In "Cold Storage", I stated that GSR 727 was an Old Chinese *ts-series. All but one of its members have affricate initials in Middle Chinese:

GSR 727 letters

MC grade

MC initial OC initial
t, f', h' I *ts- *ts-
g' *dz- *dz- < ?*N-ts(h)-
f, u, v, x, y III *ts- *Cɯ-ts-
a-c, d-e, f, z *tsh- *Cɯ-tsh-
g-i, j-k, l, m, j' *dz- *Cɯ-dz- < ?*Cɯ-N-ts(h)- or ?*Nɯ-ts(h)-
n, o-p, q, i' *tʂ- *rɯ-ts-
r, s *dʐ- *rɯ-dz- < ?*rɯ-N-ts(h)-
a'-e' *ɕ- *sɯ-tɯ-ts-

The MC grade I syllables had automatic emphasis in OC conditioned by their nonhigh vowel: e.g.,

OC *tsaŋ > *tsaŋ > MC *tsaŋ

They may or may not have had low vowel presyllables that were later lost: e.g.,

OC *Cʌ-tsaŋ > *tsaŋ > MC *tsaŋ

The MC grade III syllables had high vowel presyllables in OC that blocked the development of emphasis and conditioned upward vowel warping before being lost: e.g.,

OC *Cɯ-tsaŋ > *Cɯ-tsɨaŋ > MC *tsɨaŋ

Rhotic presyllables conditioned the retroflexion of root initials:

OC *rɯ-tsaŋ > *rɯ-tsɨaŋ > *r-tsɨaŋ > *tsrɨaŋ > *tʂrɨaŋ > MC *tʂrɨaŋ

727a'-e' 䵼 'cook'* has an unexpected palatal fricative initial in MC: *ɕɨaŋ. Since I don't know of any other cases of MC *ɕ- in affricate series, I suspect that it must have a complex origin. I proposed *sɯ-tɯ-ts- with a double presyllable by analogy with *sɯ-t- (my equivalent of Sagart's [1999: 69] *bs-t-), one source of MC *ɕ-:

727a'-e' 䵼 OC *sɯ-tɯ-tsaŋ > *sɯ-tɯ-tsɨaŋ > *stɕɯ-tsɨaŋ > *stɕɨaŋ > MC *ɕɨaŋ

cf. 725n-p 賞 OC *sɯ-taŋʔ > *sɯ-tɨaŋʔ > *stɕɨaŋʔ > MC *ɕɨaŋʔ

whose phonetic is 725a-c 尚 *Cɯ-daŋs (?*Nɯ-taŋs) > MC *dʑɨaŋh

*Written as 727f 將 OC *Cɯ-tsaŋ 'bring' (phonetic; later 'general') atop 鼎 834a-f OC *teŋʔ 'cauldron' (semantic). COLD STORAGE: WAS GSR 703 A RHOTIC SERIES?

Sagart (1999: 68) reconstructed GSR 703c 凔 'cold' as *as-hraŋ (equivalent to my *s-hraŋ) which shared a root with 755l 涼 Old Chinese *BCə-raŋ (equivalent to my *Cɯ-raŋ) 'cool'. I don't think series 703 (倉 'granary, later 'storehouse') had an *r for two reasons:

- there are no Middle Chinese *l-readings from OC *r-

- OC *Cr-clusters generally conditioned Middle Chinese Grade II; why would *s-hr- be an exception?

Here's how I would reconstruct series 703:

Prototype: *tshaŋ (emphasis automatic before the nonhigh vowel *a?)

703a-b 倉 MC *tshaŋ < OC *tshaŋ 'granary'

7.4.00:38: Schuessler (2007: 176) derived the MC initial of this word from OC *k-s-, but I think MC *ʂ- was the reflex of OC *k-s-. He linked 703a-b to 727g' 藏 'conceal' (which he reconstructed as *dzâŋ) and Written Tibetan gsang-ba 'secret'. 727g' is reconstructible as *asraŋ in Sagart's system, but 727 is a *ts-series, so I prefer to reconstruct 727g' as *N-tsaŋ. 703a-b OC *tshaŋ could be from an earlier *k-tsaŋ or *s-tsaŋ.

703c 凔 MC *tshaŋ < OC *tshaŋ and MC *tʂhɨaŋ < OC *rɯ-tshaŋ 'cold'

7.4.0:08: Although 703c and 755l OC *Cɯ-raŋ share a rhyme, I regard that as a coincidence and not as evidence for a common root.

7.4.0:31: Cognate to 812h' 凊 *tshieŋ < *sɯ-seŋ 'cold' (contra Schuessler [2007: 176]) if OC *tshaŋ < *s-saŋ?

703d 滄 MC *tshaŋ < OC *tshaŋ 'name of a river'

703e 蒼 MC *tshaŋ < OC *tshaŋ 'green'

cognate to 812c' 青 MC *tsheŋ < OC *tsheŋ 'green', whose phonetic is 812a 生 MC *ʂɨeŋ < OC *rɯ-seŋ 'live' (perhaps cognate if 'green' < 'color of life')

7.4.0:04: Could *tsheŋ < *s-seŋ (with automatic emphasis before a nonhigh vowel)?

703f 鶬 MC *tshaŋ < OC *tshaŋ 'crane'; MC *tshɨaŋ < OC *Cɯ-tshaŋ 'to tinkle'

703g 搶 MC *tshɨaŋ(ʔ) < OC *Cɯ-tshaŋ(ʔ) 'to rush'; MC *tʂhɨaŋ < OC *rɯ-tshaŋ 'to beat'

703h 瑲 MC *tshɨaŋ < OC *Cɯ-tshaŋ 'to tinkle'

703i 槍 MC *tshɨaŋ < OC *Cɯ-tshaŋ 'sharp-pointed stick'

703j 蹌 MC *tshɨaŋ < OC *Cɯ-tshaŋ 'move in a stately way; to dance'

703k 牄 (variant of 703j)

703l 創 MC *tʂhɨaŋ < OC *rɯ-tshaŋ 'to wound'; MC *tʂhɨaŋh < OC *rɯ-tshaŋs 'start work; commence, create'

703m 愴 MC *tʂhɨaŋh < OC *rɯ-tshaŋs 'grieved'

703n 瘡 MC *tʂhɨaŋ < OC *rɯ-tshaŋ 'boil, tumor'

7.4.1:14: Other scholars have reconstructed series 703 without a rhotic root initial:

Sinograph GSR MC OC
Karlgren Wang Li Li Fanggui, Baxter, Schuessler 2007 Sagart Starostin, Zhengzhang Pan This site
703a-b *tsh- *tsh- *s-hr- *sh- *skh-
703g *Cɯ-tsh-
703m *tʂh- *tʂh- *tʃh- *tshr- ?*s-hr- *shr- *skhr- *rɯ-tsh-

(Reconstructions other than Schuessler's, Sagart's, Starostin's and mine were generated using eastling.org's OC generator.)

I am not certain whether Sagart would reconstruct *s-hr- throughout series 703. I assume Sagart would reconstruct 703g 搶 with the initial of its homophone 703f 鶬.

For a while I agreed with Starostin (and, if I recall correctly, Schuessler 1987) who reconstructed OC *sh- as a source of MC *tsh-, but an aspirated s is an unusual consonant. Perhaps *sh- or *hs- was a short-lived transitional stage between *s-s- and *tsh-. (I was unaware that Zhengzhang also reconstructed an aspirated *sh- until recently.)

I don't understand why Pan reconstructed a velar in a series without any MC velars. Does he think series 703 morphemes have velar-initial cognates in other series? WAS GSR 1038 A RHOTIC SERIES?

Out of the 13 Middle Chinese syllables in the previous post, only 礐 *lɛk was written with a nonrhotic, velar phonetic

1038a-c 學 MC *ɣæwk < OC *rguk 'learn'

< *r-N-kuk, cognate to 1039a-d 告 MC *kowk/*kawh < OC *kuk(-s) 'announce, inform'?

All but one of the other members of GSR 1038 had velar initials in Middle Chinese. All belong to Grades I or II, indicating emphasis in Old Chinese. Therefore 1038 is a pure emphatic series. Moreover, the alternation of MC velars with glottal stop (1038h) may point to OC uvulars which I assume to be emphatic.

1038a-c 學 MC *ɣæwk < OC *rɢuk 'learn'

1038d-e 斅 MC *ɣæwh < OC *rɢuk-s 'teach'

1038f 覺 MC *kæwk/h < OC *rquk(-s) 'awake'

1038g 嚳 MC *kowk < OC *quk (a name)

1038h 鷽 MC *ɣæwk < OC *rɢuk and MC *ʔæwk < *rʔuk < *rʌ-quk 'name of a bird'

I tentatively assume that *-q- became glottal stop before *CV-

1038i 攪 MC *kæwʔ < OC *rquʔ

1039 only contains MC velar initials followed by Grade I (except for one instance of Grade II). Hence I could re-reconstruct 1039 as another uvular series:

1039a-d 告 MC *kowk/*kawh < OC *quk(-s) 'announce, inform'?

1039e 誥 MC *kawh < OC *quk-s 'announce, inform'

1039f-g 郜 MC *kowk/*kawh < OC *quk(-s) 'place name'

1039h 皓 MC *kawʔ < OC *quʔ 'pure, bright'

1039i 晧 MC *ɣawʔ < OC *N-quʔ 'bright, white'

cognate to 1039h?

1039j 浩 MC *ɣawʔ < OC *ɢuʔ 'great waters; vast, great'

1039k 梏 MC *kowk < OC *quk 'manacles'

also used to write 1038f (which Karlgren defined in his 1039k entry as 'straight', though he had a different gloss in the entry for 1038f)

1039l 牿 MC *kowk < OC *quk 'pen, stable'

1039m 酷 MC *khowk < OC *qhuk 'cruel'

1039n 鵠 MC *ɣowk < OC *ɢuk 'name of a bird (crane? snow-goose? swan?)'

cognate to 1117b 鶴 MC *ɣak < OC *ɢak 'crane'?

also MC *kowk < OC *quk 'center of target'

and used to write 1039j

1039o 窖 MC *kæwh < OC *rquks 'cave, cellar; secluded'

Most of the MC readings of 礐 can be derived from a uvular base:

MC *khæwk < OC *rqhuk

MC *ɣok < OC *ɢok (with an unexpected mid vowel)

MC *ɣæwk < OC *rɢuk

MC *ŋæwk < OC *N-ɢuk

(did *N-ɢ- > *ŋ- but *N-q(h)- > *ɢ-?)

MC *ʔowk < OC *Cʌ-quk

The two MC *l-readings are both Grade II, so I would expect their OC initial to have been *tr- without any uvular. Perhaps OC emphatic *r was uvular [ʀ], and the cluster *trɢ- [tˁʀɢ] merged with *tr- [tˁʀ], becoming MC *l-:

MC *lɛk < *reʳk < *drəʳk < OC *trɢək or *trɢek (with unexpected mid vowels)

MC *læk < *raʳk < *draʳk < OC *trɢak (with an unexpected low vowel)

cf. 823h 冷 MC *læŋʔ < *raʳŋʔ < *draʳŋʔ < OC *t-raŋʔ 'cold'

(the root is *raŋ, also in 755l 涼 OC *Cɯ-raŋ 'cool' and, according to Sagart [1999: 68], 703c 凔 'cold', which he reconstructed as *as-hraŋ, but I have my doubts)

I've been skeptical of my own OC *tr- > *dr- > MC *l- proposal because there are no parallel changes

OC *kr- > *gr- > MC *l-

OC *pr- > *br- > MC *l-

However, the voicing of an acute consonant before a liquid has parallels in these changes proposed by Sagart (1999: 69):

OC *sr- > (via *zr- > *zd-?) > MC *dz-

OC *sl- > (via *zl-?) > MC *z-


According to the Yunjing rhyme table, there are only 13 syllables that fit the title description. Reconstructions are mine.

*læwŋ (also Grade I *loŋ as well as Grade II *ʂæwŋ)

*læwk (also Grade I *lak; see the previous post)

*lɛj (also Grade I *ləj(ʔ/h))

*lwɛj (also Grade III *lwi(ʔ))

*lɛn (also Grade I *lan(h))

𡰠 *lwɛn (also Grade I *lwaʔ/h, Grade III *lwie)

*lɛw (al
so Grade IV *lew)


*læŋʔ (also Grade IV *leŋ(ʔ))

*lɛŋ (also Grade I *laʔ, Grade IV *lek; the Grade II reading isn't in Guangyun or Jiyun)

*lɛk (also Grade I *ɣok, *ʔowk, as well as Grade II *khæwk,*ɣæwk, *ŋæwk, and *læk)

*lwɛk (also Grade IV *lek; the Grade II reading isn't in Guangyun or Jiyun)

*lɛmʔ (also Grade III *kɨemʔ, Grade IV *tshem)

All - with the exception of 藞 - have non-Grade II readings, and all - with the exception of 礐 - have Old Chinese *r-phonetics. I tentatively assume that these syllables had initial *tr- in Old Chinese:

OC *trV > *trVʳ > *drVʳ > *rVʳ > *lVʳ > MC *l- + Grade II

cf. OC *rtV > *rʈVʳ > *ʈVʳ > MC *ʈ- + Grade II

This *tr- might have corresponded to *tʌr- in a more careful style of pronunciation. The non-Grade II readings might reflect dropped or fused presyllables:

OC emphatic presyllables > MC Grades I and IV

瀧 OC *Cʌ-roŋ > *roŋ > MC *loŋ (Grade I)

but OC *t-roŋ > *troʳŋ > *droʳŋ > *roʳŋ > *raʳwŋ > MC *læwŋ (Grade II)

顟 OC *Cʌ-riw > *rew > MC *lew (Grade IV)

but OC *t-riw > *triʳw > *driʳw > *riʳw > *reʳw > MC *lɛw (Grade II)

OC nonemphatic presyllables > MC Grade III

臉 OC *kɯ-remʔ > *kremʔ > MC *kɨemʔ (Grade III)

but OC *t-remʔ > *treʳmʔ > *dreʳmʔ > *reʳmʔ > *reʳmʔ > MC *lɛmʔ (Grade II)

Nonemphatic OC *tr- merged with *r- and became MC Grade III *l-: e.g.,

六 OC ?*(t-)ruk (cf. Written Tibetan drug) > MC *luk (Grade III)

離 OC *t-raj > *driej (used to transcribe -δρεια in Αλεξανδρεια) > MC *lie (Grade III)

離 has the phonetic 离 MC *ʈhie ~ *lie (both Grade III). Since I reconstruct *t- for 離 in OC, I could also reconstruct *t- for 离 in OC: *t-hraj. The MC *l-reading could be from OC *(t-)raj. OC *t-hr- could be from an earlier *t-k-r-.

Emphatic OC *thr- may have become MC *ʈh- followed by MC Grade II, though I don't know of any phonetic series with *l- alternating with MC *ʈh- + Grade II. MYSTERY OF THE MULTI-COLORED OX

Last night, I found a graph with some strange reconstructions in Pulleyblank's 1991 Lexicon:

犖 'multi-colored ox'

Early Middle Chinese *laɨwk > *lœœwk

Late Middle Chinese *lwaawk

(both of the above are Grade II)

Early Mandarin *lwaw

The readings of 犖 have two odd characteristics.

First, *l- and Grade II rhymes in Middle Chinese usually do not go together. The table below shows the usual pattern of correspondences:

Old Chinese

Middle Chinese initial

Middle Chinese grade




*CV-l- (with late presyllable loss)

*Cr- (partly from even earlier *rC-)

*C- (but not *l-)


Perhaps syllables like 犖 had initial *Cr-clusters that irregularly simplified to *l- after the retroflexion that led to Grade II:

*CrV... > *CrVʳ... > *rVʳ... > *lVʳ... > *l- + Grade II

(6.30.0:06: I suspect that OC *tr- might have become *l- + Grade II, whereas OC *rt- became *ʈ- + Grade II. Hence 犖 could have been *trakw < ?*t-raw-k in Old Chinese.)

Second, Chinese languages tend to avoid -wVw-type rhymes, unlike Tangut which had the rhyme -wew in Gong's reconstruction. Yet the Early Mandarin reading of 犖 in Menggu ziyun is *lwaw.

I don't know why Pulleyblank reconstructed medial *-w- in Late Middle Chinese *lwaawk but not Early Middle Chinese *laɨwk or *lœœwk. Did he project it backwards from Menggu ziyun?

Jiyun has a second Middle Chinese reading *lak (Grade I; fanqie 歷+各) for 犖. This could go back to Old Chinese *(-)rak(w) or *Cʌ-lak(w) (with late presyllable loss). Since the phonetic of 犖 is an abbreviated form of 勞 OC *Cʌ-raw or *Cʌ-law, I assume the coda of 犖 was labiovelar *-kw.

Coblin's (2007) interpretation of Menggu ziyun also includes other *-waw syllables which had Grade II and the 覺 rhyme in Middle Chinese like 犖. I suspect that the *-aaw- of the 覺 rhyme in Late Middle Chinese fused into a rounded vowel *-ɔ- that then broke into *-wa-:

Late Middle Chinese *-aawk > pre-Early Mandarin *-ɔɣ > *-ɔw > Early Mandarin *-waw

This *-waw then became modern standard Mandarin -uo [wo]. Hence 犖 is Md luo.

Since *-w- in EM *-waw is due to post-LMC breaking, there is no need to reconstruct *-w- before *-aawk in LMC.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2008 Amritavision