15.11.7.23:59: COMEBACK

I had a good reason to not post for the past two days - and to have been emphasizing emphasis in recent posts.

On Thursday and Friday, I participated in an academic conference for the first time since 2003. Guillaume Jacques wrote a report about it. My contribution was "Old Chinese Type A/Type B in Areal Perspective".

Maybe I should have renamed my talk "Typological" instead of "Areal". But apart from one mention of Salish, I did stick to Eurasia and Egypt which is right next door.

You can download my PowerPoint presentation here. I want to supplement it with a table of my stages of (pre)syllabic erosion in Old Chinese:

Notes on the phases:

1. Roots were originally disyllabic with the same six vowels in both first and second syllables. Maybe either syllable could have been stressed at this point:

2. First syllables of disyllables which were unstressed (and/or lose stress?) became presyllables with less vocalic diversity than the stressed syllables that followed them. Six vowels were reduced to an Austronesian-like four-vowel system. Presyllabic *i may have left traces in some syllables that distinguish it from other high series presyllabic vowels. With the exception of those syllables, it is impossible to determie whether a high vowel syllable had *i, *ə, or *u without non-Chinese evidence. Hence I consider this stage to be pre-Chinese.

3. Low vowel *Că- presyllables (which I have been writing as *Cʌ- on my site) conditioned emphasis:

*Că-Ca > ́*Că-Cˁa

All high series presyllabic vowels merged into *ɨ̆ (which I have been writing as *ɯ on my site). No emphasis developed after *Cɨ̆-presyllables:

*Cɨ̆-Ca > *Cɨ̆-Ca (no change)

4. The two presyllabic vowels merged into schwa. Emphasis is no longer predictable and beomes phonemic.

*Că-Cˁa > *Cə-Cˁa /CəCˁa/

*Cɨ̆-Ca > *Cə-Ca /CəCa/

5. The presyllabic vowels are lost, and presyllables become preinitials.

*Cə-Cˁa > *CCˁa

*Cə-Ca > *CCa

6. Preinitials were lost:

*CCˁa > *Cˁa

*CCa > *Ca

Presyllables could be in various degrees of reduction at any given time: e.g., in the earliest period, unstressed *Ce (stage 1) could have been optionally pronounced as *Ci (stage 2). In Middle Old Chinese, *Cə- and its reduction *C- coexisted side by side. This is analogous to the different degress of reduction of unstressed vowels in English ranging from spelling-like pronunciations to schwa or even zero. In a few cases, unstressed syllables can disappear entirely: e.g., because [bɪˈkʌz] ~ [bɨˈkʌz] ~ [bəˈkʌz] ~ [bkʌz] ~ cause [kʌz].

11.8.3:09: Applying my stages to those forms of because:

Stage 1/2: [bɪˈkʌz]

Stage 3: [bɨˈkʌz] (frontness neutralization)

Stage 4: [bəˈkʌz] (height neutralization)

Stage 5: [bkʌz] (loss; > [pkʌz] with voicing assimilation?)

15.11.4.22:33: CONSONANTAL VS. VOCALIC THEORIES OF CHINESE EMPHASIS

David Boxenhorn asked me about the implications of consonantal and vocalic theories of Chinese emphasis.

First, let me define how I interpret 'consonantal' and 'vocalic' in this context.

I advocate what could be called a vocalic theory in the sense that emphasis was ultimately conditioned by low vowels in what I call Early Old Chinese. But in Middle Old Chinese, some of those low vowels were lost, and the locus shifted to the consonant (though emphasis was phonetically present in the following vowel if not the coda). Then in Late Old Chinese, emphasis was lost, and previously predictable vocalic allophones after emphatic and nonemphatic consonants became phonemic:

Early Old Chinese (no phonemic emphasis): /Cʌpi/ [Cʌpi] > [Cˁʌˁpˁiˁ]

Middle Old Chinese (phonemic emphatic consonants): /pˁi/ [pˁiˁ] > [pˁeˁiˁ]

Late Old Chinese (no phonemic emphasis): /pei/ [pei]

So my theory could also be called consonantal or even syllabic depending on which period one is looking at and whether one is looking at phonemes or allophones.

Baxter and Sagart (2014: 69) reconstruct 36 emphatic consonants. 35 of them have nonemphatic counterparts; the 36th, *ʔʷˁ, lacks a nonemphatic counterpart *ʔʷ. Conversely, there are no nonemphatic consonants lacking emphatic counterparts. The near-total symmetry between the emphatic and nonemphatic subsets of consonants is striking; it is reminiscent of the near-total symmetry between

- the emphatic and nonemphatic subsets of the phonetic (but not phonemic!) inventory of Cairene Arabic

- the palatalized and nonpalatalized consonants in Russian

(Norman 1994, the originator of the Chinese emphatic theory, regarded Russian nonpalatalized consonants as pharygealized: i.e., what I call 'emphatic'; in any case, the palatalized consonants are not simply nonemphatic.)

If we knew nothing about Slavic language history, we might notice how other Slavic languages have smaller sets of palatalized consonants or even no palatalized consonants at all (e.g., Serbo-Croatian), conclude that Russian is conservative, and project the Russian system back into Proto-Slavic. But that would be a mistake, as we know that palatalization in Slavic was secondary and conditioned by front vowels. The short front vowel */ĭ/ was lost, and palatalized consonant allophones that had once been before it and other front vowels were reinterpreted as phonemes:

*/Cĭ/ [Cʲɪ] > /Cʲ/ [Cʲ] (after loss of short */ĭ/)

*/Ci/ [Cʲi] > /Cʲi/ [Cʲi] (nonshort */i/ retained)

The large phonetic inventory of Cairene Arabic emphatics is due to emphatic spread from five emphatic phonemes /tˁ dˁ sˁ zˁ rˁ/ and the vowel /ɑ/ (Youssef 2014); there is no need to assume that Cairene Arabic preserved a far larger inventory of emphatics than Classical Arabic. (Note, however, that emphatic /rˁ/ and a back /ɑ/ phoneme distinct from /a/ do not exist in Classical Arabic. The origins of these two phonemes are deserving of investigation. I do not assume that all non-Classical traits of modern Arabic varieties are innovations; some could be retentions of traits conserved in the nonstandard dialects of Arabic conquerors but lost in the standard.)

The precedents of Slavic and Cairene Arabic make me hesitant to project the gigantic Old Chinese inventory back into a higher node or even Proto-Sino-Tibetan. There is, to the best of my knowledge, no attested Sino-Tibetan language with such an inventory. Emphatics have not been reported in any variety of Chinese. (Perhaps they are waiting to be detected; sometimes we are blind to the unexpected.) Was the Old Chinese consonant system the last remnant of a huge proto-system that was simplified everywhere else?

I don't think so. I have never seen so many emphatics in any other proto-language. I already wrote about Afroasiatic emphatics at some length last week, so here I will merely state that Ehret's (1995) Proto-Afroasiatic has only seven voiceless emphatics which mostly form 'triads' with nonemphatic voiced and voiceless consonants:

*p' *t' *tl' *s' *c' *k' *kʷ' vs. *p *t (no *tl) *s *c *k *kʷ vs. *b *d *dl *z *j *g *gʷ

In Proto-Afroasiatic, emphatics were ejectives which generally later became pharyngealized in Arabic. (Mehri emphatics seem to be in transition.)

Johanna Nichols' (2003) Proto-Nakho-Dagestanian (Northeast Caucasian) also has such triads:

There is a phonetic reason for triads instead of tetrads with voiced as well as voiced ejectives: voiced ejectives do not and cannot exist.

The presence of voiced as well as voiceless emphatics in Old Chinese indicats that Old Chinese was not like Arabic - that its pharyngealized consonants were not from earlier *ejectives.

Interestingly, Nichols does not reconstruct pharyngealized consonants in P even though they are present in Archi and Rutul. I have not found pharyngealized consonants in Nichols' lists of Archi and Rutul reflexes. That suggests the pharyngealized consonants of those two languages are rare and possibly secondary.

I think Old Chinese pharyngealized consonants are also secondary. But why do I think low vowels conditioned pharyngealization? The low vowel a is the syllabic counterpart of the pharyngeal approximant ʕ (Pulleyblank 1997 and Operstein 2010: 177); it is to ʕ what i, ɨ, and u are to j, ɰ, and w. So I expect an Old Chinese *a-like low vowel to condition pharygealization in neighboring segments - much as back /ɑ/ does in Cairene Arabic - particularly given that northern neighbors of Chinese and their neighbors have harmonic systems in which vowel and consonant qualities are intertwined to some extent. I write the unstressed low vowel triggering pharyngealization as *ʌ, borrowing the symbol for the conventional interpretation of arae a 'bottom a' (ㆍ), the minimal low vowel of Middle Korean. I could have written it as *ă, but I wanted a symbol that was easy to distinguish from *a and that reflected my hypothesis that Chinese once had height harmony like Middle Korean.

My vocalic emphatic theory predicts that all Middle Old Chinese words with emphatic consonants once had emphasis-triggering low vowels. I used to think that Old Chinese *e and *o belonged to the same height class as *a (as they do in my reconstruction of Old Korean) and also triggered emphasis, but I am less certain of that. Maybe *e and *o-syllables also needed a preceding true low vowel to become emphatic:

*(Cʌ)Ce > *Cˁe (no presyllable needed) or *CʌCe > *Cˁe but *Ce > *Ce?

I could reinterpret *e and *o as *aj and *aw or *ja and *wa with *a (cf. Pulleyblank's *ə/*a two-vowel reconstructions of Old Chinese), but that has costs: e.g., it forces me to reinterpret *-ew as *-aɥ, etc.

Mid vowels aside, my theory predicts that Middle Old Chinese words of the type emphatic consonant + higher vowel (*Cˁi / *Cˁə / *Cˁu) should be from earlier *CACI (*A = low vowel and *I = high vowel) sequences. If Chinese borrowed such words from a polysyllabic language (e.g., Austronesian) or vice versa, the polysyllabic sources/borrowings should have begun with low vowels at the time of borrowing. Some Austronesian words with possible Old Chinese relatives pose problems for my theory; I will deal with them next time.

Conversely, if we assume that Old Chinese emphatic consonants are not innovations, then Old Chinese borrowings may preserve emphasis that was once present in the donors. The trouble is that there is no independent or internal evidence to suggest that Austronesian, Kra-Dai, Hmong-Mien, etc. had emphatics. If Old Chinese 狗 *Cə.kˁro 'dog' is a borrowing from Proto-Hmong-Mien *qluwX 'id.', why does it have an emphatic? My vocalic theory could account for the emphasis as being from a low presyllabic vowel and/or the low series vowel *o. (That would be the case even if the direction of borrowing were reversed.)

Lastly for now, my theory predicts that vowel heights for Old Chinese prefixes can be recovered if correlations between prefixes and emphasis can be made. I have yet to test this prediction.

On the other hand, the consonantal theory predicts no correlation between prefixes and emphasis since prefixes are invariably reconstructed with nonemphatics and can occur before both emphatic and nonemphatic-initial roots.

15.11.3.23:59: A CEREBRAL COUNTEREXAMPLE? OLD CHINESE 首 'HEAD'

Last night, I wrote that there was

a specific word that made me question my old uvular theory - possibly even before I saw Baxter and Sagart's uvular proposal years ago.

That word is 首 'head' which Baxter and Sagart reconstructed as *l̥uʔ. I regard that form as Middle Old Chinese.

Last week, I wrote,

the [Early] Old Chinese initial of 'head' may be from *Kl- [...] if the word is related to Proto-Austronesian *quluh and/or Proto-Tai *kraw^C (Pittayaporn 2009: 323) / *kləw^C (Li Fang-Kuei 1977). Proto-Hmong-Mien *kləuX 'road' (Ratliff 2010: 264) is a loan from Old Chinese 道 containing 首 as a phonetic. If 道 had an initial stop, perhaps 首 did too.

*quluh > *quluʔ > *qɯluʔ > *qluʔ > *l̥ˁuʔ > *l̥ˁouʔ > *tʰouʔ > *tʰauʔ

(The similarity to 頭 Mandarin [tʰou] and Cantonese [tʰau] 'head' is coincidental; 頭 and 首 are unrelated words.)

But the actual Late Old Chinese form of 首 was *ɕuʔ with a fricative from nonemphatic *l̥- and a high vowel that was never lowered by emphasis.

Maybe 首 has nothing to do with Proto-Austronesian *quluh and originally had a *k- as in the Proto-Tai word and the Proto-Hmong-Mien borrowing of its near-homophone 道 'road'.

Or maybe they are tied together.

Baxter and Sagart do not reconstruct *q- as a preinitial or in presyllables. In the dialect reconstructed by Baxter and Sagart, *q- is not automatically emphatic, and I speculate that nonemphatic preinitial/presyllabic *q- fused with nonemphatic *l into voiceless nonemphatic l̥.

If Proto-Tai *kraw^C is a loan from Old Chinese, it could be from a dialect in which the presyllabic vowel lowered after a uvular that then fronted and became emphatic (due to lower vowel-emphatic harmony):

*qul- > *qɯl- > *qʌl- > *kˁʌl- > *kˁl- > *kˁr-

That emphasis conditioned the lowering of *-u to *-aw.

Proto-Hmong-Mien *kləuX 'road' may be from yet another dialect - one which did not shift *kˁl- to *kˁr-:

道 *qʌluʔ > *kˁʌluʔ > *kˁluʔ > *kˁlouʔ > *kləuʔ

The rhyme *-əuʔ may be an intermediate stage in bending between *-ouʔ and *-auʔ.

*qʌluʔ > *kˁʌluʔ > *kˁʌlˁuʔ > *lˁuʔ > *douʔ > *dəuʔ (?) > *dauʔ

Northern Min forms such as Jianyang lau have a secondary l- that is an intervocalically lenited *-d- and not a retention of an original lateral:

*kˁʌluʔ > *kˁʌlˁuʔ > *kʌduʔ > *kʌdouʔ > *kʌdəuʔ (?) > *kʌdauʔ > *kʌlauʔ > lau

(11.4.0:24: The chronology of lenition and presyllabic loss relative to vowel changes is unknown.

For convenience, I have retained the low presyllabic vowel throughout the derivation, though it may have merged with the high presyllabic vowel at some point before the presyllable was lost. Once emphasis became phonemic for the initials of core syllables, the height of presyllabic vowels could have been neutralized without any loss of information:

I wrote "vowel" at the top of the column for stage 3 rather than "consonant and vowel" because in many cases the distinctions in stages 1 and 2 leave no traces on the initial: e.g.,

*kʌpuʔ > *kəpˁuʔ > *kəpouʔ

*kɯpuʔ > *kəpuʔ > *kəpuʔ

Emphatic *pˁ- and nonemphatic *p- have merged into nonemphatic *p- in stage 3; only the vowels are distinct.

On the other hand, emphatic *lˁ- and nonemphatic *l- did not merge with each other. The former hardened and merged with emphatic *dˁ- as nonemphatic *d-, whereas the latter weakened to *j-.)

15.11.2.23:59: DID BACK CONSONANTS CONDITION EMPHASIS IN OLD CHINESE?

David Boxenhorn asked me if emphasis in Old Chinese (OC) could be conditioned by presyllabic back consonants instead of low vowels: e.g.,

*Q.C- > *Cˁ- rather than *Cʌ.C- > *Cˁ-

My old answer was that both could condition emphasis, though low vowels in presyllables and syllables proper were the primary sources. I thought that

- uvular (and pharyngeal?)-initial syllables

- and perhaps also uvular-initial presyllables

conditioned emphasis (indicated below with *ˁ). Contrast the development of uvulars with velars in this earlier reconstruction:

Low vowels: emphasis and secondary uvulars

*qa > *qˁa vs. *ka > *qˁa

*qe > *qˁe vs. *ke > *qˁe

*qo > *qˁo vs. *ko > *qˁo

*Cʌ.CV > *CˁV

e.g., *Cʌ.kV > *qˁV

High vowels: emphasis with original uvulars but no emphasis with velars

*qə > *qˁə vs. *kə > *kə

*qi > *qˁi vs. *ki > *ki

*qu > *qˁu vs. *ku > *ku

*Cɯ.CV > *CˁV

except: *qɯ.CV > *CˁV

However, Baxter and Sagart's OC reconstruction contrasts emphatic and nonemphatic uvulars: e.g., *q vs. *qˁ, etc. Such a distinction is rare in the world's languages; it is currently in Archi and Rutul in the Caucasus, far from China. I would rather not reconstruct an exotic distinction - at least not at an early level - so I prefer to reconstruct primary nonemphatic uvulars and secondary emphatic uvulars:

Original nonemphatic uvulars and velars retained after higher vowels:

*qə > *qə; cf. *kə > *kə

*qi > *qi; cf. *ki > *ki

*qu > *qu; cf. *ku > *ku

*Cɯ.qV > *qV; e.g., *Cɯ.kV > *kV

Secondary emphatic uvulars and velars developed after lower vowels:

*qa > *qˁa; cf. *ka > *kˁa

*qe > *qˁe; cf. *ke > *kˁe

*qo > *qˁo; cf. *ko > *kˁo

*Cʌ.qV > *qˁV; e.g.,*Cʌ.kV > *kˁV

Such a complex system eventually broke down when most of the uvulars left gaps to be filled by emphatic velars which backed:

*q(ˁ)- > *ʔ-; new *q- from *kˁ-

*q(ʰˁ)- > *x- (usually*; phonetically [χ] before lower vowels?); new *qʰ- from *kʰˁ-

*ɢˁ- > *ɢ-; new *ɢ- from *gˁ-

*ɢ- > *ʁ- > *ɰ- > *j-; new *ɢ- from *gˁ-

The new system only had one series of uvulars and no phonemic emphasis (though emphasis may have persisted at the phonetic level):

(11.3.1:09: The three stages contain roughly the same amount of complexity distributed in different ways. The locus of a binary distinction traveled rightward over time:

Early Old Chinese: low vs. high vowel presyllables

*/Cʌ.pi/ [Cʌ.pi] > [Cˁʌˁ.pˁɪˁ]

*/Cɯ.pi/ [Cɯ.pi]

Middle Old Chinese: emphatic vs. nonemphatic core syllable initials; Baxter and Sagart's reconstruction corresponds to this stage

*/pˁi/ [pˁəˁɪˁ]

*/pi/ [pi]

Late Old Chinese: lower vs. higher vowels

*/pei/ [peɪ]

*/pi/ [pi]

I originally wanted to use back consonants in the examples above, but the uvular-velar shifts would complicate the three-way contrast.)

That's the big picture. Tomorrow I'll look at a specific word that made me question my old uvular theory - possibly even before I saw Baxter and Sagart's uvular proposal years ago.

*11.3.0:45: See Baxter and Sagart (2014: 102-105) for less common reflexes of *q(ʰˁ)-:

- Middle Chinese *ɕ- via *x- before front vowels including secondary fronted *a

- Proto-Min *kʰ-

- Middle Chinese *ʈʰ- from *qʰr- in eastern dialects

For simplicity, I have only listed reflexes of *q(ʰˁ)- without preinitials or presyllables at the Middle Old Chinese level. Such preceding elements resulted in even more reflexes: e.g., Middle Old Chinese *t.qʰ- became Late Old Chinese *tɕʰ- (see Baxter and Sagart 2014: 160; their Middle Chinese initials are almost identical to my Late Old Chinese initials).

15.11.1.15:41: WHY DO OPPOSITES ATTRACT? THE MEHRI DEFINITE ARTICLE Ḥ(Ə)-

Last week I mentioned the Mehri definite article a- which only occurred before emphatics and voiced consonants.

Mehri has another definite article ḥ(ə)- which only occurs before glottal stops, f-, and voiced nonemphatics:

Why does voiceless pharyngeal-initial ḥ(ə)- occur before voiced nonpharyngealized consonants? This distribution is what I call diachronic detritus; the synchronically inexplicable result of an earlier sound change. Rubin (2010: 71) wrote that

Many of the nouns with the definite article ḥ(ə)- have an etymological initial ʾ- [i.e.., glottal stop], which is sometimes reflected in the long ā of the definite article ḥā-.

Exceptions have y- from *y- (e.g, yūm 'days') or are due to analogy.

So ḥ(ə)- may have originally only been before *ʔ- which was lost before voiced consonants and *y-. Maybe this was the earlier distribution of articles:

1. *ʔa-

> a- before emphatic and voiced consonants

a- remains in front of ʔ- from voiced *ʕ-

> *ʔ- > zero before voiceless consonants other than *ʔ-

2. *ḥə-

> before *ʔ- (> zero in *ʔ- + voiced consonant clusters) and *y-

Why did *y-words have both *ʔa- and *ḥə-?

Was *ḥə- lost before *ʔ- + emphatic/voiceless consonant clusters?

Why is ḥə- before voiceless nonglottal f-? (I presume Mehri f is from *p- since there is no p in Mehri; was the lenition of *p due to Arabic influence?)

ḥə-f- would not be surprising if that sequence were from *ḥə-ʔw- whose cluster fused into -f-, but such a word would have a nondefinite form with (ʔ)w-, and Rubin did not note any (ʔ)w- ~ ḥə-f-alternations.

That scheme omits the plural definite article hə- ~ ha- which only occurs

- before nouns with the voiceless fricatives s- and ɬ̠- and a CCōC pattern

- the high-frequency nouns həbɛ̄r 'the camels' and hərbāt 'the companions' (sg. ərbāt)

Are the last two forms relics of a period when h-articles were more common? They have nothing in common with the other h-article forms apart from ending in long vowel + consonant sequences.