In part 5, I should have explicitly stated my criteria for the reconstruction of presyllables at the Proto-Kra level: at least one form in each of the two major branches of Kra (Southwestern and Central-Eastern) with a lenited or fused initial. (By 'fused' I mean initials like Paha d- < *nt- which may be from *NV-t-.) Even so, there is no guarantee that each branch had the same presyllable. In part 4, I cited an example of the same root with different presyllables in different Laha languages. If there is presyllabic variation within Laha, there could have been such variation within larger Kra groupings.

If there is only evidence for a presyllable in only one branch, that presyllable may either be a retention from PK lost in the other branch or an innovation absent in the other branch.

Moreover, a trace of a presyllable found only in one language in one branch does not necessarily entail reconstructing a presyllable in the common ancestor of that branch: e.g., a presyllable in pre-Qiaoshang may not have been inherited from Proto-Gelao, Proto-Western Kra, or Proto-Southwestern Kra.

With the above in mind, let's look at the four Proto-Kra words that Ostapirat reconstructed with voiced retroflex stops. (The numbering continues from part 5.)

6. 'leaf' (p. 227)

Ostapirat: *ɖiŋ

This site: *CV-diŋ

SW: Qiaoshang zen

CE: Paha ðɛɛŋ, Buyang ʔdiaŋ

7. 'crow' (p. 239)

Ostapirat: *ɖəŋ

This site: *CV-dəŋ

SW: Qiaoshang zã, Laozhai dõ, Wanzi thaŋ < *d-

CE: Paha ðaŋ, Buyang ʔdaŋ

Although I would be happier if I knew of a Southwestern Kra word for 'leaf' with a stop initial, these two cases are straightforward: lenited initials in both Qiaoshang and Paha point toward a PK presyllable.

8. 'raw' (p. 236) k-d-

Ostapirat: *(k-)ɖep

This site: *(kV-)dep

SW: Qiaoshang zĩ (p. 114) or zen (p. 142), Laozhai dæ, Ta Mit Laha kthop < *k-d-

CE: Buyang ʔdip (but no Paha form!)

Without a Paha form with initial ð-, I cannot confirm that there was a presyllable in Central-Eastern Kra.

Qiaoshang of the Western branch and Ta Mit of the Southern branch point toward a presyllable in Proto-Southwestern Kra, but it would be dangerous to project a *kV- up into Proto-Kra. Moreover, there is no guarantee that the presyllable in pre-Qiaoshang had a velar initial like the k- still in Ta Mit today. Pre-Qiaoshang and pre-Ta Mit may have added different presyllables to the same root.

I am not even certain if the same root is involved since the two different Qiaoshang forms (is one an error?) have unexpected nasality. Could the pre-Qiaoshang form be something like *CV-dem with a nasal counterpart of *-p?

To complicate matters further, the Proto-Tai cognate of this word is *ʔdl/rip in Li Fang-kuei's system of reconstruction with an uncertain liquid medial. The liquid is in Saek rip and, outside Tai, in T'en lip (Li Fang-kuei 1977: 130).

Once I would have reconstructed Proto-Kra-Dai *rV-ʔd- to account for these liquids in Kam-Tai and Ostapirat's Proto-Kra retroflex *ɖ-, but I no longer think PK had retroflexes.

Now I wonder if the PKD form had a simple *ʔd- and different branches independently added different presyllables: *kV- in pre-Ta Mit (and pre-Qiaoshang?) but *rV- in Proto-Kam-Tai:

PKT *rV-ʔd- > *ʔdr- > T'en l- but d- in Siamese and most Tai languages other than Saek which has r-

If there were Kra-internal evidence for *rV-, I would reconstruct *rV- at the PKD level.

9. 'navel' (p. 222) m-d-

Ostapirat: *m-ɖaɯ A

This site: *NV-daɯ

SW: Qiaoshang zo, Laha dau

CE: Paha naau, Buyang ʔduə

Paha n- is from *nd- < *NV-d-. *m- is Ostapirat's notation for his generic presyllabic nasal. I prefer to write it as *N- since its point of articulation cannot be determined. If it was *m- (or some other nonalveolar nasal), it assimilated with the following *d- in pre-Paha: *md- > *nd- > n-.

Ostapirat and I assume that pre-Paha and pre-Qiaoshang both had a nasal-initial presyllable, but this may be incorrect. Pre-Paha could have discarded the original presyllable in favor of *NV-, or pre-Paha and pre-Qiaoshang independently added different presyllables to an inherited root with initial *d-.

There is no Kra evidence for a liquid in 'navel', even though there is a liquid in *ʔbl/rɯ, the Proto-Tai cognate according to Li Fang-kuei's reconstruction system. The initials for this morpheme* vary in Tai: b-, d-, l-, and even n- (Li Fang-kuei 1977: 91-92). Perhaps this *ʔbl/r- is from an earlier *ʔb-d- (< ?*ʔV-PV-d-).

Other Kra-Dai languages listed in Peiros' online database on Starostin's site have initials such as

Be and Lakkia m- (< ?*m-d-)

Proto-Hlai *zw-

Proto-Kam-Sui *ʔblw- (Thurgood; I assume *b' = *ʔb)

(both perhaps < *b-dw- < *b-d- with a labial preinitial conditioning *-w- insertion** and the lenition of *-d- to *-z- before presyllables in Hlai [cf. Qiaoshang])

No single presyllable can be reconstructed at the Proto-Kra-Dai level; the only consonantal common denominator of 'navel' may be *d-:

Proto-Kra-Dai *d(w)-
Proto-Kra *NV-d- (assuming Paha n- reflects the original PK presyllable) Proto-Hlai *zw- (< ?*b-d(w)-) Proto-Kam-Tai *d(w)-
Be m- (< ?*m-d-) Lakkia m- (< ?*m-d-) Proto-Tai *ʔbl/r- (< ?*ʔb-d-) Proto-Kam-Sui *ʔblw-

Proto-Hlai and Proto-Kam-Sui *-w- could be a retention from Proto-Kra-Dai.

*I use the term 'morpheme' because there is no Siamese word dɯɯ 'navel'. The actual word is sa-dɯɯ with a sa- prefix which is a reduction of saaj 'line; band'.

**I proposed a similar process in Tangut: e.g.,

TT1857 swe R34 1.33 'pure' < *P-se


(8.2.0:18: Added modern Kra forms.)

Ostapirat (2000) reconstructed nineteen Proto-Kra words with retroflexes in his appendix. Nine have voiceless retroflex stops:

1. 'egg' (p. 224)

Ostapirat: *ʈəm A (A-D represent PK tonal classes)

This site: ?*CV-təm (I prefer to interpret PK tonal classes in terms of segments: e.g., A = final sonorant.)

SW: Qiaoshang zø, Laha tam

CE: Paha ðam, Buyang tam

(SW = Southwestern and CE = Central-Eastern, the two major divisions of Kra according to Ostapirat)

2. 'bite' (p. 238)

Ostapirat: *ʈai B

This site: ?*CV-taih

SW: Qiaoshang zei, Laha tai

CE: Paha ðaai (no other CE forms available)

These two words have lenited initials in two distant branches (Qiaoshang z-, Paha ð-) so I think it is safe to project presyllables conditioning lenition up to the PK level:

PK *CV-t- > Proto-Southwestern Kra *CV-t- > pre-Qiaoshang *CV-d- > *CV-ð- > Qiaoshang z-

PK *CV-t- > Proto-Central-Eastern Kra *CV-t- > pre-Paha *CV-d- > *CV-ð- > Paha ð-

3. 'eye' (p. 220)

Ostapirat *m-ʈa A

This site: ?*NV-ta < *?mata

(cf. Proto-Malayo-Polynesian *mata < Proto-Austronesian *mata 'eye')

SW: Qiaoshang ze, Laha taa

CE: Paha daa, Buyang taa

4. 'dry in sun' (p. 240)

Ostapirat *m-ʈak D

This site: ?*NV-tak

SW: Qiaoshang zɒ, Wanzi tei

CE: Paha daak, Buyang taak

(8.2.0:15: Why does Pubiao have hraak instead of the expected taak? Is the unexpected hr- evidence for a presyllable with initial *r- in pre-Pubiao, if not at higher levels?)

These two words have a lenited z- in Qiaoshang and a voiced d- < *nt- in Paha, so I think a nasal-initial presyllable conditioned lenition in Qiaoshang

*NV-t- > *NV-d- > *NV-ð- > z-

and fused with a stop initial in Paha:

*NV-t- > *nt- > *nd- > d-

Ostapirat reconstructed *m- as a generic preinitial nasal in PK. I don't know of any Kra-internal evidence for specifying a labial nasal. Perhaps Ostapirat was influenced by the m- of Austronesian words for 'eye' which have been compared to Kra-Dai words for 'eye'. (8.2.0:10: It is debatable whether the resemblance between the words is due to borrowing or a genetic relationship.) Even if 'eye' had PK *m-, this does not necessarily mean that 'dry in sun' had *m- as opposed to some other nasal.

(8.2.1:36: A nasal presyllable is not a PK innovation because it is also supported by Sui ndaa A1 [Ostapirat 2000: 4]. It is unlikely that two different branches of Kra-Dai (Kra and Sui [of the Kam-Tai branch] independently added nasal-initial presyllables to *ta. Note, however, that the Tai language Saek has praa < *ptaa. Could *pt- be from *mt- with *m- denasalizing to assimilate to the following oral obstruent?)

5. 'louse (head)' (p. 225)

(this section heavily revised 8.2.0:56 to incorporate the Wanzi tonal evidence)

Ostapirat *C-ʈu A

This site: ?*CV-tu

SW: Wanzi ta A2 (rather than the expected ta A1)

CE: Paha ðhuu, Buyang tuu

Since there is no Qiaoshang cognate, the only evidence outside Paha for a presyllable is the irregular A2 tone of the Wanzi form which might reflect a voiced-initial presyllable (bold = voicing):

*CV-t- + A1 > *Ct- + A1 > *Cd- + A2 > *d- + A2 > t- + A2

Wanzi t- can be from *d- as well as *t- (Ostapirat 2000: 109).

The Paha initial ðh- could reflect a presyllable present in pre-Paha or some other level lower than PK. The aspiration in Paha could be a trace of a velar initial in the lost presyllable:

*kV-t- > *kV-d- > *kV-ð- > *kð- > *xð- > ðh-

Another possible source of Paha ðh- in 'louse' is PK *C-tj- (which would be *CV-tj- in my notation; see Ostapirat 2000: 214 and Guillaume Jacques 2001: 104).

These hypotheses about Paha could be combined with the voiced-initial presyllable implied by Wanzi tone A2:
PK *gV-t(j)- > pre-Wanzi *gd- > *d- > Wanzi t-
PK *gV-t(j)- > *gV-d(j)- > *gV-ð(j)- > *gð(j)- > *ɣð(j)- > Paha ðh-

Presyllables do not seem to affect Paha tones: e.g., PK *NV-ta 'eye' became Paha daa A1, not daa A2. (The A1 tone reflects the original voiceless *-t-, not the voiced d- derived from *Nt-, the fusion of the presyllable and *-t-.)

Ostapirat (2000: 215) reconstructed *d- in PK presyllables but not grave *g- or *b-. *g- or *ɣ- (cf. the Written Tibetan preinitial H- [ɣ]) could be substituted for the *ɦ- that Ostapirat (2000: 182, 185) reconstructed in Proto-Central-Eastern Kra: e.g.,

'rope': PCEK *gV-ʃ- (Ostapirat's *ɦ-ʃ-) > Paha jh-

'five': PCEK *gV-m- (Ostapirat's *ɦ-m-) > Paha mh-

Ostapirat (2000: 245) reconstructed PK 'five' as *r-ma A, even though I know of no Kra-internal evidence for *r- as opposed to another voiced consonant. (Wanzi mpu lacks the expected retroflex vowel pointing toward *r-.) Ostapirat's choice of *r- may have been influenced by the *l- of Proto-Austronesian *lima. (He could not reconstruct *l-m- because the expected Wanzi reflex would be mpl-, not mp-.) PK *g- and *r- could have merged as PCEK *ɦ-:

PK *g-C- > *ɣ-C- > PCEK *ɦ-C-

PK *r-C- > *ʀ-C- > *ʁ-C- > *ɣ-C- > PCEK *ɦ-C-

Perhaps pre-Wanzi replaced the *rV- presyllable inherited from PK with a nonrhotic-initial presyllable. DID PROTO-KRA HAVE RETROFLEX INITIALS? (PART 4: CENTRAL KRA)

At the end of part 3, I asked,

Why can't I regard some presyllables as Qiaoshang innovations? Why is it necessary to project them all the way back to the PSWK level?

Or to the Proto-Kra level? If feature X is in one branch of a language family, it need not be in the common ancestor of the family: e.g., Indic retroflexes do not go back to Proto-Indo-European.

Presyllabic variation is not unknown in Kra: e.g., these forms in Laha languages for 'star' (Ostapirat 2000: 37):

Nong Lay kluŋ (with k- as a trace of a velar-initial presyllable)

Ta Mit ma luŋ (with a labial-initial presyllable)

Ostapirat reconstructed early Laha *kV-l- for 'star' and presumably regarded the Ta Mit presyllable as an innovation.

I don't think presyllables before alveolars are Qiaoshang or even Proto-Gelao, Proto-Western, or Proto-Southwestern Kra innovations because Qiaoshang z- corresponds to the lenited initial ð- in Paha, which belongs to the Central branch of the Kra family. I believe that pre-Qiaoshang and Paha independently retained Proto-Kra presyllables; it is less likely that both languages independently happened to add presyllables to the same words.

(8.1.0:45: I have added Buyang forms to exemplify how Eastern Kra has no traces of presyllables [unless ʔd- is a compression of an earlier *CV-d-, as I suspect it may be].)

Example Pages in Ostapirat (2000) Qiaoshang PG/PWK /PSWK Paha Buyang
egg 113, 178 *CV-t- ðam tam
crow (v.) 22-23, 114, 183 zã *CV-d- ðaŋ ʔdaŋ
hawk 114, 188 *kV-l- ðaaŋ laaŋ

If I were unaware of Paha, I would might regard presyllables in 'egg' or 'crow' as Qiaoshang innovations because I know of no evidence for them in other Kra languages.

Cases in which Qiaoshang z- does not correspond to Paha ð-:

Example Pages in Ostapirat (2000) Qiaoshang PG/PWK/PSWK Paha Buyang
dry in sun 114, 240 *NV-t- daak taak
eye 113, 179 ze daa taa
thick 114, 237 *CV-n- naa naa
bird 114, 184, 224 zau nhook no cognate
fat 114, 184, 235 nan nɛn
pillar 114, 182 *NV-ts- dʑhuu θu
mountain 114, 229 zɤu *CV-dz- no cognates
choose 114 zen

(8.1.1:01: The -h- in Paha nhook 'bird' and dʑhuu 'pillar' is conditioned by combinations of non-A2 tones with voiced initials and is not a trace of a presyllable. The lack of -h- in A2 tone words may be due to dissimilation: A2 could have conditioned *breathy voice in pre-Paha, and when an -h- developed after all *voiced initials, it could have dropped before breathy vowels. Breathiness was later lost.)

(8.1.1:25: All sibliants other than *ɦ-(t)ʃ- have merged into Buyang θ- [Ostapirat 2000: 181-182]. so Buyang θ- is not necessarily from a lenited *ts-. Ostapirat [2000: 181] listed one apparent exception, ɕiak 'pestle' < *tʃ-, but did not explain it in his text, and did not list it as a Buyang form in his appendix. I wonder if it is from a variety of Buyang other than the E-Cun variety Ostapirat generally uses.)

I know of no non-Qiaoshang evidence for reconstructing *CV-n- or *CV-dz-.

Ostapirat (2000: 179, 182) reconstructed Proto-Central-Eastern Kra *nt- and *nts- as sources of Paha d- and dʑ-. These prenasalized obstruents could be contractions of Proto-Kra *NV-t- and *NV-ts- preserved in the southwestern branch down to the Proto-Gelao level. (It's also possible, but less likely, that a Proto-Kra nasal-initial presyllable was replaced by a nonnasal-initial presyllable at some point in the development of the southwestern Kra languages.) Note that Laozhai has palatal tɕ- instead of retroflex tʂ- for 'pillar', possibly indicating a nonrhotic (but not necessarily nasal) initial presyllable (unless the form in Ostapirat 2000: 114 is a typo).

Next: Revising Ostapirat's retroflex-initial reconstructions of Proto-Kra etyma. DID PROTO-KRA HAVE RETROFLEX INITIALS? (PART 3: SOUTHWESTERN KRA)

Lachi, sister of the Gelao languages, has alveolars and palatals corresponding to Ostapirat's Proto-Gelao alveolars and retroflexes (which I reconstruct as presyllable + alveolar sequences):

# Ostapirat's Proto-Gelao My Proto-Gelao Lachi
1 *t, *ʈ- *(CV-)t- t-
2 *d, *ɖ- *(CV-)d- tɦ-
3 *n-, *ɳ- *(CV-)n- n-
4 *l-, *ɭ- *(CV-)l- l-
5 *ts-, *tʂ- *(CV-)ts- tɕ-
6 *dz-, dʐ- *(CV-)dz- tɕɦ-
7 *r- *r- lɦ-, kɦɣ- (see below)
8 *hr- *hr- unknown; l-?

Ostapirat (2000: 146-147) projected the above Proto-Gelao initials back into Proto-Western Kra, the common ancestor of Gelao and Lachi. (7.31.0:36: But clearly PWK *r- could not have developed into Lachi kɦɣ-. See below.)

Laha, southern sister of the Western Kra languages (Gelao and Lachi), has a similar pattern of correspondences:

# Ostapirat's Proto-Western Kra My Proto-Western Kra Laha
1 *t, *ʈ- *(CV-)t- t-
2 *d, *ɖ- *(CV-)d- d-
3 *n-, *ɳ- *(CV-)n- n-
4a *l- *l- l-
4b *ɭ- *CV-l- kl-
5 *ts-, *tʂ- *(CV-)ts- c-
6 *dz-, dʐ- *(CV-)dz- unknown; j-?
7 *r- *r- khl-
8 *hr- *hr- unknown

The PWK initials can be projected back into Proto-Southwestern Kra with at least a couple of exceptions. Ostapirat (2000: 167)* reconstructed PSWK *kV-l- as the source of Lachi kl- and my PWK *CV-l- (which can now be rewritten as *kV-l-). He also reconstructed PSWK *kV-r- as the source of Lachi khl- and my PWK *r- (which in some or all cases may be revised as PWK *kV-r-):

Proto-Southwestern Kra Laha Proto-Western Kra Lachi
*kV-r- khl- *r- (or *kV-r-?) lɦ- (was the presyllable *kV- already lost in PWK, or did it drop in pre-Lachi?)
*kV-r- kɦɣ-

Southwestern Kra languages generally lost most presyllables without a trace with a few exceptions: the z-initials of Qiaoshang, the retroflex affricates of Laozhai, and the k(h)l-clusters of Laha.

Next: Why can't I regard some presyllables as Qiaoshang innovations? Why is it necessary to project them all the way back to the PSWK level?

*7.31.00:21: Ostapirat wrote presyllables as C- rather than as CV-. I have changed his notation to mine for consistency. DID PROTO-KRA HAVE RETROFLEX INITIALS? (PART 2: GELAO)

Over the last few days, I've been rethinking what I meant to say about retroflexes in Kra. I'm going to adopt a leaves-to-trunk strategy and look at the reflexes of Ostapirat's (2000) proto-reflexes in individual Kra languages.

I assume that Ostapirat's (2000: 25) subgrouping of Kra is correct:

Southwestern Kra Central-Eastern Kra
Western Kra Southern Kra Central Kra Eastern Kra
Gelao Lachi Laha Paha Buyang Pubiao

In this post, I will focus on the Gelao languages which are mostly spoken in Guizhou. Gelao has three branches: Southwestern, Northern, and Central, represented in Ostapirat (2000) by Laozhai, Qiaoshang, and Wanzi. Laozhai and Wanzi alveolars have two types of correspondences with Qiaoshang alveolars, palatals, and retroflexes:

# Laozhai Wanzi Qiaoshang
Type 1 correspondence Type 2 correspondence
1 t- t- t- z-
2 d- t(h)-
3 n- n(t)- ndʑ-
4 l- l-
5 tʂ-, tɕ- s- none
6 dʐ- ts(h)-
7 ʐ- z- ʐ-
8 ʐ- h- s-

Ostapirat reconstructed the following Proto-Gelao consonants to account for those correspondences:

# Type 1 correspondence Type 2 correspondence
1 *t- *ʈ-
2 *d- *ɖ-
3 *n- *ɳ-
4 *l- *ɭ-
5 none *tʂ-
6 *dʐ-
7 *r-
8 *hr-

In his system, all Type 2 correspondences reflect Proto-Gelao retroflexes. Yet only four of his proto-retroflexes have retroflex reflexes in Qiaoshang and/or Laozhai. Moreover, why would a voiceless retroflex stop voice in initial position?

I propose two subtypes of Type 2 correspondences. The first subtype (in bold) reflects alveolars which were lenited after lost presyllables whereas the second subtype reflects true retroflexes:

# Type 1 correspondence Type 2 correspondence
1 *t- *CV-t-
2 *d- *CV-d-
3 *n- *CV-n-
4 *l- *CV-l-
5 none *tʂ-
6 *dʐ-
7 *r-
8 *hr-

The nonlenited/lenited alveolar correspondences (1) and (2) are reminiscent of this nonlenited [ɗ] < *t ~ lenited [z] < *CV-t- alternation in Vietnamese:

đao [ɗaaw] < *taaw ~ dao [zaaw] < *CV-taaw 'knife'

borrowed from Middle Chinese 到 *taw

7.30.0:30: *-t- voiced intervocalically in both Vietnamese and Qiaoshang, merging with *-d- (voiced segments in bold):

*CV-tV > *CV-dV

Note, however, that intervocalic *-n- and *-l- are not known sources of Vietnamese d- [z].

7.30.1:07: Proto-Gelao intervocalic *-n- may have shifted to pre-Qiaoshang *-nd(ʑ)- (cf. how Proto-Gelao *n- > Qiaoshang ndʑ-) and then merged with *-d- or *-(d)z- (see below).

Proto-Gelao *-l- could have lenited to pre-Qiaoshang *-ɮ- before shifting to z-.

7.30.1:40: I am puzzled by (5) and (6). Laozhai has retroflexes but Qiaoshang doesn't. Compare (5) with (9) and (6) with (10):

# Proto-Gelao (Ostapirat) Proto-Gelao (alternative) Laozhai Wanzi Qiaoshang
5 *tʂ- *CV-ts- tʂ-, tɕ- s- z-
6 *dʐ- *CV-dz- dʐ- ts(h)-
9 *ts- *ts- ts- s- ts-
10 *dz- *dz- dz- ts(h)-

My alternative reconstruction explains why (5) and (6) did not develop like (7) and (8) in Qiaoshang:

Nonrhotic (5 and 6):

*CV-ts-, *CV-dz- > merged as *CV-dz- > z-

Rhotic (9 and 10):

*r- > *ɽ- > ʐ-

*hr- > *hɽ- > *ʂ- > s-

But why does Laozhai have retroflexes (and inexplicably a palatal tɕ- in 'pillar' - is that a typo?) instead of alveolars for (5) and (6)? Perhaps (5) and (6) were Proto-Gelao *rV-ts- and *rV-dz-. The presyllabic vowel dropped in pre-Laozhai and the clusters fused into retroflex affricates:

*rV-ts- > *rts- > tʂ-

*rV-dz- > *rdz- > dʐ-

Perhaps other coronal presyllable initials merged with *r- before *ts- and *dz-: e.g.,

*lV-ts- > *lts- > *rts- > tʂ-

Old Chinese *r- + affricate clusters are one probable source of Middle Chinese retroflex affricates:

掫 OC *r-tsu-ʔ > MC *tʂuʔ

(also has nonretroflex MC readings: OC *tsu > MC *tsəw, OC *tsu > MC *tsu)

驟 OC *rdzus > MC *dʐuh

(phonetic: 取 OC *tshoʔ > MC *tshuoʔ; like Karlgren [1957: 55], I cannot explain why a *-o phonetic was used to write *-u words)

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