07.11.10.21:21: EIGHT REASONS *Y (PART 5)

In part 4, I mentioned the possibility that Old Chinese had a series of palatalized or palatal consonants: e.g., *ch, *sh. These consonants would be reconstructed in some instances where Middle Chinese has a front vowel instead of an expected central or back vowel: e.g.,

者 OC *ca' > MC *chyæ' [tɕjæ^ʔ]

石 OC *jak > MC *jiek [dʑiek]

However, graphs for palatal-initial syllables appear as phonetics in graphs for nonpalatal syllables:

都 OC *ta > MC *to

拓 OC *thak > MC *thak

which also serves as another spelling of 摭 OC *cak > MC *chyek [tɕiek]

I proposed reconstructing *Cy-clusters instead:

者 OC *tya', phonetic in 都 OC *ta

石 OC *dyak, phonetic in 拓 OC *thak

(also alt. spelling of 摭 OC *tyak)

Do such clusters correspond to Benedict and Matisoff's Proto-Tibeto-Burman *Cy-clusters (or palatal consonants)? Below is a list of all PTB *coronal-y forms with proposed Chinese cognates in Matisoff 2003:

PTB *g-t(y)ik, *tyak 'one' : 隻 OC *tek

PTB *tya-n 'red' : 丹 OC *tan 'red'; probably not cognate to 朱 OC *to 'red', 綪 OC*C-sings 'dark red', 縉 OC *tsins 'pale red' (unless this is from *s-tin-s), 紫 OC *tse' 'purple' (unless this is from *s-te'; its true PTB cognate may be *t[s]aay 'red')

PTB *m-d(y)ak 'good' (see PTB *lyak below)

PTB *dyal 'lip': 唇 OC ?*mdur 'lip' (Matisoff [2003: 539]) no longer believes they are cognate; he prefers to compare 唇 to PTB *m-ts(y)ul 'lip/beak')

PTB *dyung 'insect': not cognate to 蟲 OC *rlung; a loan from late OC *Dung [ɖuŋ] or *Düng [ɖyŋ])?

PTB *nya 'woman' : 女 OC *rna' 'woman' (added 11.10.22:28)

PTB *nya-n 'red' : 赧 OC *rnan' 'blush'

PTB *s-nyam 'think' : 念 OC *nyəm(')s 'think', 恁 OC nəm' 'think' (added 11.10.22:35)

PTB *s-nyam/*s-nem 'soft' : 荏 OC *nəm' 'id.' (added 11.10.22:36)

PTB *nyap/*nip/*nup 'sink/enter' : 入 OC *nəp 'enter' (added 11.10.22:31)

PTB *nyap/*nip 'crush/compress/press on' : 躡 OC *rn(y)ap 'trample' (added 11.10.22:31)

PTB *tsyəw 'hand' : not cognate to 手 OC *hnu' 'hand'

PTB *tsyow 'boil/cook' : not cognate to 煮 OC *ta' 'id.'

(similar to but still not cognate to 炒 Md chao < OC *C-saw' 'fry'; is the PTB form borrowed from a late OC *Chhæw' [tʂ^hæwʔ]?)

PTB *tsyak 'red, blood; gold' : not cognate to 赤 OC *k/t-hlyak

unless PTB *tsy- ([tsj] or [tɕ]?) is from an earlier *khly- or *thly-)

PTB *tsyat 'break/cut' : 絕 OC *dzot (< ?*N-s-tot?) 'cut off/break off', 脃/脆 OC *tshots 'brittle' (< ?*C-s-tot-s) 'brittle', 折 OC *tet (< ?*tyat) 'break/bend/destroy'

is PTB *tsy- < *s-ty-, and was the PST root *t(y)Vt?

PTB *tsyap 'join/connect' : 接 OC *ts(y)ap 'contact'

PTB *dzya-n/k 'eat/food' : 餐 OC*tshan 'eat/food/meal' (phonetic [not in my font] is *dzan); not cognate to 飤 OC *s-lək-s 'food/give food to', 食 OC *m-lək 'eat'

PTB *gyi/dzyi 'ride horse' : the former looks like a borrowing from some later stage of Chinese: 騎 late MC *gïy < early MC *gïe < late OC *gïay <  OC *gay, and the latter (source of Tangut dzeey 'ride') might be an unrelated native TB word; its resemblance to Md qi [tɕ^hi] < Late MC *gïy is purely coincidental, but its resemblance to Old Khmer jih 'ride' may not be - is Written Burmese cih a borrowing from a Mon cognate to OK jih?

PTB *dzyuk 'belong' : borrowing from 屬 MC *juok [dʑuok] <*N-tuk 'be attached to; belong'

PTB *sya 'flesh/meat' : not cognate to 身 OC *hling 'body' (for *-ng cf. 銵 Md keng < MC *khEng < OC *khrling < *kr-hling 'sound of clinking metal' [onomatopoetic; if not for *r-, not too different from Eng clink!])

also not cognate to 獸 MC *shuh < OC *Cus 'beast' (initial unknown, but the OC initial could not be *s[y]-, and the vowels do not match)

PTB *s(y)ar 'louse' : not cognate to 蝨 OC *srit < *srik 'louse' unless this is from a pre-OC *syarik, which is almost certainly not the case, as nothing indicates that PTB preserves the first halves of disyllables; see below for the true PTB cognate of 蝨

PTB *sywar/*sywa-n/t : 散 OC *sans and 撒 MC *sat < OC ?*sat 'disperse' (not attested in OC; earlier spelling was 殺+米 with phonetic 殺 OC *r-syat)

PTB *m-syil 'wash' : 洗 OC *syər/l 'wash' (added 11.10.22:19)

PTB *syim 'sweep' : 侵 OC *tshim < ?*C-sim 'sweep over/invade' (its right side is a drawing of a broom and a hand)

PTB *syip/syup 'whisper' : 咠 OC *ts(h)ip 'whisper' (sound-symbolic coincidence?; cf. Eng whisper with -isp-)

PTB *syu(w) 'grandchild' : 孫 OC *sun 'id.'

PTB *syey(-s) 'know' : 悉 OC *sit 'all'; 'know (thoroughly)'

PTB *syow 'rat' : not cognate to 鼠 OC *hna' 'id.'

PTB *zyaaw/zyu(w) 'decay' : may be loan from 庮 late OC *zu (or *zü [zy]) < OC *s-lu' or *s-yu', an otherwise unattested prefixed variant of *lu(') or *yu(') 'decay'

(This is the only definite *zy-root in Matisoff [2003]. Perhaps its rarity implies a foreign source: i.e., late OC.)

PTB *(s-)lya 'tongue' : 舌 OC *ml(y)at 'id.', maybe 臄 OC *gak if it is from *glak (no OC-internal evidence for an *-l-)

also 舐 OC *mle' 'lick' (Schuessler 2007: 467)

PTB *l(y)ak/*m-d(y)ak/*l(y)ang 'good/beautiful' : possibly loans from

麗 late OC *leh < OC *reks 'beautiful' (< 'well-proportioned') < 'pair'

良 late OC *lïang < OC *rang

or should 麗 and 良 be reconstructed with OC *Cə-l-?

Is PTB *-d(y)- an attempt to imitate a flap pronunciation of OC *r?

Cognate to 易 OC *lek < ?*lyak 'at ease'?

PTB *(g-)l(y)ak 'hand/arm' : 翼 OC *lək 'wing'

Further comparisons in Schuessler (2007) but not in Matisoff (2003):

PTB *m/s-lyak 'eat, feed' (of animals) : 食 OC *m-lək 'eat', 飤 OC *s-lək-s

PTB *s-lyam 'flame' : 炎 OC *wl(y)am 'id.'

PTB *lep/*lyap 'flat object' : 葉 OC *l(y)ap 'leaf', 蝶 OC *lep < ?*lyap 'butterfly', 牒 OC *lep < ?*lyap 'tablet'

PTB *s-r(y)ang/*sring 'alive/green/raw' : 生 OC *sreng or *r-seng 'live/bear/be born'; MC Shæng implies a variant ending in *-(y)ang

maybe also 靑 OC *C-seng 'green, blue'

Could the PST form have *r-s-? (Benedict/Matisoff's PTB does allow *r-s only before *a: *rsa 'vein', *r-sang 'lizard', *r-sak 'breath')

PTB *s-r(y)ak 'stay the night' : not cognate to 宿 OC *suk 'id.'

PTB *s-ryak/*s-rik 'pheasant' : 翟 OC *lekw < ?*lyakw 'id.'

PTB *s-ryak 'grease/oil' : 液 OC *lyaks 'fluid' (the *r : *l correspondence is irregular, so could the PTB form be a loan from an OC dialect with *r < *l?)

PTB *(z)ryang 'uncle/a superior' : not cognate to 尚 OC *dang(')s 'superior', which is probably cognate to 上 OC *dang(') 'rise', possibly even 登 OC *təng 'climb' if OC *d- in the above words is from *N-t-

superficially looks like 郎 OC *rang 'gentleman' (< 'person working in a palace') < 'corridor of a palace' (hence the graph has the element 阝 indicating places, not people)

PTB *rya-t 'laugh' : not cognate to 口+至 OC *dit 'laugh'

PTB *s-r(y)am 'sharp' : 銛 OC *s-hl(y)am 'id.' (the *r : *hl correspondence is irregular, so could the PTB form be a loan from an OC dialect with *[h]r < *hl?)

PTB *g-r(y)ap 'stand' : 立 OC *rəp 'stand'

PTB *s-r(y)ik 'louse' : 蝨 OC *srit < ?*srik 'louse'

(PTB *s-ryut 'worse' : 劣 OC *rot 'inferior' [Schuessler 2007: 358]; added 11.10.22:24)

PTB *g-ryum 'salt' : 鹽 OC ?*rə-yam 'salt'

This took a lot longer to compile than I expected and I'm out of time, so I'll stop here. I originally intended to conclude this post with an analysis, but now I'd like to postpone that until I look at other correspondences of PTB *-y- in OC.

Next: Which OC consonants correspond to PTB grave consonant + *-y- clusters?

07.11.9.1:33: EIGHT REASONS *Y (PART 4)

This may count as a reason 'y not'.

If Old Chinese had a *-a(C) : *- ya(C) distinction (with *-C = *-', *-'s, *-s, *-ks, *-k) reflected in Middle Chinese as *-ïə('/h) : *-yæ('/h), then this distinction could have occurred after grave consonants, as in Written Tibetan. Yet the distribution of *-y- in this environment seems to be the opposite of the WT distribution. Looking at Starostin's database and converting his OC *ia to my *ya, I found that the only initials (using my reconstruction) preceding *ya were coronals:

*t-, *th-, *d-

*rtsh-/*tshr-, *s-

*hl-, *l-, *sl-

The only exceptions might have contained velars:

車 'chariot' if it was *khya rather than *k-hlya (if from a root *la 'to carry' [Sagart 1999: 204] or borrowed from an Indo-European word for 'wheel' with *l)

邪 'depravity' if it was *slngya (is its phonetic 牙 *rnga?) rather than *slya or *snglya

Why are there no reconstructible instances of OC*pya or *kya? Did OC *-y- disappear after grave (velar and labial) initials without leaving a trace in MC? Did OC and Proto-Tibeto-Burman (or other Sino-Tibetan languages, if PTB never existed) each retain only part of the original distribution of Proto-Sino-Tibetan *-y-?

Conversely, is OC *-y- an innovation on the Chinese side? Did *-y- develop after coronals in some OC dialects? Or did coronals palatalize in those dialects: e.g., *s- > *sh-? That might be expected before palatal vowels like *i or even *e, but why would it happen before a nonpalatal vowel *a? Is there any language in which *sa became sya or sha?

A third explanation is that some OC dialects had preserved a set of palatal or palatalized consonants in addition to the dental / alveolar set.

If such dialects existed (and I doubt they did), their distinction between palatal(ized?) and dental / alveolar consonants is not at all reflected in sinography, implying that the originators of the script had merged the two series. For example, 也, representing the final particle which could be reconstructed as *l^yay', or *ʎay', is phonetic in 匜 *lay 'ritual vessel' and other graphs whose readings had no initial palatal(ized) consonants.

I would prefer to reconstruct 也 'final particle' as *lyay' with the same initial consonant as 匜 *lay 'ritual vessel'. This solution allows me to reconstruct one root initial *l for both readings of 蛇:

*m-lyay 'snake' (with the prefix *m- for small animals)

*lay 'pretentious'; also the second half of the disyllabic, partly reduplicative word 委蛇 *'waylay 'compliant'

Moreover, I would expect conservative palatal(ized?) consonants to correspond to Matisoff's PTB palatals.

Next: What do those PTB palatals correspond to in OC?

07.11.8.2:40: EIGHT REASONS *Y (PART 3)

Laurent Sagart's The Roots of Old Chinese (1999) has been a major influence on my thoughts about Chinese for the past eight years. After I read it for the first time in 2000, I abandoned Starostin's Old Chinese reconstruction and adopted Sagart's modification of Baxter's (1992) reconstruction.

One thing that has long bothered me about that 1999 reconstruction is its inability to explain why OC *-a in nonemphatic ('Type B') syllables became both *-ïa and *-yæ in Middle Chinese: e.g.,

余 OC *^bla > MC *yïə 'I' (p. 241)

邪 OC *^bs-la > MC *zyæ 'depravity' (p. 240)

(I have replaced others' MC forms with my reconstructions. These two words belonged to different MC rhyme classes, regardless of reconstruction. The prefix *s- does not have any effect on the rhyme in Sagart's reconstruction. See below for an instance of OC *^bs-la > MC *zïə.)

Pulleyblank proposed that the split in MC rhymes reflected OC vowel length:

余 OC *-a > MC *-ïə

邪 OC *-aa > MC *-yæ

But I think vowel length had a very different role in OC*.

Starostin reconstructed a diphthong for the second type of case:

Starostin OC *-a > MC *-ïə

Starostin OC *-ia > MC *-yæ

He could not reconstruct *-ya because his OC had no non-coda *y.

A couple of years ago, I wondered if a diphthong *ia or a medial *y could be avoided by reconstructing *-æ:

OC *-a > MC *-ïə

OC *-æ > MC *-yæ

Did OC have a seven-vowel system?

I don't think so, because there is no evidence for *-æ(C)(C) rhyme categories distinct from *-a(C)(C) rhyme categories in OC. In Shijing, *-a words interrhyme with *-æ words: e.g., this pair in 4.2.1.4 involving a word which also happens to be the name of its rhyme category:

I would rather reconstruct *-ya instead of *-æ. There was no separate rhyme category for 邪, because 邪 OC *slya had the same vowel as 魚 *nga.

邪 also had a second reading meaning 'slow'. In Sagart's system, both 邪 'depravity' (> MC *zyæ) and 邪 'slow' (> MC *zïə) would be reconstructed as OC *^bs-la. In my current system, 'depravity' was OC *slya with medial *-y- and 'slow' was OC *sla without it. (OC *sla 'slow' was also written as 徐 with 余 OC *la as its phonetic.)

All of the above examples involve open syllables. However, an OC *a : *ya distinction can also account for rhyme category splits in closed nonemphatic syllables. The basic pattern is:

OC *-aC > MC *-ïaC

OC *-yaC > MC *-ieC

*In short, I suspect that MC 'level tone' often reflects OC long vowels and MC 'rising tone' often reflects OC short vowels. The patterns of correspondence are more complex than that and require an full explanation elsewhere.

07.11.8.2:40: EIGHT REASONS *Y (PART 3)

Laurent Sagart's The Roots of Old Chinese (1999) has been a major influence on my thoughts about Chinese for the past eight years. After I read it for the first time in 2000, I abandoned Starostin's Old Chinese reconstruction and adopted Sagart's modification of Baxter's (1992) reconstruction.

One thing that has long bothered me about that 1999 reconstruction is its inability to explain why OC *-a in nonemphatic ('Type B') syllables became both *-ïa and *-yæ in Middle Chinese: e.g.,

余 OC *^bla > MC *yïə 'I' (p. 241)

邪 OC *^bs-la > MC *zyæ 'depravity' (p. 240)

(I have replaced others' MC forms with my reconstructions. These two words belonged to different MC rhyme classes, regardless of reconstruction. The prefix *s- does not have any effect on the rhyme in Sagart's reconstruction. See below for an instance of OC *^bs-la > MC *zïə.)

Pulleyblank proposed that the split in MC rhymes reflected OC vowel length:

余 OC *-a > MC *-ïə

邪 OC *-aa > MC *-yæ

But I think vowel length had a very different role in OC*.

Starostin reconstructed a diphthong for the second type of case:

Starostin OC *-a > MC *-ïə

Starostin OC *-ia > MC *-yæ

He could not reconstruct *-ya because his OC had no non-coda *y.

A couple of years ago, I wondered if a diphthong *ia or a medial *y could be avoided by reconstructing *-æ:

OC *-a > MC *-ïə

OC *-æ > MC *-yæ

Did OC have a seven-vowel system?

I don't think so, because there is no evidence for *-æ(C)(C) rhyme categories distinct from *-a(C)(C) rhyme categories in OC. In Shijing, *-a words interrhyme with *-æ words: e.g., this pair in 4.2.1.4 involving a word which also happens to be the name of its rhyme category:

I would rather reconstruct *-ya instead of *-æ. There was no separate rhyme category for 邪, because 邪 OC *slya had the same vowel as 魚 *nga.

邪 also had a second reading meaning 'slow'. In Sagart's system, both 邪 'depravity' (> MC *zyæ) and 邪 'slow' (> MC *zïə) would be reconstructed as OC *^bs-la. In my current system, 'depravity' was OC *slya with medial *-y- and 'slow' was OC *sla without it. (OC *sla 'slow' was also written as 徐 with 余 OC *la as its phonetic.)

All of the above examples involve open syllables. However, an OC *a : *ya distinction can also account for rhyme category splits in closed nonemphatic syllables. The basic pattern is:

OC *-aC > MC *-ïaC

OC *-yaC > MC *-ieC

*In short, I suspect that MC 'level tone' often reflects OC long vowels and MC 'rising tone' often reflects OC short vowels. The patterns of correspondence are more complex than that and require an full explanation elsewhere.

07.11.7.7:55: EIGHT REASONS *Y (PART 2)

Suppose that

- Sino-Tibetan is real: Chinese is related to Tibetan, Burmese, Tangut, etc.

- Tibeto-Burman is real: Tibetan, Burmese, Tangut, etc. comprise a branch of Sino-Tibetan distinct from Sinitic.

- Proto-Tibeto-Burman had a non-coda *y

- Old Chinese had no non-coda *y; Middle Chinese initial *y- is from OC *l in nonemphatic syllables, and a medial *-y- may have developed in some nonemphatic syllables before front vowels.

Given those assumptions, did Proto-Sino-Tibetan have a non-coda *y or not?

Starostin's solution, judging from his online database, was to reconstruct PST *y and *l:

The merger of PST *y and *l to OC *l is unusual. I would expect a change in the opposite direction:

*l > *y

Could the *y-lessness of OC be conservative? If it is, then it could be projected back into PST, and PTB *y could have been weakened from an original palatalized*l^y or a palatal *ʎ that retained its lateral quality in OC:

This *l^y or *ʎ would belong to the class of consonants that Benedict (in Matisoff 2003: 65) reconstructed as *tsy, *dzy, *sy, *zy. However, there are three problems:

- There already is an *ly in Benedict's PTB reconstruction. If PTB *y- < PST *l^y-/ʎ, then where did PTB *ly- come from?

- The frequency of PTB medial *-y- suggests that PST *Cl^y-/*Cʎ- clusters were also frequent, which seems unlikely.

- Variation occured between PTB *y and *r, not PTB *y and *l (Matisoff 2003: 45). This may suggest that PTB *y in part or in whole came from PST *r^y. But there already is an *ry in PTB, and if PTB *y- < PST *r^y-, then where did PTB *ry- come from?

Instead of a lateral or a rhotic, one could project the uvular *ɢ first reconstructed for OC by Pan Wuyun (1997) and recently adopted by Baxter and Sagart (2007) back into PST:

However, this would still not account for the origin of *y in PTB clusters like *py-, *ky-, etc. unless they came from *Cɢ- (> ?*Cɣ-) clusters (cf. the Cɣ clusters of Japhug rGyalrong mostly from Proto-rGyalrong *Cw- [not *Cɢ-; Jacques 2003: 331-332; other sources here*]).

For the time being, I wonder if the simplest solution is best:

Next: What made me say *ya to yod?

*tp-, *kp-

*rb-

*pk-

07.11.6.1:30: EIGHT REASONS *Y (PART 1)

Karlgren (1940, 1957) reconstructed *y in roughly half the syllables of Middle Chinese and projected this feature back into Old Chinese.

As far as I know, Pulleyblank (1962) was the first to propose that MC initial and medial*y was secondary. He proposed that MC initial and medial*y developed in OC syllables with long vowels. Hence he did not reconstruct any OC*y at all (1962: 141-142). Some later scholars (e.g., Starostin 1989 and Sagart 1999) also did not reconstruct OC *y in non-coda position.

However, I have never been totally convinced that OC was *-y-less. Pulleyblank later brought *y back into his OC reconstruction and Schuessler (2007: 96) also included a non-coda *y in his OC reconstruction. This *y (Bodman's 'primary yod') does not correspond in a simple manner to Karlgren's OC *y.

One reason that I have long suspected that OC had a non-coda *y is that other archaic Sino-Tibetan languages have or had it: Written Tibetan, Written Burmese, rGyalrong (even at the proto-rGyalrong level [Jacques 2003]), and Tangut*. Matisoff (2003: 15, 59) reconstructed non-coda *y in Proto-Tibeto-Burman.

OC superficially looks like Written Tibetan and, to a lesser extent, rGyalrong and the non-ST language Old Khmer, because these languages are 'edgy' rather than 'nuclear' in terms of their phonological typology. All of those other 'edgy' languages had y. Ubykh, perhaps the 'edgiest' language ever**, also had y.

Of course, I cannot simply reconstruct *y in OC just because presumably related and typologically similar languages have or had y. Probability is not precise, and it's perfectly possible for a language to lack non-coda y (e.g., northern Vietnamese which has [z] < *[k]y). Perhaps *y-loss was a Chinese innovation absent in the rest of Sino-Tibetan. I will present better reasons for OC non-coda *y in the next seven installments.

*All Tangut reconstructions known to me have *y and/or a medial *-i- which may have been phonetically [y]: Nevsky (1960), Kychanov and Sofronov (1963), Nishida (1964), Hashimoto (1965), Sofronov (1968), Shi et al. (1983), Li Fanwen (1986), Gong Hwang-cherng (in Li Fanwen 1997), and Arakawa (1999).

The Tibetan transcriptions of Tangut contain both y- and -y-, though the latter is not as common as the above reconstructions would imply.

Nevsky (1960 II: 204) also reconstructed a palatalized *k^y in

TT0226 GOLD 1.66 (Nevsky's *k^ye, *k^yä)

that corresponds to others' reconstructions with -y- or -i-: e.g., Gong's kiẹ and Arakawa's ?kyiq (my guess using his 1999 system). But not all researchers reconstruct -y- or -i-:

Nishida's kE (1964: 191, 194; 1966: 336 - yet in 1964: 51, 60, he reconstructed rhyme 1.66 as -Ịě - and stated that 1.66 did not occur with velar initials [though it clearly did]. -E is his reconstruction for rhyme 1.33.)

Hashimoto's käy (my guess using his system)

Sofronov's kẹi

Shi et al.'s kã (my guess using their system)

Li Fanwen's kee

**Ubykh is 'edgy' in the sense that its 84 consonants bear a far greater functional load than its two phonemic vowels. However, it is not a monosyllabic or sesquisyllabic 'edgy' language, so it has simple syllable structures (e.g., CVC) unlike Written Tibetan which allows CCCCVCC syllables like bsgrags 'proclaimed'.

07.11.5.1:45: SAT-UMPED

薩 Tangut period NW Chinese ?*sa, the meaningless sinograph used to transcribe

TT0981 KILL sja 1.20

in the Pearl, has an unusual structure.

Like KILL which has WOOD on top, 薩 has a seemingly meaningless botanical top element (艹 'grass'). I presume that 艹 in 薩 was intended for 'graphic harmony' with 菩, the graph with which it is often paired. (菩薩 Late Old Chinese *bɔsat represented Skt bo[dhi]sat[tva].) Perhaps WOOD in KILL is meant to be a phonetic element derived from the top of 薩 TPNWC ?*sa. One might even guess that KILL as a whole is a distortion of 薩, but there is no resemblance between its bottom elements and the bottom elements of 薩.

One might expect the bottom of 薩 to be its phonetic, just as the bottom of 菩 is its phonetic. Yet as far as I know, there is no graph 阝+產. The function of 阝 'mound' is unknown. The lower right element, 產 LOC *ʂɛn 'produce', is an odd choice for a phonetic for three reasons:

- it has a retroflex onset instead of *s-

- it has *ɛ instead of *a

- it has a nasal coda rather than *-t

In the Eastern Han transcription material I have on hand (Coblin 1983), 薩 represented

Skt sat (38, 39, 74); Pali or some other Middle Indic sat (182)

Skt sad (138)

Skt sar (45)

Skt saartha (228)

Skt tha (!?) (55)

All of those syllables only vaguely resemble 產 LOC *ʂɛn.

Potential phonetics pronounced LOC *sat

摋 'slap from the side'

殺+米 'let go' (later written 撒)

contained the semantically unappealing phonetic 殺 'kill'. 散 *san meant 'scatter'. Perhaps the creator of 薩 wanted to avoid transcribing Buddhist terminology with elements bearing negative connotations. 產 'produce' is more positive.

Next: A puzzling production.

07.11.4.23:56: YET ANOTHER 'KILLER' PROBLEM

殺 'kill' MC *ʂɛt < as if from OC *r-set (but rhymed with *-at words in Laozi as if it were *r-sat)

belongs to a class of words that apparently had *a in Old Chinese but had Middle Chinese readings suggesting Old Chinese *e. Two others are

山 'mountain' MC *ʂɛn < as if from OC *rsen (but rhymed with *-an words in Shijing as if it were OC *rsan and transcribed Greek -xan-)

獲 'to catch' MC *ɦwɛk < as if from OC *r-wek (but rhymed with *-ak words in Shijing as if it were OC *r-wak)

Yesterday, it occurred to me that these were not cases of irregular development or dialect mixing in the admittedly heterogenous Middle Chinese rhyme dictionary tradition. While trying to find some justification for Gong's -j- in Tangut

TT0981 KILL sja 1.20

I wondered if

- the OC cognate of sja 1.20 was *r-sjat

- MC *ɛ was the regular development of later OC *rja:

*r-sjat > *srjat > *srjet > *ʂɛt

I noticed at Google Books that Baxter (1992) reconstructed *srjat and *srjan corresponding to my *r-sjat and *rsjan. I cannot access the page with 'catch', but Sagart (1999) and Jacques (2006), who both use a modification of Baxter's system, reconstruct it as *wrak (rewritten in my emphatic notation) without *-j-. I would consistently reconstruct 'catch' as *r-wjak.

My proposal of OC *rja as a source of MC *ɛ not only leads to a nice correspondence between Gong's Tangut sja 1.20 and the root of OC *r-sjat, but also eliminates what appeared to be the irregular raising of OC *a. Reconstructing *ja in OC syllables without *r and/or emphasis also has explanatory power. (Note that my OC *ja has what Bodman would call a 'primary yod'; this *j does not necessarily correspond to the *j of Karlgren's and Li Fang-kuei's OC, which is equivalent to nonemphasis in my reconstruction.)

However, if this OC *-j- is real, *-j-less cognates seem to indicate that it was an infix (cf. Sagart's now defunct *-j- infix hypothesis for syllables that I reconstruct as nonemphatic):

殺 'kill' OC *r-sjat

cognate to 摋 OC *sat 'slap from the side'

獲 'to catch' OC *r-wjak

cognate to 穫 OC *wak 'to harvest'

I know of no cognates of 山 OC *rsjan 'mountain', so I am not sure if *-j- really is an infix in it. Perhaps medial *-j- reflects an earlier unstressed vowel *i which could have been in both prefixes and roots:

殺 OC *r-sjat < ?*ri-sat

獲 OC *r-wjak < ?*ri-wak

山 OC *rsjan < ?*risan (disyllabic root?)

Next: 'Eight' reasons for and against *ja.