even earlier *ʔc-, the palatal counterpart of *ʔp-, *ʔt-, and *ʔk-. Modern Khmer j- may be partly from *ʔj- (Pinnow 1980: 124) which could partly go back to *ʔc-.

In any case, all of the above examples of voiceless obstruent onset voicing before vowels involve implosives, not fricatives like Germanic *f- and *s-. So I still wonder if *f- *s- > v- z- is unique to Germanic. I wish there were an UPSID-like database of known sound changes, though I realize that 'known' is open to question since history is far more controversial than synchronic phonetics.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2008 Amritavision