15.1.24.23:50: THE GOLDEN GUIDE: LINE 104: TANGRAPHS 516-520

104. I have a long list of topics, and I can't make up my mind about what to write about next, so I'll fall back on the Golden Guide. Two lines to the last surname ...

Tangraph number	516	517	518	519	520
Tangraph
Li Fanwen number	1720	1456	4686	0881	4760
My transcription	1ven1	1chhi2	1khwan4	2se4	1an1
Tangraph gloss	the surname Ven	the surname Chhi; Sanskrit chi, che?	the surname Khwan; transcription of Chinese 郡 *3khwin3 'administrative region'	the surname Se	the surname An
Word	the surname 隗 Wei (*2wi3) or 韋 Wei (*1wi3)	the surname 翟 Zhai (*4chhe2)	the surname 權 Quan (*1khwan2)	the surname 薛 Xue (*4se4)	the surname 安 An (*1an1)
Translation	Wei, Zhai, Quan, Xue, An

Now I have Kotaka's six-part series on the Golden Guide on hand, so I'll use that as reference in addition to Nie Hongyin and Shi Jinbo's article. Kotaka's notes make me realize how poorly understood the relationship between the phonologies of Tangut, Tangut period northwestern Chinese, and Sanskrit still is.

516: I suppose this analysis of 1720 somehow describes the Ven family:

=+

1720 1ven1 = right of 1105 1khon4 'to give' + left of 5659 1ver1 'luxuriant'

Grade I 1720 appears in the Tangut translation of Sunzi as a transcription of the Chinese Grade III surname 隗 *2wi3 which had no nasality. Why wasn't 隗 transcribed as Grade III *vi3?

Nie and Shi identified 1720 as the Chinese Grade III surname 韋 *1wi3 which also had no nasality. Kotaka noted that 韋 was transcribed as

5287 1vi1 (with Grade I, not III!)

in The Forest of Categories, so he thought 1720 was unlikely to be 韋. The fact that *wi3 (with different tones) was transcribed as 1ven1 and 1vi1 may indicate that Chinese *-i3 was unlike either Tangut -en1 or -i1 and had no exact match in Tangut.

517: 1456 is a fanqie character:

=+

1456 1chhi2 = 1796 1chhuq3 'to lure' + 4972 1chi2 'to amuse'

The Tangraphic Sea states that 1456 is for transcribing mantras. Arakawa (1997: 116) thought 1456 might represent Sanskrit chi or che, but Sanskrit ch- was normaly transcribed as Tangut tsh-, not chh-. Moreover, the rhyme -i2 is not in Arakawa's table of attested Sanskrit transcriptions, implying that -i2 was somehow unlike Sanskrit short i, long ī, or long e [eː]. (Sanskrit has no short e.)

The use of 1456 1chhi2 for Chinese *4chhe2 may imply that neither Tangut -i2 nor Tangut -e2 precisely matched Chinese *-e2.

518: 4686 is a semantic compound:

=+

4686 'administrative region' = 4719 2keq2 'boundary' + 2725 1wo2 'circle'

4686 1khwan4 is a poor vocalic match for Chinese 郡 *3khwin3 'administrative region'.

4686 appears in Sunzi as a transcription character for 權 *1khwan2 which can be a surname. The vowel types match but the grades (Tangut IV, Chinese II) don't.

519: The analysis of 0881 is unknown.

The left side may not be phonetic since I cannot find any se4-graphs containing it. Perhaps it describes the Se family: e.g., if it is short for

4773 2luq3 'silk',

the Se might have be known as sellers of silk (and Se brings to mind Latin sericum 'silk', though the similarity is probably coincidental).

The right side must be from 2888 2my1 'surname'.

Nie and Shi identified 0881 as a transcripiton of Chinese 薛 *4se4, but Kotaka pointed out that 薛 was transcribed as

3683 2sa4 (first syllable of 3683 2532 2sa4 1de4 'day after tomorrow' 1de4 is 'day')

in The Forest of Categories.

520: 4760 is a fanqie character:

=+

4760 1an1  = top of 4940 2y4 '' + bottom left and center of 4685 1an1

4940 either represented initial glottal stop or zero. Homophones defined 4760 as a surname and its homophone 4685 as a place name, but 4760 also appeared in the place name

4760 3628 1869 1an1 1ghwan4 1po1  ' 安原堡 *1an1 1(ngg)wan3 1po1 Anyuan Fortress'

The correspondence of Tangut gh- to Chinese *ngg- or zero should be investigated.

15.1.23.23:46: MEETING TANGUT CROWS

Do non-Chinese Sino-Tibetan languages have different sets of consonants corresponding to the (labio)velars and (labio)uvulars that Baxter and Sagart (2014) reconstructed? I would like to see the comparative evidence in Pan Wuyun's "喉音考" (On laryngeals, 1997) mentioned in Baxter and Sagart's 2010 paper on uvulars. Baxter and Sagart focused on Chinese-internal evidence, though they did point out that they thought

Written Tibetan g- corresponded to Old Chinese *ɢʷ- (which I used to reconstruct as *w-)

Written Burmese ဟောင်း <hoŋḥ> 'old' might be cognate to 公 'father, ruler' (< 'elder'?) which they now reconstruct as *C.qˁoŋ (*Cə.qˁoŋ in 2010)

cf. the velar-velar correspondence of

Written Burmese ကိုး  <kuiḥ> : Old Chinese *kuʔ 'nine'

It would be especially nice if comparative evidence supported the pharyngealized/nonpharygealized distinction in Baxter and Sagart's velars and uvulars.

I proposed that uvulars may be one source of Tangut Grade II based on these comparisons:

2750 1ghu2 < *ɢu 'head' : Old Chinese 后 *ɢˤ(r)oʔ 'sovereign', Written Tibetan mgo 'head'

4046 1khi2 < *CI-qha 'bitter' : Zhongu qʰɐⁿde 'bitter', Ronghong Qiang qʰɑ(q)

but Old Chinese *kʰˤaʔ has a velar!

Could Tangut clarify whether Old Chinese 烏 *qˤa  ~ 鴉 *qˤra 'crow' and 迓 *ŋˤ<r>ak-s (*m-[qʰ](r)ak-s?) 'to meet'* had uvulars?

Nishida (1964: 203) and Grinstead (1972: 114) glossed

1550 3110 2ka1 0jiq3  < *kra or *qa  + S-ji(-H) or *SI-ja(-H)

as 'crow', but the Chinese gloss in the Timely Pearl is 老鴟, literally 'old bird of prey' ('old' is a common noun prefix; cf. English old expressing familiarity rather than age). In any case, 2ka1 is Grade I rather than Grade II *2ka2 which I would expect for a cognate of Old Chinese *qˤ(r)a.

2ka1 can appear by itself, but 0jiq3 is a bound morpheme. (0 indicates an unknown tone.)

2261 0176 2on4 1na'3

2261 is also the second syllable of 1ta1 2on4 'swallow'; apparently an on was a kind of bird, and a crow was a 'black on'.

2262 2114 1jwon3 1leq2

The graph for 2114 is derived from 2262 'bird' plus 0176 'black'.

As for 迓 *ŋˤ<r>ak-s (*m-[qʰ](r)ak-s?) 'to meet', the only vague match I could find was Grade II

4040 1khu'2 'to invite'

which could be from *khru-X, *qhu-X, or *qhru-X. *qhru is the most likely since its cluster matches that of Japhug qru 'to invite'*.

''.*X (the pre-Tangut source of the mysterious phonemic attribute that I write with an apostrophe as a convenient subsitutte for a prime symbol) must be an affix***, as cognates identified by Guillaume Jacques (2014: 59) lack it:

2791 2khu4 < *Cɯ-qhru-H 'to call, invite'

3254 2khu4 < *Cɯ-qhru-H 'imperial edict'

(The presyllable *Cɯ- conditioned Grade IV and the *-H conditioned the second tone.)

Guillaume compared 4040 to Written Burmese ကြို  <krui> 'to meet someone on arrival' which has <k-> corresponding to Japhug q-. If Japhug q- corresponded to Old Chinese *q-, does that mean *q- remained as a stop in Written Burmese clusters?

*q- > h- in ဟောင်း <hoŋḥ> 'old'

*qr- > kr- in ကြို  <krui> 'to meet someone on arrival'

*1.24.4:48: The phonetic series of 迓 is largely uvular, and 迓 may be related to 御 *m-[qʰ](r)aʔ 'to ward off', so perhaps 迓 could be reconstructed with a uvular root initial.

**1.24.4:57: Although Guillaume Jacques (2014: 58) proposed Japhug qru as a cognate, he reconstructed the pre-Tangut form of 4040 as *khjoo without a uvular or *-r-. His *-j- is a carryover from Gong Hwang-cherng's Tangut reconstruction.

***1.24.4:57: I conventionally write *-X as a suffix, but it could have been a prefix or infix.

15.1.22.22:40: A 'DENTAL' UVULAR SERIES

Perhaps 鴉 *q(r)a once had an *m-prefix for animals that justified the choice of 牙 *m-ɢˤ<r>a 'tooth' as a phonetic but vanished without a trace.

Baxter and Sagart (2014) reconstructed most other members of Grammata Serica Recensa series 0037 牙 'tooth' with the structure *N-Qra:

I added 0047 since it too contains 牙 'tooth' as a phonetic. (Karlgren did not recognize that. Hence he placed it in a separate series. It seems that the trend is to combine his series. I can't think of any series of his that has been split by later scholars.)

Notes on individual members:

0037a 牙 *m-ɢˤ<r>a: I long assumed that 'tooth' had *ŋ- and might have been a loan from a Southeast Asian ŋa-word for 'ivory'. But if Baxter and Sagart are correct, then the word could have spread throughout Southeast Asia after *m-ɢˁ- fused into a nasal. Or *m-ɢˁ- was borrowed as a nasal. In either case, the only non-Chinese support I know of for a medial liquid is Bahnar ŋə-la 'ivory' (from Schuessler 2007: 550). I could not find that word in SEAlang's Bahnaric data which contains other words that appear to be reflexes of Shorto's (2006) Proto-Mon-Khmer *[m]laʔ 'ivory' and *bluək 'id.' I presume Shorto considered ŋa-words to be borrowings. Perhaps Chinese traders lacking their own word for tusks called them 牙 'teeth' and Southeast Asians borrowed that word for 'tusk'. It is unlikely the semantic shift went in the other direction: i.e., Southeast Asians sold ŋa 'tusks' to the Chinese who then adopted that word for 'teeth'.

0037b is a variant of 0037a.

0037c 庌 *[N]-qʰˤraʔ 'covered galleries' belongs to the same word family as ⾑ *qʰˤraʔ (this site: *qʰraʔ) 'cover'.

0037d 芽: Is this the word for 'teeth' applied to plants? Are sprouts like teeth growing from the earth?

0037e 訝: Not listed. Synonymous with 0037f 'to meet', so I supposed it was also *ŋˤ<r>ak-s.

Can also mean 'astonished'. Could 'astonished' be *[N]-qʰˤrak-s which would be in the same word family as 虩 *qʰrak 'to fear'?

0037f 迓 *ŋˤ<r>ak-s: I would rather not reconstruct a velar word in a uvular series, but I think Baxter and Sagart did so because it belongs to a velar word family with

0699d 迎 *ŋ<r>aŋ(-s) 'to meet'

0766n' 輅 (no reconstruction given; *ŋˤ<r>ak-s since 0766 is a mostly velar series) 'to meet'

0788a 屰 *ŋrak 'to go against'

(23:36: Also cf. Written Burmese ငြား <ŋrāḥ> 'to meet'.)

Then again, Schuessler (2009: 551) thought 0037f was cognate to

0060l 御 *m-[qʰ](r)aʔ 'to ward off'

0037g 雅 *[N-ɢ]ˤraʔ ~ *N-ɢˤraʔ: I guess Baxter and Sagart are less sure about how to reconstruct 'kind of musical instrument' than 'proper, refined' because they regard the latter as cognate to 夏 *N-ɢˤraʔ 'great' whereas the etymology of the former is unknown (at least to me), so there is no word-family evidence to favor a specific preinitial or root initial.

0047a 邪 *sə.ɢA ~ *sə.la ~ *ɢ(r)A ~ *[ɢ](r)A: Until now I would have reconstructed something like *sɯ-ŋlja to accomodate its various readings and 牙 which I would have reconstructed as *ŋra or *rŋa.

(My *ja is equivalent to Baxter and Sagart's cover symbol *A for an *a which has an unusual Middle Chinese reflex. See pages 223-224 of their book. I once considered reconstructing a seventh vowel, but as they pointed out, there is no rhyme evidence for one.)

I think they feel compelled to reconstruct 邪 *sə.la 'to walk slowly' with a liquid because it is an alternate spelling of

0062p 徐 *sə.la 'to walk slowly'

in the Classic of Poetry - or at least the text as we have it now. How old is the use of 邪 for 'walk slowly'? Could it postdate the merger of *ɢ- (my *sɯ-ɢ-) and *l- as *j-? There is no doubt that 0062 is a lateral series.

15.1.21.23:54: WHY DO SOME CHINESE CROWS HAVE 'TEETH'?

於 *Cɯ-qa (Baxter and Sagart: *[ʔ]a) > Middle Chinese *ʔɨə 'in'

烏 *qa (Baxter and Sagart: *qˤa) > Middle Chinese *ʔo 'crow'

which were written with variations of a drawing of a crow and are regarded as members of the same phonetic series (Grammata Serica Recensa 0061). Ideally I'd want them to have the same initial consonant, though Baxter and Sagart reconstruct them with different consonants while leaving the option of *q- open for 於. (Brackets indicate uncertaintly; in this case, *[ʔ] means 'either *ʔ or something else that has the same Middle Chinese reflex as *ʔ: i.e., *q.)

The Middle Chinese readings of the 烏/於 phonetic series only has initial *ʔ-, so without additional evidence, there is no way to tell whether that Middle Chinese *ʔ- was from Old Chinese *ʔ- or *q-.

Transcriptions such as 烏弋山離 'Alexandria' and 烏桓 'Avar' from the Records of the Grand Historian (c. 100 BC) tell us that 烏 was something like *a toward the end of the first millennium AD, though the fine details are uncertain: e.g., was the initial consonant in the underlying dialect(s) zero, a glottal stop (with or without pharyngealization?), or even a pharyngeal fricative *ʕ- (cf. how Arabic ʕ- is borrowed as phonemic zero in English)? They do not rule out the possibility of another initial consonant at an earlier period: e.g., *q-.

烏 *qa (Baxter and Sagart: *qˤa) > Middle Chinese *ʔo 'crow'

鴉 *qra (Baxter and Sagart: *qˤra) > Middle Chinese *ʔæ 'crow'

as members of the same *qa-word family.

By establishing that 牙 was a uvular series, Baxter and Sagart solved the mystery of why Middle Chinese 牙 *ŋæ is phonetic in 鴉 whose reading probably never had a nasal.

If Schuessler (2007: 83, 517) is correct, 鴉 never had *-r- (which is unlikely to be an infix in 'crow'*) and its Middle Chinese low vowel is an archaism. Hence 烏 and 鴉 could have been homophones, and Mandarin 烏鴉 wuya 'crow' would be a reduplication if pronounced in Old Chinese as *qa qa.

Moreover, *qa matches the global sound-symbolic archetype for 'crow': e.g., Sanskrit kāka-. (See Wiktionary for more examples. Thai kaa 'crow' could be a borrowing from Chinese, though it could also be the independent product of sound symbolism.)

*1.22.4:52: Old Chinese *-r- indicates double or multiple objects (Baxter and Sagart 2014: 58) and would be out of place in 'crow' unless 鴉 *qra once meant something like 'a flock of crows'.

If 鴉 'crow' had *-r-, then 烏 *qa and 鴉 *qra were a word family only in the weak sense that they were based on the same sound symbolism (cf. the English 'word family' of gl-words: gleam, etc.). I would not consider *qra to be a derivative of a root √*qa.

Perhaps 鴉 *q(r)a once had an *m-prefix for animals that justified the choice of 牙 *m-ɢˤ<r>a 'tooth' as a phonetic but vanished without a trace.

15.1.20.23:59: GIVING IN TO THE UVULAR HYPOTHESIS

For over a decade I had a vague notion that Old Chinese had distinct velar and uvular phonetic series, but I never worked out any criteria to sort out which was which (beyond assuming that uvular series normally retained or conditioned low[ered] vowels - an assumption I retain today). Baxter and Sagart (2014) present such criteria, building upon Pan Wuyun's 1997 proposal: e.g., phonetic series mixing velar and laryngeal initials in Middle Chinese (MC) pronunciation such as 1173/1189/1190 were once uvular in Old Chinese (OC).

1173a 公 MC *koŋ < OC *C.qˁoŋ (this site: *C.qoŋ) 'father, prince; impartial'

cf. Karlgren's *kuŋ, Schuessler's *klôŋ 'prince', *kôŋ 'impartial'

Schuessler compared *klôŋ 'prince' to Khmer ខ្លោង <khloŋ> 'chief' and *kôŋ to Written Tibetan dgung 'middle'

1173g 瓮 MC *ʔoŋʰ < OC *qˁoŋ-s (this site: *qoŋs; earlier  'earthen jar'

cf. Karlgren's *ʔuŋ, Schuessler's *ʔôŋ

1189a 妐 MC *tɕuoŋ < OC *t-qoŋ (this site: *tɯ.qoŋ) 'father-in-law'

cf. Karlgren's *tjuŋ, Schuessler's *toŋ

1190a 松 MC *zuoŋ < OC *sə.ɢoŋ (this site: *sɯ.ɢoŋ) 'pine'

cf. Karlgren's *dzjuŋ, Schuessler's *s-loŋ

Once I would have more or less agreed with Schuessler and reconstructed *-l- in all three series:

*kloŋ, ʔloŋs, tɯ-loŋ, sɯ-loŋ

But as Baxter and Sagart (2009: 233) noted, medial *-l- is not supported by comparative evidence (aside from Schuessler's Khmer comparison. For instance, Written Burmese ဟောင်း <hoŋḥ> 'old', a possible cognate of 公 '', lacks <l>. I have already mentioned -l-less Written Tibetan dgung 'middle', a possible cognate of 公 'impartial'. Furthermore, the 公 series does not have most of the Middle Chinese initials typical of uncontroversial lateral series:

MC *d- < OC *lˁ-

MC *ɕ- < OC *l̥ -

MC *tʰ- < OC l̥ˁ-

Baxter and Sagart (2014: 171) solved an old mystery for me: the etymology of Taiwanese 予 hōo [hɔ˧ ] 'to give' whose proto-Min initial is *ɣ-. Until Sunday morning I thought hōo could have been a substratum word of unknown origin. But Baxter and Sagart regarded it as a descendant of 與 'to give' which they reconstructed as *m-q(r)aʔ. *m-q- fused into *ɢ- which weakened to *ɣ-. My reconstruction of 與 as *Cɯ-laʔ (whose *C may have been *k-*) could not be related to the Taiwanese word. Moreover, the phonetic series of 與, like that of 公, also lacked the three Middle Chinese initials that I listed earlier. Baxter and Sagart's approach accounts for more facts. So at last I give in and accept their reconstruction as a point of departure.

1.21.4:05: I forgot to mention the reason I italicized in in the title of the article. My attempt to integrate the Baxter-Sagart hypothesis with my own ideas about the origins of 'emphasis' in Chinese requires me to reconstruct high-vowel presyllables in type B ('nonemphatic') syllables with uvulars: e.g.,

與 OC *mɯ-q(r)aʔ (cf. my old *Cɯ-laʔ) > MC *jɨə

One problem with my hypothesis is that I am required to reconstruct presyllables in type B function words which were more likely to be monosyllabic: e.g..,

於 OC *Cɯ-qa (B&S: *[ʔ]a) > MC *ʔɨə 'in'

於 is a variant of a drawing of OC 烏 *qa (B&S: *qˤa) 'crow', so I assume the two words were nearly homophonous.

于 OC *Cɯ-q(r)aʔ (B&S: **ɢʷ(r)a) > MC *wuo 'in'

Compare the synonymous type A syllable 乎 *ɢa (B&S: *ɢˤa) 'in' which became Middle Chinese *ɣo without a high vowel characteristic of type B syllables in Middle Chinese.

Baxter and Sagart's reconstructions of the first two words for 'in' are simpler, but their system has many more uvulars than mine.

1.21.4:12: After finishing the body of my post, I realized that 與 'to give' and 予 'to give' (whose character represents Taiwanese hōo) were both *Cɯ-laʔ in my old reconstruction, whereas they now have different consonants in Baxter and Sagart's reconstruction (and in my revision of my reconstruction):

與 B&S: *m-q(r)aʔ, this site: *mɯ-q(r)aʔ

予 B&S: *laʔ, this site: *Cɯ-laʔ

Is it a coincidence that these two verbs rhyme?

*In my old reconstruction, 與 *Cɯ-laʔ 'to give' may have been homophonous with 舉 *kɯ-laʔ 'to raise' whose *k- survives today in Cantonese geoi [kɵy˨˥]. Could 'to raise' have become 'to give' as in the case of Japanese ageru 'to raise, give'?

15.1.19.15:29: THE (K)HÍ-STORY OF HƠI

For many years I have been puzzled by hơi [həːj] 'gas, air, breath, odor', an early Chinese loan in Vietnamese corresponding to Sino-Vietnamese 气/氣* khí [xi] which was borrowed later. Although this word now has h- in Cantonese, that fricative is from *kʰ- which independently weakened to [x] in Vietnamese sometime between the 17th century (when it was [kʰ] and first romanized as kh-) and modern times. Schuessler (2009: 305) reconstructed *kʰ- for Late and Early Old Chinese. How can Vietnamese hơi be older than khí (as indicated by its rhyme and tone**) yet have a seemingly newer initial?

Yesterday morning the solution came to me as I read Baxter and Sagart (2014)'s section on 气. On page 170, Baxter and Sagart reconstructed 'breath' as

*C.qʰəp-s > *C.qʰət-s > *kʰət-s

which then became Late Old Chinese *kʰɨəs, Early Middle Chinese *kʰɨ(ə)jʰ, and Late Middle Chinese *kʰɨi. That last form was borrowed as khi.́

The uvular root is preserved (possibly with an infix) as

吸 *qʰ(r)əp > Middle Chinese *xip 'to inhale'

A "tightly attached preinitial *C-" (p. 169) conditioned the fronting of uvular *qʰ- to velar *kʰ-. If that preinitial (probably a prefix in the case of 气***) were not present, *qʰ- would have weakened to *x-. And that is what I think happened to the source of hơi in the Chinese dialect spoken in Vietnam under Chinese rule:

*qʰəp-s > *qʰət-s > *χəs > *xəjʰ

The h- of hơi is from that *x- and has nothing to do with Cantonese h- from *kʰ-.

I hypothesize that Old Chinese *ə partly bent upward to *ɨə after velars but remained unbent after uvulars****. (It may have lowered and backed to *ʌ.) The vowel stayed unbent even after uvular *χ- fronted to velar *x-. Hơi preserves that unbent vowel.

*The complex variant 氣 is well known in the West as Mandarin qi / ch'i and Japanese/Korean ki.

**The rhyme -ơi [əːj] is closer to Early Middle Chinese *-ɨ(ə)jʰ than to Late Middle Chinese *-ɨi which is reflected in khí.

The ngang tone of hơi [həːj] in old loans corresponds to earlier Chinese *-ʰ < *-s and the sắc tone in newer loans.

***The period in *C.qʰəp-s indicates that Baxter and Sagart (2014: 7) were "not confident that a presyllable [or preinitial] is a synchronic prefix". It does not rule out the possiiblity that the presyllable or preinitial was a prefix prior to Old Chinese.

In this case, I think *C- was some sort of nominalizing prefix added to *qʰ(r)əp 'to inhale'.

I do not know why Baxter and Sagart reconstructed *(r) indicating the possibility of *-r- in 'inhale' but not 'breath'. The Middle Chinese reflexes of *-rəps and *-əts are identical. Perhaps semantics make *-r- unlikely in 'breath', as 'breath' does not belong to Baxter and Sagart's (2014:57-58) three categories of *-r-infixed words:

1. actions with multiple agents, patients, or locations or repeated actions

2. intensified stative verbs

3. double or multiple objects

I am certain that there was no *-r- in the source of hơi, because *qʰrəp-s would have become *xɛjʰ (= xeajH in Baxter and Sagart's Middle Chinese notation) and would have been borrowed into Vietnamese as *he [hɛ].

****Baxter and Sagart (2014) allow what I call upward bending after bare nonpharygealized uvulars: e.g., 氣 originally represented a verb 'to present food' (hence its semantic component 米 'rice') which they reconstructed as

*qʰət-s > my Late Old Chinese *xəs > my early Middle Chinese *xɨ(ə)jʰ (= xj+jH in Baxter and Sagart's Middle Chinese notation)

I don't distinguish between pharyngealized and nonpharygealized uvulars. In my reconstruction, only downward bending of higher vowels could occur after uvulars unless they were preceded by a high-vowel presyllable or converted to velars at an early stage. Hence I reconstruct 'to present food' as *Cɯ.qʰət-s.

The *C- of 'to present food' might have been identical to the *C- of 'breath', facilitating the use of the character for the former to write the latter.

'Breath' might have been *Cɯ.qʰəp-s at an earlier stage, though the unstressed presyllabic vowel must have dropped by the time uvulars weakened because *Cɯ.qʰ- became a fricative *χ-, whereas *C.qʰ- became a velar stop *kʰ-.

15.1.18.23:45: FOUR SERIES OR EIGHT? OLD CHINESE (LABIO)VELARS AND (LABIO)UVULARS

As I read Baxter and Sagart (2014), I am becoming increasingly convinced that they are right in many ways, yet I remain skeptical about their eight series of back consonants:

I reconstruct only four series:

*K : *Kʷ : *Q : *Qʷ

Here is how Baxter and Sagart's series map onto mine:

All researchers agree that Old Chinese had two types of syllables: A and B. I consider A syllables to be 'emphatic' and B syllables to be 'nonemphatic'. Baxter and Sagart project the A/B distinction far back, whereas I think it only became phonemic after the loss of presyllables that conditioned the distinction.

Syllables with nonuvular initials and lower vowels (*A = *e, *a, *o) and syllables with uvular initials (regardless of vowel) are type A unless preceded by a high-vowel presyllable:

*Cɯ-CA > *Cɯ-CIA > *CIA

*IA represents warped lower vowels (*ie, *ɨa, *uo).

Syllables with nonuvular initials and higher vowels (*I = *i, *ə, *u) are type B unless preceded by a low-vowel presyllable:

*Cʌ-CI > *Cʌ-CAI > *CAI

*AI represents warped higher vowels (*ei, *əɨ, *ou).

*X represents weakened (former) uvulars:

*q- > *ʔ- (followed by the shift *K- > *Q- before lower vowels and warped higher vowels)

*qʰ- > *x- before higher vowels and warped lower vowels; χ- before lower vowels and warped higher vowels

*ɢ- > *ɣ- before higher vowels and warped lower vowels; *ʁ- before lower vowels and warped higher vowels

*ɣ- later weakens to *j-, merging with *j- from stage 1 *l-, etc.

In stage 1, uvulars are phonemic.

In stage 2, uvulars are allophones of velars and need not be specified: e.g., *[qɑ χɑ ʁɑ] /ka xa ɣa/.