15.10.31.22:33: DID *I CONDITION *A-FRONTING IN OLD CHINESE?

Two nights ago I reconstructed Old Chinese (OC) 射 with presyllabic *-i- without comment:

*mi-lak (*m-ljak?) 'to hit with an arrow' ~ *mi-lak-s (*m-ljak-s?) 'to shoot with a bow, archer'

One mystery in OC phonology is why *a behaves in two different ways in syllables of the type *T(s)a(k/j): e.g.,

蛇 with two readings:

Early OC *mlaj 'snake' > Late OC *ʑjæj with a front vowel

Early OC *laj (second syllable of 'compliant') > Late OC *jɨaj with a nonfront vowel

Here are some approaches to the problem:

- Starostin (1989) reconstructed *a and *ia; the later is equivalent to my *ja above.

- A few years ago, I reconstructed *a and *æ.

- Baxter and Sagart (2014) used the notation *a and *A, stating that "Our *A is not intended as a seventh vowel; it is an explicitly ad hoc notation that basically means 'a case of OC *-a which for as yet unexplained reasons becomes MC -jae instead of MC -jo [i.e., the regular reflex].' "

My fourth approach involves the frontness of *-i- moving into the following syllable:

Early OC *mi-laj > *mljaj > *nɮjaj > *ɮjaj > Late OC *ʑjæj 'snake'

Early OC *laj > *lɨaj > *ɮɨaj > *ʑɨaj > Late OC *jɨaj (second syllable of 'compliant')

There was a chain shift *nɮ- > *ɮ- > *ʑ- > *j-. Prenasalization shielded *nɮ- from weakening to *ʑ- like *ɮ-.

Normally *a-breaking is conditioned by a high-vowel presyllable. 'Compliant' may have had only one such syllable, and the remainder of the word harmonized with it:

委蛇 Early OC *Cɯ-q(r)oj-laj > Late OC *ʔuoj-jɨaj

I assume presyllables had short vowels that were reduced versions of the six in main syllables: *i, *ə, *u, *e, *a, *o. The last three merged into a low vowel I write as *ʌ. The first three merged into a high vowel I write as *ɯ except in a few cases where later fronting is a trace of *i. Perhaps at one point there was a triangular three-vowel system in presyllables somewhat like that of Pacoh: *i vs. a nonfront high vowel *ɯ vs. *ʌ.

Unfortunately, I do not have any evidence for *-i- other than the fronting in later OC. Why was this fronting in such a specific environment (*coronal + *a + *-k or *-j)? Why didn't fronting occur in *-ŋ final syllables which usually develop like *-k syllables? *-ŋ has more in common with *-k than *-j, yet fronting only occurred before the latter two codas.

15.10.30.20:13: WHAT WAS THE RANGE OF OLD CHINESE ROOSTERS?

The Chinese character 酉 currently only represents the word for the tenth Earthly Branch conventionally translated as 'rooster' in English.

The character is a drawing of a wine vessel and is used as a semantic element in characters for words with alcoholic semantics. Moreover the words 酉 for 'tenth Earthly Branch' and 酒 'wine' rhyme. So it seems likely that 酉 was originally devised for a word 'wine vessel' that was cognate to 酒 'wine'. However, are there any texts in which 酉 means 'wine vessel'?

Premodern dictionaries list three other definitions for 酉:

1. 就 (Shuowen, c. 100 AD) which has many possible translations (e.g., 'to go to'); I don't know which one was intended

2. 飽 (Guangyun, 1008 AD) 'satiated'

3. 老 (Guangyun, 1008 AD) 'old'

How old are those definitions? Is there textual support for them?

All these definitions had *u in Early Old Chinese (EOC) like 酉 and 酒. Did 酉 represent five unrelated near-homophones that later greatly diverged in Late Old Chinese (LOC)?

1. 酉 'tenth Heavenly Branch': EOC *N-ruʔ (Baxter and Sagart 2014: 372) > LOC *juʔ

2. 酉 'wine vessel', possibly cognate to 酒 EOC *tsuʔ (Baxter and Sagart 2014: 347) > LOC *tsuʔ

3. a cognate of 就 '?': EOC *[dz]u[k]s > LOC *dzuh

4. a cognate of 飽 'satiated': EOC *pʌ-ruʔ > *prˁuʔ > LOC *pɔuʔ > *pæuʔ

5. a cognate of 老 'old': EOC *Cʌ-ruʔ > *rˁuʔ > LOC *louʔ > *lauʔ

Why was 酒 EOC *tsuʔ 'wine' written with a phonetic 酉 EOC *N-ruʔ 'wine vessel'? Was a common rhyme and similar semantics sufficient to justify the choice of a phonetic with completely different initials? And what is Baxter and Sagart's justification for reconstructing a nasal prefix in 酉 EOC *N-ruʔ?

But what if my OC palatal hypothesis is correct, and 酉 'wine vessel' and 酒 were cognates with similar initials? (Also see these two follow-up posts.)

酉 'wine vessel': EOC *N-cuʔ > *ɟuʔ > LOC *juʔ

酉 'tenth Heavenly Branch' might have had *N-c- or original *ɟ-

酒 EOC *cuʔ > LOC *tsuʔ

According to the palatal hypothesis, 就 might have been EOC *N-Cu(k)-s (*C = an unknown palatal) which would have been a good phonetic match for 酉 *ɟuʔ.

However, there is no reason to believe that 飽 and 老 ever had any palatals. Nor is there any reason to reconstruct palatal prefixes. So why would 'satiated' and 'old' be written with a palatal phonetic 酉? Did the use of 酉 for *r-words reflect a dialect or dialects in which *ɟ (already lenited to *j?) and *r had converged (or even merged?): e.g., to *ʑ and *ʐ? (Starostin reconstructed Eastern Han *ʑ- as the source of Postclassic and Middle Chinese *j-.) It would not be unreasonable to write forms like 飽 *pʐˁuʔ 'satiated' and 老 *ʐˁuʔ 'old' as 酉 *ʑuʔ. Such forms would not be in early texts in which *ɟ and *r were distinct. If 酉 represented 'satiated' and 'old' in early texts, I would have to abandon this explanation.

APPENDIX: A KRA-DAI VERSION OF THE PALATAL HYPOTHESIS?

While researching this post, I rediscovered a 2008 post in which I compared the Common Tai ʔj- : Jiamao Hlai tsh- correspondence to the *j- ~ *ts- alternations in Middle Chinese that I derive from Old Chinese palatal *ɟ- ~ *c-. Last year I proposed an implosive palatal stop *ʄ- as a source of later Tai ʔj-. Norquest (2007: 13) regarded Jiamao Hlai as a "non-Hlai language which has been in close contact with Hlai", so Jiamao Hlai tsh- may reflect a Hlai obstruent at the time of borrowing. Could that early Hlai obstruent have been *ʄ-? Can *ʄ- be reconstructed at the Proto-Kra-Dai level?

According to Norquest (2007: 338), Jiamao tsh- is in borrowings that had Proto-Hlai *tɕh-; even earlier Jiamao borrowings of words with that initial phoneme have ts- from pre-Hlai *c-. Is pre-Hlai *c- a reflex of a Proto-Kra-Dai *ʄ-? Unfortunately, I can't find the only example of a Common Tai ʔj- : Jiamao Hlai tsh- correspondence that I have (CT ʔjuu B : J tshu 'to be') from Shintani (1991: 2) in Norquest (2007).

Norquest (2007: 348) reconstructed pre-Hlai *lj- as a source of Jiamao unaspirated ts- in loanwords. That reminds me of how I used to reconstruct 酉 as EOC *luʔ though its phonetic series had *ts- in Middle Chinese (e.g., 酒 Middle Chinese *tsuʔ 'wine'). The difference is that pre-Hlai *lj- was borrowed into pre-Jiamao as *lj- which hardened to *dʑ- and devoiced to ts-, whereas there is no reason to believe that the affricate of 酒 Middle Chinese *tsuʔ is the product of hardening and devoicing; those features must be projected back into the Old Chinese reading of 酒 though they conflict with all evidence pointing toward a voiced initial for its phonetic 酉.

15.10.29.22:46: WHY WAS OTTOMAN TURKISH ض <Ḍ> PRONOUNCED IN TWO DIFFERENT WAYS?

To answer that question, I looked at three old textbooks:

Hagopian (1907: 9): "It is generally pronounced as a hard [= emphatic] z, but sometimes as a hard d."

Barker (1854: 2): "d hard, and sometimes z"

Vaughan (1709: xxvi, 2): ’z without any reference to a stop (Oddly, ظ <ẓ> was romanized as an affricate ’dz!)

Was there a new layer of Arabic loans after1709 in which Arabic was borrowed as d instead of z? I can't imagine why that would be the case, so my guess is that Vaughan left out the d-variant, and that the variation reflects two or more strata of borrowing from Arabic before 1709.

Looking at Embarki (2013: 25-26) to go back a millennium in Arabic itself, I see that Al-Khalīl (d. 786) grouped ض <ḍ> with شج <sh> and ج <j> as 'arched' (شجريه shajriyya) in Kitāb al-`ayn. That implies <ḍ> sounded somehing like sh and j. Could it have been a lateral fricative [ɮˤ] or affricate [dɮˤ]?

Sībawayh rejected the pronunciation of ض <ḍ> and ظ <ẓ> as [θ] as "bad" in Al-Kitāb (793). This theta s reminiscent of Proto-Semitic *θˁ which is the source of Arabic /ðˁ/ ~ /zˁ/ written as ظ <ẓ>. I assume <ẓ> *ðˤ and (a continuant pronunciation?) of <ḍ> first merged into *ðˤ, devoiced to become [θˤ], and merged with [θ].

Retsö (2013: 435, 439-440) collected data pointing toward an earlier lateral pronunciation of the consonant written as <ḍ>:

- "Traces of such an articulation are found in some modern dialects in the southern peninsula"

- The name of the pre-Islamic Arabic god Ruḍā appears in 7th c. BC cuneiform as <ru.ul.da.a.ú> with <...l.d...>

- Arabic al-qaḍī was borrowed into Spanish as alcalde with -ld-

- "In the Modern South Arabian languages we find a laterialized and glottalized apico-alveolar consonant that etymologically corresponds to Arabic /ḍ/": e.g., Mehri ź /ɬ̠ʼ/ (I presume).

Yet he concluded "there is no real evidence that the present-day [stop] realization of the ḍād is secondary". I do not understand why.

Ehret (1995: 481) reconstructed the Proto-Semitic source of this consonant as a stop *dˁ as well as a lateral*ɮ ultimately going back to a Proto-Afroasiatic lateral *dl which was "probably" an affricate (i.e., [dɮ]?). I stated my objections to a stop interpretation here. Would Retsö regard Arabic /ḍ/ as a retention of Proto-Semitic source of this consonant as a stop *dˁ?

All that reminds me of the Tangut initial that Tai Chung-pui (2008: 201) reconstructed as ld-. Tibetan transcriptions for that initial include zl- and even a single instance of c-. I suspect it was a lateral affricate [dɮ] like Proto-Afroasiatic *dl. I have yet to look into the origin of ld-.

One ld-word with external cognates is

5710 1ldiq3 'arrow' (transcribed in Tibetan as ldi(H), zliH, d-ya; see Tai Chung-pui 2008: 198)

which Guillaume Jacques (2014: 161) derived from pre-Tangut *S-lje and cognate to Japhug zdi < *l- 'arrow'. I used to think 5710 was cognate to Old Chinese 矢 *hliʔ < *sl-? 'arrow'. Now I think it might be cognate to Tibetan mda < *mla 'arrow' (via Bodman's law) and Old Chinese 射 *mi-lak (*m-ljak?) 'to hit with an arrow' ~ *mi-lak-s (*m-ljak-s?) 'to shoot with a bow, archer', as *-a often rose to -i in Tangut.

Could 矢 and 射 be the zero and *a-grades of a root *lj-K? But there is no external support for a *-j- in the root. Unlike Guillaume, I don't think there was a medial *-j- in the Tangut word for 'arrow'. No other cognates contain -j-.

Was there a consonant between *S- (which conditioned Tangut vowel tension that I write as -q) and *-l- in pre-Tangut that fused with *-l- to become an affricate [dɮ]?

Even if I believed in reconstructing *-j- in 'arrow', I could not derive [dɮ] from *-lj- before *S- as Guillaume's *S-ljo 'head' became

0124 2luq3 (transcribed in Tibetan as lu; see Tai Chung-pui 2008: 220)

with l-, not ld-.

I also wouldn't reconstruct *-j- in 'head'; I think the pre-Tangut form was *S-luH which is close to Old Chinese 首 *hluʔ < *sl-? 'head'.

However, the Old Chinese initial of 'head' may be from *Kl- rather than *sl- 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.

15.10.28.21:48: WHAT IS THE ORIGIN OF THE VOICING CONTRAST IN ARABIC EMPHATICS?

Last night, I wrote (in haste as always),

In Afroasiatic (including Mehri), there is a three-way contrast between emphatics and voiceless and voiced nonemphatics

But as I've long known, that is certainly not accurate for Modern Standard Arabic which has a four-way contrast in its alveolar stops: /tˁ t dˁ d/. And some speakers have also a four-way contrast in their alveolar sibilants: /sˁ s zˁ z/. (Other speakers have /ðˁ/ instead of /zˁ/, but no one has a voiceless dental fricative /θˁ/.) However, there is no four-way contrast for nonalveolars.

According to Islam Youssef (2006: 13, 16), Cairene Arabic has voiced and voiceless emphatics as well as voiced and voiceless nonemphatics throughout its inventory of allophones, but has only five emphatic consonant phonemes which are all coronals: /dˤ tˤ sˤ zˤ rˤ/. The distinction between /r/ and /rˤ/ (Youssef 2006: 25) is absent from Modern Standard Arabic.

To be on the safe side, I've inserted "usually" in my statement since a voicing contrast in Afroasiatic emphatics is unusual in my extremely limited experience. If Arabic is not alone, I'd like to know.

In any case, Ehret (1995: 481-482) did not reconstruct a voicing contrast in Proto-Afroasiatic (PAA) emphatics. Here are his sources for the voiced and voiceless emphatic pairs in Arabic:

PAA *tʼ Proto-Semitic *tˁ > A /tˁ/

PAA *dl > Proto-Semitic *ɮ ~*dˁ > A /dˁ/

Omotic has ejective reflexes. I wonder if Ehret did not reconstruct this consonant as an emphatic at the PAA level because he wanted to avoid having two lateral emphatics. His PAA reconstruction has either zero or one emphatic per consonant class.

I think the Arabic stop might be an innovation since other Semitic languages in the comparative table at David Boxenhorn's blog have fricatives. I would rather not have Proto-Semitic *dˁ lenite independently multiple times while remaining intact in Arabic.

PAA *tlʼ > Proto-Semitic *sʼ ~ *sˁ > A /sˁ/

PAA *sʼ lost its emphasis and merged with *s in pre-Proto-Semitic and hence also in Arabic.

PAA *čʼ > Proto-Semitic *θˁ (or *tʲʼ?) > A /ðˁ/ ~ /zˁ/

Ehret cited Omotic languages which have čʼ today.

Did A /ðˁ/ become /zˁ/ to be the emphatic counterpart of /z/ so that both voiced emphatics were alveolars?

10.29.14:41: This shift also reduced the markedness of the segment since alveolar fricatives are more common than dental fricatives.

The last three sound changes have no parallels in Old Chinese unless *lˤ- became *ɮ- before hardening to *d-. I have yet to see any Afroasiatic-Sinitic parallels in emphatic evolution despite the existence of emphatic-conditioned Semitic-Sinitic vocalic parallels. Voiceless obstruents never became voiced in Old Chinese (though the opposite often occurred).

15.10.27.20:48: HOW IS MEHRI LIKE PROTO-INDO-EUROPEAN (AND UNLIKE OLD CHINESE)?

Voiced obstruents, or at least voiced stops, devoice in pausa. In this position, both the voiced and emphatic stops are ejective, losing the three-way contrast (/kʼ/ is ejective in all positions). Elsewhere, the emphatic and (optionally) the voiced stops are pharyngealized. Emphatic (but not voiced) fricatives have a similar pattern, and in non-pre-pausal position they are partially voiced.

Rubin (2010: 14) wrote (emphasis mine),

As Johnstone also notes, it is not completely clear how the glottalic [= Wikipedia's 'emphatic'] consonants fit into the categories of voiced and voiceless. Johnstone (AAL [Afroasiatic Linguistics], p. 7) wrote that they are "perhaps best defined as partially voiced." What is certain is that the glottalic consonants pair with voiced consonants when it comes to certain morphological features, for example the appearance of the definite article (§ 4.4) and the prefix of the D/L-Stem (§6.2).

Let's look at Rubin's coverage of those two features:

The definite article a- is found before the consonants b, d, ð, ð̣, g, ġ, j, ḳ, l, m, n, r, ṣ, ṣ̌, ṭ, w, y, z, and ź (voiced and glottalic consonants), though not all nouns beginning with those consonants take the article a-.

[...]

The definite article is also used with nouns beginning with ʾ, though only when the ʾ derives from etymological ʿ.

[...]

The definite article a- usually does not occur (or, one could say it has the shape Ø) before the consonants f, h, ḥ, k, s, ś, t, t, and x (voiceless, non-glottalic consonants). (p. 69)

The prefix a- appears only in [D/L-Stem verbs] when the initial root letter is voiced or glottalic, similar (but not identical) to the distribution of the definite article (see § 4.4). (pp. 93-94)

Aspiration of most of the voiceless non-glottalic [= nonemphatic] consonants constitutes an important element in the distinction of glottalic/non-glottalic pairs. (p. 14)

Although emphatics conditioned similar vowel changes in Mehri and Old Chinese (OC), the resemblance between the two stops there:

- In Afroasiatic (including Mehri), there is usually a three-way contrast between emphatics and voiceless and voiced nonemphatics, but in OC, there was a six-way contrast defined by voicing, aspiration, and emphasis: e.g.,

Mehri: /tˁ t d/ : cf. Proto-Indo-European (and Archi [Northeast Caucasian] and Ubykh [Northwest Caucasian]) */tʼ t d/

OC: *tˁ *tʰˁ *dˁ *t *tʰ *d

- There is no pairing between emphatics and voiced nonemphatics in Old Chinese.

Cf. how ejectives and voiced nonejectives did merge in many Indo-European varieties.

- Aspiration plays no role in distinguishing between emphatics and voiceless nonemphatics in Old Chinese.

But it may have distinguished between ejectives and nonejectives in Proto-Indo-European: e.g., */tʼ t d/ could have been *[tʼ tʰ dʱ].

These differences between Mehri and OC do not necessarily invalidate what I could call the 'Chinese Glottalic Theory'. But they do imply that there are limits of using Mehri (or any similar language) for predicting phonetic phenomena in OC.

I wish that OC did not have unique typology. Maybe it didn't. I would not be surprised if emphasis played a role in the development of Tangut vowels.

15.10.26.23:59: HOW IS MEHRI LIKE OLD CHINESE?

I have known for some time about the existence of the Modern South Arabian languages (not to be confused with Old South Arabian). However, I had never read anything about them. I couldn't even name one until tonight when I discovered Rubin's The Mehri Language of Oman (2010). I immediately went to the section on vowels and found synchronic shifts reminscent of diachronic shifts in Old Chinese:

I have converted Rubin's notation into IPA as used in the Wikipedia article on Mehri.

Lowering in Mehri is conditioned by three classes of consonants:

Glottalic: /tʼ θʼ ɬ̠ʼ sʼ kʼ/ (no known cases of /ʃʼ/ followed by /iː uː eː/)

Guttural: /χ ʁ ħ ʕ ʔ h/

Liquid: /r l/ if "there is normally a glottalic or guttural consonant elsewhere in the root" (Rubin 2010: 29; i.e., glottalic and guttural consonants' lowering effects can spread 'through' them: e.g., [məʁrajb] 'well-known' which I presume is phonemically /məʁriːb/)

10.27.11:52: This phenomenon is reminscent of the 'transparency' of Thai sonorants (not just liquids) in tonal development: e.g., อร่อยʔàrɔ̀ɔy 'delicious' should in theory have an r-conditioned falling tone on its second syllable, but it has a glottal stop-conditioned low tone as if the -r- didn't exist. However, in Thai, the conditioning consonant has to precede the 'transparent' sonorant, whereas in Mehri, the consonants conditioning lowering can be in the order liquid ... glottalic/guttural as well as the reverse: e.g., [məlawtəʁ] 'killed' (masc. pl.) which I presume is phonemically /məluːtəʁ/.

Wikipedia calls the glottalic consonants 'emphatic' and lists pharyngealized allophones for all but /kʼ/. Thus they may be comparable to the Old Chinese pharygealized 'emphatic' consonants that conditioned the lowering of high vowels. However, Mehri has a limited set of mostly coronal emphatics, whereas Old Chinese had an emphatic counterpart of every single nonemphatic consonant and even *ʔʷˁ which had no nonemphatic counterpart (Baxter and Sagart 2014: 69).

Early Old Chinese as reconstructed by Baxter and Sagart did not have any uvular or pharyngeal fricatives. I have hypothesized that Old Chinese had pharyngeal fricatives for Old Chinese, but there is no strong evidence for them. I do, however, believe Late Old Chinese developed uvular fricatives accompanied by lowered vowels.

The Old Chinese counterparts of Mehri glottals and liquids did not condition lowering. I would not have expected Mehri /ʔ h/ to condition lowering since I have never heard of nonpharygealized glottals having that effect in any other language.

15.10.25.23:59: HOW DID CAMISIA BECOME KAMĪZ?

I was surprised that Latin camisia 'shirt' was borrowed into Arabic as قميص qamīṣ with three features I woudn't expect:

1. uvular q for [k]

2. a long ī for short [i] (to match an existing a ... ī vowel template? lengthening in some intermediary language?)

3. an emphatic ṣ for nonemphatic [s]

Why not borrow the word as *kamis? And why does Urdu has -z in qamīz? I assume kamīz and kamīj show different degrees of Indicization (i.e., avoidance of un-Indic q and z).

10.26.0:26: While I'm on this topic, a kamīz is half of a shalvār kamīz outfit. Why was Persian شلوار‎‎ shalvār borrowed into Arabic as سروال‎ sirwāl instead of شلوار‎‎ *shalwār? Was i ... ā an existing vowel template? Why not preserve the consonants instead of backing s and reversing l and r?

10.26.3:26: Is kamīz a direct borrowing from Portuguese camisa [kɐmizɐ]? Is qamīz a compromise between qamīṣ and kamīz?