18.6.13.23:27: EARLY OLD CHINESE PRESYLLABIC *I

Baxter and Sagart (2014: 224) use the notation *A for an Old Chinese *a that has unexpected Middle Chinese reflexes with palatal elements: e.g.,

土 Old Chinese *tʰˁaʔ > Early Middle Chinese *tʰɔˀ 'earth'

is phonetic/semantic in

社 Old Chinese *m-tʰAʔ > Early Middle Chinese *dʑiæˀ 'sacrifice to the spirit of the soil'

instead of Early Middle Chinese †dʑɨəˀ without any palatal vowels

in fact, the phonetic series of 土 has no examples of Early Middle Chinese †-ɨə

Baxter and Sagart speculated that what they write as *A could actually reflect the effect of some unknown preinitial consonant on the vowel.

I propose that we are actually seeing vocalic transfer: the effect of some unknown first syllable vowel on a second syllable vowel.

That occurred to me as I was working out the development of 抯 'to pull out of water' which I originally intended to cite in the addendum to my last post until I decided to use more straightforward examples.

Early Old Chinese: It is not clear whether the first variant has a prefix *r(i)- or an infix *-r-. *Ci- might be the same syllable as *r(i)-; if so, then there is a root *tsaʔ with only two prefixes, *ri- and *Ni-.

Middle Old Chinese: The *i in the first syllable has caused the following *a to break to *ia. No breaking occurred in *tsrˁa, either because its *i was lost before vocalic transfer (see the table below) or because it never had an *i (i.e., its *-r- was an infix).

Early vs. late *i-loss

*tsrˁaʔ has pharyngealization because it lacked a high vowel that would have blocked pharyngealization.

*r- and *-ts- underwent metathesis: *r-ts- > *tsr-.

Late Old Chinese:

*tsr- fused to retroflex *tʂ-.

*N-ts- fused to voiced *dz-.

Pharyngealization was lost.

*a broke to *ɤa after retroflexes.

*a fronted to *æ to assimilate with the preceding front vowel *i.

Early Middle Chinese:

*ɤa became *ea (to avoid *ɤ, a vowel absent elsewhere in the system) which then fused to *æ.

Final glottal stop *-ʔ became creaky voice (written with a subscript tilde).

Mandarin:

*æ backed to /a/.

*ts palatalized to /tɕ/ before *-i-.

The diphthong *iæ became /jɛ/.

'Deeper' phonemicizations are possible: e.g., Pulleyblank's (1984: 52) /iă/ or my /jə/.

Open syllables with *voiceless initials and nonglottalized, nonbreathy vowels developed tone 1.

Syllables with *voiceless obstruent initials and *creaky voice developed tone 3.

Syllables with *voiced obstruent initials and *creaky voice developed tone 4 via assimilation:

*dziæ̰ > *dzʱiæ̰ > *dzʱiæ̤ > *tsʱiæ̤ > /tɕjɛ4/

The *voiced initial became breathy voiced, and the *creaky voice in the following vowel became *breathy voice. Then the initial devoiced and lost its breathy voice to dissimilate from the following *breathy voiced vowel. Ultimately the vowel lost the breathiness that conditioned what is now Mandarin tone 4.

6.19.12:41: Contrast the three readings of 抯 'to pull out of water' with those of other words written with the same phonetic 且: 且/祖 'ancestor', 沮 'to leak', 沮 'marsh', and 菹 'marsh':.

Karlgren (1957: 32) regards the character 菹 as primarily representing Old Chinese *tṣi̯o 'to pickle', equivalent to my *r(ɯ).tsa.

*tsaʔ 'ancestor'

Possibly *CV.tsaʔ if the first syllable was lost without a trace. Who knows, maybe at some earlier point *CVC or even more complex first syllables were possible.

*NV.tsaʔ 'to leak'

I use periods to indicate breaks between syllables and - in later stages - the borders between presyllables and syllables if I cannot detect a morpheme boundary. I don't know of any 'leak' word family with different prefixes before a root √tsaʔ, so I assume *NV.tsaʔ was a disyllabic root rather than √tsaʔ plus a nasal prefix.

*ri.tsa 'marsh'

I don't know of any 'marsh' word family with different prefixes before a root √tsa, so I assume *ri.tsaʔ was a disyllabic root rather than √tsa plus a nasal prefix.

In Early Old Chinese, the higher vowels of first syllables mostly merged into *ɯ (my symbol for 'unknown high vowel' inspired by the Middle Korean minimal vowel ㅡ /ɯ/). *i is recoverable if it triggered vocalic transfer after acute initials, but I otherwise can't tell if *ɯ was from *i, *ə, *u, or even some other vowel like *y that no longer existed in Old Chinese.

How did three words become six? 'Marsh' developed four variants:

1. Conservative: Middle Chinese *tsiæ implied by the fanqie of Jiyun (1037, after the Middle Chinese period but still based on Middle Chinese phonology) is the direct descendant of *ri.tsa.

No real Mandarin descendant. /tɕjɛ1/ is a reading generated by reading the Jiyun fanqie as Mandarin, not a naturally transmitted word. It would be convenient to have a term for this kind of artificial modern reading.

This variant is not recorded in the phonological tradition but is reconstrucible on the basis of Mandarin /tɕy1/ unless that is a reading by analogy with 且 /tɕy1/.

Windows' IME says 沮 can also be read /tɕy1/.

3. Depalatalized first vowel and suffixed: *ri.tsa > *rɯ.tsa-s

I have no idea what the final *-s is doing.

This variant has no standard Mandarin descendant.

The phonetic series of 且 is 'mixed' in the sense that it includes characters for three types of Early Old Chinese (sesqui)syllables:

1. *(Cʌ.)CV: e.g., 且/祖 *tsa/ʔ 'ancestor'

If there was a low-vowel presyllable *Cʌ-, it wouldn't have affected the vowel in a following *a-syllable, so its presence or absence is impossible to detect. I only reconstruct it whenever it leaves a trace.

2. *Cɯ.CV: e.g., 沮 *Nɯ.tsaʔ 'to leak', 且 *Cɯ.tsa 'many'

3. *Ci.CV: e.g., 菹 *ri.tsa 'marsh', 且 *Ci.tsʰaʔ 'moreover'

'many' and moreover' might have a common root √(Ci.)tsa with a prefix or a root presyllable that conditioned aspiration in 'moreover':

*Ci.tsa-ʔ > *Ci.tsiaʔ > *C.tsiaʔ > *tsʰiaʔ?

The vowels of the three types (low *ʌ, high front *i, and high back *ɯ) are reminiscent of the three vowels of open presyllables in Pacoh (Watson 1964: 144): /a i u/¹. Watson's examples are:

/pa.piː/ 'to converse'

/ti.noːl/ 'post (n.)'

/ku.ceːt/ 'to die'

Vietnamese chết /cét/ 'to die' has no trace of the presyllable still in Rục ku.cíːt  'to die'.

Here are Early Old Chinese words with similiar presyllables and their post-vocalic transfer Middle Old Chinese descendants:

腦 *Cʌ.nuʔ > *nˁauʔ  'brain'

*Cʌ could be *pʌ-; cf. Proto-Austronesian *punuq 'brain'; see this post on the mismatch of the first vowels

借 *Ci.sak > *tsiak 'to loan, borrow'

*Ci- could be *ti-

牀 *kɯ.dzraŋ > *k.dzrɨaŋ 'bed'

Appendix: More examples of Early Old Chinese *Ci- words

I collected these from Baxter and Sagart (2014: 223-226) and Schuessler (2009: 45, 64). Middle Old Chinese forms follow to show the aftermath of vocalic transfer.

1. 邪 *Ci.ɢa > *ɢia 'interrogative particle'

2. 者 *Ci.taʔ > *tiaʔ 'nominalizing particle'

I am not comfortable with sesquisyllabic particles in a language whose roots are typically monosyllabic. But then again, maybe in EOC, sesquisyllabic roots were the norm.

3. 奢 *si.tʰa > *stʰia   'extravagant'

but not all characters with the phonetic 者 had *i: e.g.,

煮 *sɯ.taʔ 'to cook'

possibly cognate with Tangut

𗟞

4664 1liq1 < ?*S.ta < ??*Si.ta 'to cook'

The *i assumes cognancy with 炙; see below.

4. 車 *ki.kla > *kkia > *kʰlia 'chariot'

also *kɯ.kla > *klɨa 'id.'

cf. Proto-Indo-European *kʷékʷlo- 'wheel'

5. 寫 *Ci.saʔ > *siaʔ 'to depict'

6. 遮 *Ci.ta > *tia 'to cover'

7. 野 *mi.laʔ > *liaʔ 'open country'

also *mɯ.laʔ > *mlɨaʔ 'id.'; the Middle Chinese reading *dʑɨə̰ has an irregular *d-

里 *mɯ.rəʔ 'village' on the left might be phonetic; is the resemblance to Japanese mura coincidental? If the Japanese word were really related, I would expect †muro < ††-ə.

8. 昔 *Ci.sak > *siak 'in the past'

9. 柘 *Ci.taks > *tiaks 'mulberry tree'

In Proto-Min (PM), the merger of *i with *ɯ in presyllables was total before *-ak; there are no traces of *i, and both *Ci.Cak and *Cɯ.Cak have become PM *Ciok (Baxter and Sagart 2014: 226):

10. 炙 *si.tak > *tiak 'to roast'

PM *tšiok

cognate to 煮 *sɯ.taʔ 'to cook'?

is the resemblance to Japanese tak- 'to burn, to cook rice' coincidental?

11. 尺 *Ci.tʰak 'foot (measure')

PM *tšhiok

12. 石 *Ci.dak or *Ni.Tak 'stone'

PM *džiok

13. 螫 *Ci.l̥ak > *l̥iak 'to sting'

PM *tšhiok

14. 𥼶 *Ci.l̥ak > *l̥iak 'to wash rice'

PM *tšhiok

15. 射 *mi.lak > *mliak 'to hit with an arrow'

PM *džiok

16. 借 *Ci.sak > *tsiak 'to loan, borrow'

PM *tsiok

*C- fused with *s- into *ts-

17. 席 *Ci.dzak > *ziak 'mat' with irregular *z

PM *dziok

cf. 藉 *Ci.dzak-s > *dziaks 'mat'

Unsolved mystery: Why did *-i- only leave traces in syllables ending in *-a, *-aʔ, *-as, and *-ak(s)?

¹But in closed sesquisyllables, only /ə/ is possible: e.g., /ɓəmɓar/ 'to divide by two' < /ɓar/ 'two'. I do not know yet whether Old Chinese had closed sesquisyllables.

18.6.12.23:43: REFLEXES OF PROTO-TAI *P.T- IN SAEK

Earlier today (in a table in an addendum I finished on 6.14) I mentioned the 'famous' Saek word for 'eye' (praː) which attracts attention because it's not like Thai taː or similar words in other Tai languages. Pittayaporn (2009: 323) reconstructs its Proto-Tai source as *p.ta which elegantly accounts for the p-, -r- (< *-t-), and t-.

That made me curious about whether Proto-Tai *p.t- always became pr- in Saek. Going through Pittayaporn's list of Proto-Tai reconstructions, I see that Proto-Tai *p.t- has two different reflexes:

1. pr- as in 'eye' (above) and pra:j 'die' (Pittayaporn 2009: 357)

2. t- as in tɤ: 'gizzard' (Pittayaporn 2009: 330)

The presyllable *p.- must have been lost in the ancestor of Saek 'gizzard'; it is reconstructible on the basis of Bao Yen pʰɤɰ whose aspiration is from *-r̥- < *-r- < *-t- (cf. Cao Bang tʰɤj with the same source of aspiration).

Pittayporn (2009: 328) reconstructs Proto-Tai *p.tak 'grasshopper' even though that word has no reflexes in Saek or Bao Yen. Does it have any reflexes with p-like initials? I think he reconstructs *p.t- on the basis of forms like Cao Bang and Shangsi tʰak which have aspiration from  *-r̥- < *-r- < *-t- (as in Bao Yen). Even without Saek or Bao Yen or anything labial, the pattern of initials in Cao Bang and Shangsi matches that of *p.t-words rather than *t-words:

If Proto-Tai 'grasshopper' were simply *tak, the Cao Bang and Shangsi reflexes would be †tak with †t-.

6.15.10:16: Old Chinese had many words of the 'gizzard' type that had variants with and without presyllables: e.g., 扶 'to crawl'.

At a stage even before Early Old Chinese, the word may have been *Ni-pʰa, *Nə-pʰa, or *Nu-pʰa with a high series vowel that was later reduced to *ɯ in an unstressed position and ultimately lost.

In Early Old Chinese, the word had developed a variant without a presyllable. *pʰa is comparable to English 'cause, a variant of because without a presyllable be-. Presyllable loss - and other forms of reduction - are not entirely mechanically predictable. Just because because could lose its be- doesn't mean that it always did, much less that all be-words had such variation: e.g., there is no monosyllabic variant †lieve of believe.

In Middle Old Chinese, the high vowel of the presyllable conditioned the warping of *a to *ɨa. The variant without a presyllable had no high vowel and was subject to developing pharyngealization. I write pharyngealization after the initial consonant, but it was a quality of the entire syllable.

In Late Old Chinese, *N-pʰ- fused into *b-. *ɨ rounded to *u after labials. Pharygealized *a backed to *ɑ. Pharyngealization disappeared after leaving its mark on the vowel.

In Early Middle Chinese, *a raised and rounded to *o after *u. *ɑ raised and rounded to *ɔ.

In Late Middle Chinese, the vowels raised further: *uo > *u, *ɔ > *o. *b- became breathy *fʱ before *u.

In Mandarin, breathiness conditioned tone 2 before being lost. Open syllables without that breathiness or any laryngeals developed tone 1. *o raised even further to /u/.

痡 'suffering' and 鋪 'to spread out' both have two variants, one with a presyllable and one without. The bare version happens to be homophonous with the monosyllabic version of 'to crawl'.

*pʰ-, unlike *b-, did not develop a breathy reflex in Late Middle Chinese. As a result, Late Middle Chinese *fu became Mandarin /fu1/ rather than /fu2/ with tone 2 conditioned by *breathiness.

I suspect that the sesquisyllabic (and even earlier disyllabic) versions of 痡 'sufferihg' and 鋪 'to spread out' had very different first halves: e.g., *kupʰa and *pipʰa, etc. The original first consonants are not recoverable, and all that can be said about the original first vowel was that it was nonlow; a low series vowel (*a *e *o) would not have conditioned the warping of *a to *ɨa. *ɯ is my symbol for an unknown high series vowel. So the 'homophony' of 痡 'sufferihg' and 鋪 'to spread out' is an illusion caused by my agnostic notation *Cɯ-pʰa; the two words may not have been homophonous until Middle Old Chinese.

I don't know why 鋪 'to spread out' is written with the 金 'metal' radical. The sesquisyllabic version of 'to spread out' has a more common spelling 敷 with the radicals 方 'direction' and 攵 'action with hand'¹ which make more sense. 敷 is not a spelling of the monosyllabic version *pʰa.

Schuessler (2007: 173) regards 鋪敷 'to spread out' to be cognate to 布 *pa-s 'to spread out' and 博 *pa-k 'wide'. The aspirated initial *pʰ- may be from some earlier cluster like *kp- (which is absent from Baxter and Sagart's 2014 reconstruction). Perhaps the earliest reconstructible form of 鋪 'to spread out' is *kɯ-pa. The two Middle Old Chinese forms would then both reflect the presyllable.

In Stage 1, there is only one form of the word.

In Stage 2, the word develops a monosyllabic variant *kpa.

In Stage 3, the vowel of *kpa remains unbent since there is no presyllabic high vowel to condition the bending of *a to *ɨa.

In Stage 4, the presyllabic vowel of *kɯ-pɨa was lost.

In Stage 5, *kp- became *pʰ- - a change that probably also occurred in Middle Korean centuries later.

In Stage 6, the variant without a high vowel developed pharyngealization.

I forgot about the use of 布 *pa-s 'to spread out' to write 'cloth' (a borrowing from an Austroasiatic language: cf. Katu [Kantu dialect] kapaːs 'cotton', Kuy kpah 'cloth', and Sanskrit kārpāsa- 'cotton', also an AA borrowing) which fits my hypothesis of an earlier *k- in 'to spread out', a native word that happened to sound like 'cloth'. The *k-p-word was later reborrowed with disyllabic spellings:

幏布 *kæh-pɑh 'cotton' (c. 100 AD); is the first *-h for foreign *-r-, or was this spelling coined by someone who still had *kr- in 幏: *krɑh-pɑh?

古貝 *kɔˀ-pɑɕ 'cotton' (c. 430 AD)

See Schuessler (2007: 173) for further discussion, though he does not reconstruct *k- in the Old Chinese words for 'cloth' or 'to spread out'.

¹There is no Chinese word 攵 'action with hand'; the gloss refers to the use of 攵 *(r-)pʰok 'to beat' as a component in other characters. (The word 'to beat' is more commonly written 撲 which is not a component in other characters.)

18.6.11.23:59: DID SAEK SHIFT *Z- UNDER VIETNAMESE INFLUENCE?

Last night I stumbled upon found this passage in Pittayaporn (2009: 296):

In Saek, *z- became /j-/ merging with PT *ˀj-, probably due to influence from North-Central Vietnamese, where original *z- has become /j-/ (Alves 2007).

Northern Vietnamese has /z/ corresponding to /j/ in central and southern Vietnamese. I think Saek would be or would have been in contact with central Vietnamese. (It's not clear if there are Saek villages in Vietnam anymore.)

One might conclude that the north preserves a /z/ that became /j/ elsewhere. This would then be parallel with Saek. But I am not sure that is the case. Here are the data:

By 'northern' I mean Hanoi and Vinh (the latter is north central); 'nonnorthern' refers to Huế (at the center) and Saigon. (I don't want to say 'south' because Huế is certainly not in the south.)

Capital letters stand for obstruents with unspecified voicing: e.g., *C could be voiceless *c or voiced *ɟ.

Hyphens before consonants indicate the presence of an unspecified presyllable: e.g, *-C- represents *c or voiced *ɟ. preceded by a presyllable.

Exactly what the Middle Vietnamese spellings gi- d- r- stood for is not certain. I can only say that none of those three consonants were /z-/ or /j-/. I think it's possible that gi- and d- became /j-/ without a *z-phase. But maybe Saek is evidence for such a phase.

Or is it? The /z-/ of Vietnamese postdates the 17th century and long postdates the devoicing of original *voiced obstruents (possibly by the late first millennium AD). On the other hand, Saek *z- is original. Did Saek have *z- and a full set of voiced obstruents as late as the 18th century - almost a thousand years after Vietnamese devoiced its voiced obstruents?

6.14.2:21: I don't think what I wrote above is clear. Let me try again.

Phases of Vietnamese

Vietnamese consonants can be said to have gone through five phases which I will illustrate with hypothetical examples for simplicity:

Phase 1: Early Old Vietnamese:

presyllables present

no tones

no lenition

phonemic voicing in obstruents

I am not sure Early Old Vietnamese ever had *(d)z-. It is perhaps telling that Early Middle Chinese 字 *dzɨʰ 'written character' was borrowed as ́*ɟɨːʰ (now chữ) rather than as †zɨːʰ which would have become †tữ. Later Early Middle Chinese 字 *dzɨʰ became Late Middle Chinese 字 *tsɨ̣ and was borrowed again into Vietnamese; see phase 3 below.

Phase 2: Middle Old Vietnamese:

*-r- > *-r̥- after a voiceless initial

subphonemic tones conditioned by voicing before main vowel: *voiceless > unmarked ngang tone, *voiced > grave accent for huyền tone

tones conditioned by final consonants may date between phase 1 and phase 2

Phase 3: Late Old Vietnamese:

voicing (lenition) of medial obstruents: *-t- > *-d-

*-r̥- > *-ʂ-

devoicing of voiced obstruent initials

words formerly distinguished by obstruent voicing now distinguished only by tone which had become phonemic

Late Middle Chinese 字 *tsɨ̣ 'written character' (with a devoiced initial) was borrowed as ́*sɨ̣ː (now tự). (For simplicity I use a Vietnamese tone mark even for Late Middle Chinese.)

Phase 4: Middle Vietnamese:

presyllables lost

*Cʂ- > s- /ʂ/

Drag chain *s- > *t- > /ɗ/

Italicized forms are 17th century spellings; those spellings of consonants remain in use today. đ is /ɗ/, but the phonetic value of d is uncertain. [d] is the simplest interpretation, but [dʲ] and [ð] are also possible.

Phase 5: Modern Vietnamese: different reflexes of Middle Vietnamese s and d depending on dialect. s lost retroflexion in Hanoi (but not in Vinh which has /z/ like Hanoi and unlike the nonnorth dialects; Thompson 1987: 98). The picture for d is less clear. Two scenarios:

Scenario 1. All dialects shifted d to /z/, and nonnorthern dialects shifted /z/ to /j/

Scenario 2. d shifted in different ways; no shared /z/-phase

There is no doubt that Proto-Tai *z- became /j-/ in Saek. The question is whether that shift in Saek reflects the influence of Vietnamese given scenario 1. Let's suppose scenario 1 is true. Phase 4 is in the 17th century and phase 5b perhaps starts in the middle 19th century. (The last traces of Middle Vietnamese consonantism seem to disappear after the early 19th century.) So the Saek change would have to be dated between the 17th and 19th centuries. But if the Saek change were that recent, Saek would have had *z- - and presumably other Proto-Tai voiced obstruents such as *g *d *b- - as late as the 17th or even 18th century. That doesn't seem likely given that its neighbor Vietnamese had undergone devoicing prior to borrowing from Late Middle Chinese during phase 3 (circa the 10th century).

Phases of Saek

Saek has gone through some of the same changes as Vietnamese up to phase 3, though the details differ:

Phase 1: Proto-Tai:

presyllables present (rewritten here as *Cə- instead of as *C.- as in Pittayaporn's notation)

no tones

no lenition

phonemic voicing in obstruents

drag chain shift: *-t- > *-d- > *-r-; contrast with Vietnamese phase 3 in which  *-t- > *-d-; 

Phase 3:

loss of presyllabic vowels

*pər- > *pr-; *pr̥- > *pʰr-

subphonemic tones determined by initial consonant (Including presyllabic consonants unlike Vietnamese) after lenition (again, unlike Vietnamese)

To facilitate comparison with Vietnamese, I use Vietnamese tone notation: zero for tone A1 and a grave accent for tone A2.

Tones conditioned by final consonants may have developed between phase 1 and phase 3.

Phase 4:

drag chain shift: *pd- > *pr- > r-, *d- > tʰ-, *z- > j-

words formerly distinguished by initial voicing now distinguished by tone wh\ich has become phonemic

My guess is that lenition and devoicing happened independently in Vietnamese and Saek, whereas tonogenesis did not - Vietnamese phase 3 and Saek phase 3 may have been simultaneous.

Phases of Cao Bang

On 6.11, I thought Saek having *z- and other voiced consonants as late as the 18th century was improbable, but Tai languages on the Sino-Vietnamese border never underwent devoicing (PIttayaporn 2009: 110). Compare the phases of Cao Bang with those of Vietnamese and Saek:

Phase 1: Proto-Tai: same as Saek phase 1

Phase 2:

loss of presyllabic vowels

*-r- > *-r̥- after a voiceless initial (as in Vietnamese and Saek)

Phase 3:

Chain shift: *pt- > *pr̥- >  *pʂ-

subphonemic tones determined by voicing of consonant before vowel (contrast with Saek)

To facilitate comparison with Vietnamese, I use Vietnamese tone notation: zero for tone A1 and a grave accent for tone A2.

Tones conditioned by final consonants may have developed between phase 1 and phase 3.

Phase 4:

*pr̥- > *tr̥- > tʰ-

elimination of *voiceless-voiced clusters and chain shift: *pd- > *d- > dʱ-

*pʂ- > *pɕ- > pʰj-

*z- > *s- > tʰ-; *z- devoiced but this seems to be an anomaly; see my 6.13 entry; the fortition is reminiscent of Vietnamese (see Phan 2013 for examples of *s- > /tʰ/ in Vietnamese: eg., *sit > thịt 'meat'¹) but probably occurred independently much later. Phan (2013: 65) regards fortition of fricatives as "common in Southeast Asia and should not be considered a shared innovation."

tone A2 still strongly associated with voiced initials but has become phonemic due to the devoicing of *z-

Finally, for reference:

Phases of Thai/Lao

Thai and Lao never underwent lenition; medial *-t- and *-d- remain as stops today.

Phase 1: Proto-Tai: same as Saek and Cao Bang phase 1

Phase 2:

loss of presyllabic vowels

*-r- > *-r̥- after a voiceless initial (as in Vietnamese, Saek, and Cao Bang)

Phase 3: More or less represented by Thai and Lao spelling (but Lao has no <z>; *z- corresponds to ຊ <j>)

reduction of *pC- to *t- and *ɗ- (not *d-!); was there an intermediate geminate stage *tt- and *dd-?

*-r̥- > -ʰ-

subphonemic tones determined by initial consonant (Including former presyllabic consonants unlike Vietnamese)

To facilitate comparison with Vietnamese, I use Vietnamese tone notation: zero for tone A1 and a grave accent for tone A2.

Tones conditioned by final consonants may have developed between phase 1 and phase 3.

Phase 4

drag chain shift: *ɗ- > d- > *tʰ-

words formerly distinguished by initial voicing now distinguished by tone wh\ich has become phonemic

the Vietnamese notation, though convenient, is misleading, as tones A1 and A2 have undergone splits and, in Thai, a merger.

The development of tones A1 and A2 in Thai and Lao

All of the phases above are my speculations built upon the work of Gage ("Vietnamese in Mon-Khmer Perspective", 1985) and Pittayaporn (2009). The relative chronology is only approximate; some but not all changes could be reordered with the same final results.

¹The nặng tone written with a subscript dot normally indicates a *voiced initial. It is tempting to reconstruct a change *z- > /tʰ/ as in Cao Bang. But support for *z- in native words is weak. The tone may reflect a lost voiced prefix.

18.6.10.23:59: EMPHATIC SAND

Tonight I found the section on the Middle Korean emphatic particle za at random in Lee and Ramsey (2011: 194). The earliest attestations of it I can find in Old Korean are in two 鄉歌 hyangga

毛等居叱沙

*motʌn kəs sa

'all thing EMPH'

- 慕竹旨郎歌 （c. 700)

一等沙

*hʌtʌn sa

'one EMPH'

- 禱千手觀音歌 (c. mid-8th century)

It occurred to me that the 'sand' spelling of that particle¹ obviously must predate the lenition of *s to Middle Korean z.

If a *z-pronunciation had existed in Old Korean, it could have been spelled with Middle Chinese

嵯嵳𣩈㽨瘥𥰭䑘艖蒫醝䰈鹺䴾齹虘蔖䠡䣜躦𪘓 *dza

or 邪䓉耶椰瑘𥯘鎁釾𦭿𦰳斜䔑擨 *ziæ².

(There was no Middle Chinese syllable *za. This gap is not accidental. I should look into it.)

It turns out that 邪 'evil' is attested as a phonogram in Old Korean hyangga, but 俞昌均 Yu Chhang-gyun (1994: 76) interprets it as a symbol for *ra (cf. its possible Old Chinese reading *la in Schuessler 2009: 56). There have been many attempts to reconstruct the pronunciation of Old Korean. Has anyone interpreted 邪 as *sa (possibly tempted by its modern Sino-Korean reading sa) or *za? I don't have any other sets of hyangga readings on hand. Another thing to look into when I get the chance.

¹6.11.21:29: It never occurred to me to use Unicode superscript numerals for endnotes until now. No more long strings of asterisks.

It's theoretically possible that the 'sand' spelling in this text postdates the 8th century, as these poems survive in 三國遺事 Samguk yusa (1285) whose earliest surviving copy is from 1512. Even if these poems are actually from c. 700 AD, their spellings could have been altered in the centuries between then and 1512.However, I know of no other evidence pointing toward some other Ur-spelling of the emphatic particle. The 口訣 kugyŏl phonogram for *sa ~ *za is 氵 which is almost certainly an abbreviation of 沙 'sand', the most common sa-character with the left-hand component 氵 'water'. Kugyŏl manuscripts from the Koryŏ dynasty (918-1392) predate 1512; one need not worry about potential errors in their transmission.

²6.11.23:44: Nearly all of these characters are rare and therefore not likely candidates for phonograms which tended to be high-frequency characters. So one might argue that the Old Korean particle was *za but not written as such because there was no high-frequency characters with a similar reading other than 邪 *ziæ 'evil' which was already being used for *ra if Yu (1994) is correct. However, if *s had already lenited to *z in Old Korean, I would expect to see other phonogram spellings unambiguously reflecting lenition. But I know of none offhand. Although one might argue that *s lenited before other consonants, that possibility could only be confirmed if there were *(d)z-spellings of later z-words. No such spellings seem to exist.

The only *(d)z-phonogram in Yu's (1994: 75-78) catalog of phonograms in hyangga are the aforementioned 邪 *ziæ 'evil' and

齊 Middle Chinese *dzej 'equal' : Yu's Old Korean *tsjə (my *tse)

which, like 邪 *ziæ 'evil', does not represent an Old Korean syllable corresponding to a Middle Korean z-syllable. So if Old Korean already had *z-syllables, they were not written with Chinese *(d)z-characters and cannot be detected.

That Chinese dialect had a reflex of Middle Chinese *ɲ- that corresponds to z in Middle Korean Sino-Korean readings. But there was no Middle Korean Sino-Korean reading †za. So it seems Old Koreans had no good options for writing *za if they had such a syllable - and I still don't think they did.

(The questions of what that Chinese dialect's reflex of *ɲ- was and how it was borrowed into Old Korean - as *z- or as something else that became z- in Middle Korean - remain open. The simplest solution is to assume that Chinese dialect had something like the *ž- of Liao Chinese. This was borrowed into Old Korean as *z-, a consonant originally only in borrowings. Later, Middle Korean lenited *s in native words, resulting in a new /z/ that shared the fate of the old borrowed one: both /z/ soon disappeared from the Seoul dialect. [But does any Korean dialect today have a trace of /z/ in Sino-Korean words?)