Four days ago, I proposed that Tangut lh- may partly come from an earlier *kl-.

Last night, when looking for Ronghong Qiang reflexes of Matisoff's Proto-Tibeto-Burman *k-, I discovered that LaPolla and Huang (2003: 363) regarded RH hla 'slip' as originating from PTB *kla 'fall' (Matisoff 2003: 595 listed variants with g- and -k/-y/-t). The obvious Old Chinese cognate is 落 *qʌlak 'fall' (Matisoff 2003: 480; my reconstruction).

If *kl- became lh- in Tangut, Tangut words for 'fall' should begin with lh-, but they don't:

TT2481 FALL-INT0-TRAP lyị 1.67 < ?*s-la

TT2557 FALL gyi 1.11 < ?*N-kla or ?*N-gla

TT3305 FALL/LOSE kow 2.47 (cognate to TT5271?)

but Jacques (2007: 148) translated this as 'administrative official'

TT3566 TEARS-FALL liẹ 1.66 < ?*s-l-

TT4552 FALL lyï 1.29

TT5271 FALL kiew 1.44 (cognate to TT3305?)

Another word which might have begun with lh- but doesn't is

TT0238 lyụ 2.52

translated by Grinstead (1972: 128) and Jacques (2007: 112) as HEAD and by Li Fanwen (1986: 437) as 腦 BRAIN. (Amazingly, it was homophonous with TT4391 BODY lyụ 2.52, cognate to Written Tibetan lus!)

lyụ 2.52 looks like the Tangut cognate of Old Chinese *hlu' 'head', which may be cognate to / a loanword from Proto-Austronesian *quluH1 'head'. Matisoff (2003) doesn't list any *lu-like Proto-Tibeto-Burman reconstruction, but Sagart (2002: 5) listed Lushai lu 'head'.

Here are two possible paths of development based on my hypotheses about emphasis in Chinese:

*qulu' > *kɯlu' > *klu' > *hlu' [l̥uʔ]

*q- > *k- before high vowels

*qulu' > *qlu' > *χlu' > *xlu' > *hlu'

deemphasized to assimilate to nonemphatic *l

I could account for the two patterns of correspondence in Tangut in two different ways:

Solution 1:

*kVl- > Tangut l- ('fall', 'head')

*kl- > Tangut lh- ('manure')

cf. 屎 Old Chinese *hli' 'excrement'

Solution 2:

*kl- > Tangut l- ('fall', 'head')

*s-l- > lh-

this would necessitate deriving the root of gDong-brgyad rGyalrong tɯ ɣli 'manure' from*k-l- with a prefix *k- not in pre-Tangut

TT1313 MANURE lyï 1.29 would descend from a bare root, whereas TT1313 lhyị 2.60 would descend from an *s-prefixed root

did *s- cause *l to devoice and also trigger tensing of the vowel, resulting in TT?

Tangut lh- corresponds to eight gDong-brgyad rGyalrong onsets in Jacques (2003):

FrequencygDong-brgyad rGyalrongProto-rGyalrongTangutNotes
4j-*lj-lh-did Tangut have a devoicing prefix absent in the rGyalrong cognates of these lh-words?
1ɬ- ("aspirated lateral"; rare)(unknown)ɬ- also corresponds to l- in one case
1ɣl-*kl-lh- + tense vowelTangut lh- < ?*voiceless consonant + l-?
1pj-*pj-lh-unlikely correspondences; reject

I can boil this down to:

Proto-rGyalrong *l(j)- : Tangut lh- < *C-l- (without the tenseness trigger)

Proto-rGyalrong *C(ə-)l- : Tangut lh- < *C-l- (with the tenseness trigger)

Next: I don't think a simple prefix was enough to trigger tenseness. There may have been something more. What was it? PST 'TEN': VELAR AND UVULAR?

Middle Chinese 計 *keyh 'calculate' has a palatal-initial phonetic 十 *jip 'ten'. This implies that the MC *j of 'ten' originated from a palatalized velar:

Old Chinese *gip > MC *jip

'Calculate' did not palatalize because it had an emphatic initial:

Old Chinese *keps [qeps] > MC *keyh [qejh]

Since other syllables written with 十 were nonemphatic, I presume that the emphasis in 'calculate' was caused by an earlier emphatic prefix that was later lost:

*Cʌ-keps > *Cʌ-qeps > *qeps

'Ten' has cognates outside Sinitic:

Written Burmese kyip (used "when counting rational beings"), Limbu gip, Miju kap ~ kyep, Mikir kep, Maring chip, Yawdwin (So. Kukish) gyip (Matisoff 2003: 353)

Matisoff (2003: 353) reconstructed Proto-Tibeto-Burman *gip.

The Tangut word for 'ten' had a velar initial with a low vowel:

TT4489 TEN ɣạ 2.56

Although this could have come from an earlier *sV-kap (the tenseness of the vowel implies an earlier prefix, perhaps with initial*s-), I think it might be cognate to Qiang and rGyalrong forms with uvulars:

Mawo χɑ- (in words for tens)

Taoping χɑ- (low falling) (in words for tens)

but Ronghong has ho- ~ ha- ~ hɑ- with a glottal (LaPolla and Huang 2003: 383)!

Japhug (kɤmɲɯ): sqi

Japhug (gSar-rdzong): sqɪj

Zbu sɐʁɐʔ

from Proto-rGyalrong ?*sqi, with lowering in Zbu

(Somang ɕtɕé < Proto-rGyalrong * stɕ- is noncognate. It may be related to chay, the regular WB word for 'ten'. Matisoff [2003: 671] listed a second PTB root for 'ten', *ts(y)i(y) ~ *tsyay.)

I suspect that these all descend from an earlier *sV-qa, with the following changes:

- lenition of -q- to -ɣ- in Tangut, *s-prefix dropped after causing vowel tensing

- lenition of *sV-q- to χ- in Mawo and Taoping; further to h- in Ronghong

22:09: cf. my proposal of *s-q- > *χq- > *χ- in Old Chinese

- 'brightening' (raising) of the vowel in rGyalrong PST 'HEAD': VELAR OR UVULAR?

Guillaume Jacques proposed that Old Chinese 后 'ruler' is cognate to Written Tibetan mgo 'head'.

Since 后 has no nonemphatic members in its phonetic series, I reconstruct it as OC *go' [ɢoʔ] with an original emphatic initial.

I believe that OC *g was derived from

- a pre-OC (i.e., Proto-Sino-Tibetan?) uvular

- a pre OC (PST?) *g before non-high vowels

So it's possible that the ancestor of WT mgo and OC *go' was either *ɢo or *go.

(Matisoff [2003: 20, 225] regarded Tibeto-Burman uvulars as secondary and listed Benedict's Proto-Tibeto-Burman *m-gaw 'head' with a velar. This reconstruction does not appear in Matisoff's indexes.)

The Zhongu dialect of Tibetan may preserve earlier uvulars (Sun 2003: 783). Its word for 'head' is ngo-tʂə (Sun 2003: 814), implying an earlier velar. (I am assuming the word was not borrowed from another Tibetan dialect or subjected to a uvular to velar shift. It is not known whether all earlier uvulars were preserved in Zhongu or not; some or many may have become velars.)

However, Qiang varieties point toward a uvular initial:

Ronghong qə- (in compounds involving the head)

quətsu 'lift the head' (LaPolla and Huang 2003: 160-161) is not cognate, since qu is a root for verbs which may have nothing to do with heads

Mawo qə-, qa- (ditto)

Taoping (low falling tone)

There are other q-forms in 'head'-related words in these two dialects, but I am only citing the Ronghong morpheme identified as 'head' by LaPolla and Huang (2003: 357-358) and its Mawo and Taoping counterparts.

Elsewhere in Qiangic, Pumi qho55 (Jacques 2004: 153) has a uvular initial, but rGyalrong languages have velars (Jacques 2004: 233):

Japhug tɯ-ku

Zbu tə-kuʔ

(Somang tə-kó cannot resolve this issue, since Somang has lost its uvulars.)

I wondered if this ku could be from an earlier *qo (for the vowel, see Jacques [2004: 234]). In my current theory of emphasis in Old Chinese, *qu became *ku, so perhaps this change also occurred in Japhug: *qo > *qu > ku. This would predict that Japhug has no instances of qu, and I could not find any in Jacques (2004). However, Japhug does permit qhu - e.g., qhu 'later; after; back' - so it seems that the absence of qu is a random gap and not the result of a sound change. I would rather not claim that

*qo > ku


*qho > qhu (not khu)

for the sake of a single etymology.

I now wonder if there are actually two (unrelated?) lookalike words for 'head', one with a uvular initial and one with a velar initial. The velar version might be at the end of the Qiang words for 'forehead':

Ronghong zduxku

Mawo zdəʂku

Taoping da (low falling) χku (mid)

but I cannot identify the initial morphemes, whose homophones don't have plausible semantics:

Ronghong zdu 'deer'; 'male marker suffix'

Taoping da (low falling) 'directional / past tense prefix'

The initial of

TT3619 HEAD ɣu 1.4 (transcribed in Tibetan as dgu, bgu)

could be from a prefixed velar or uvular.

Proto-Lolo-Burmese 'uh (Matisoff 2003: 477) could be from an earlier *quh. Cf. the correspondence of Written Burmese 'ap 'needle' to uvular-initial forms. See here.

Matisoff (2003: 477) regarded PLB 'uh as cognate to WT dbu, but the correspondence ' : b is bizarre. I think PLB may be related to WT mgo. WT dbu could also be related to mgo if its b < *gw, but there is no other evidence for a labiovelar or labiouvular initial in this word.

Matisoff (2003: 198) compared WT dbu to Old Chinese 頭, but 頭 was*N-to without any labial at all and is too new to be of Proto-Sino-Tibetan origin. See Sagart (1999: 156) for its true etymology. DWELLING IN METAL

Chinese 鋸 'saw' is written as 金 'metal' (semantic) + 居 'dwell' (phonetic).

居 'dwell' in turn looks has 古 'old' as phonetic.  (Is 尸 'corpse' a distortion of something else?)

古 'old' was something like *[qɑʔ] in Late Old Chinese. I can't think of any reason to reconstruct a cluster, prefix, or presyllable in this word, though a more complex earlier form can't be ruled out.

古 'old' is phonetic in 苦 Late Old Chinese *[qhɑʔ] 'bitter'. 'Bitter' has uvular initials in Qiang:

Mawo qhɑ

Ronghong qhɑq

Taoping qhɑ

It also has a uvular initial in some Tibetan dialects (Sun 2003: 783):

Zhongu qhɐ

Shibazi qhaa55

Yanyun qhɐ-nde

Ren'entang qhʌ

Although Written Tibetan has no uvulars, it's possible that these dialects preserved uvulars that shifted to velars elsewhere.

(Japhug rGyalrong kɯ-qiaβ 'bitter' is not cognate to the above words. The initial lacks aspiration and -i- and < *-p correspond to nothing elsewhere.)

Hence I suspect that the Proto-Sino-Tibetan word for 'bitter' was something like*qha, preserved more or less intact in Early Old Chinese as 苦 *qhaʔ. 'Old' was probably nearly homophonous apart from the absence of aspiration: 古*qaʔ. The EOC vowel I write as *a may have already been backed after uvulars (*[ɑ]), though its backness did not become phonemic until Late Old Chinese which distinguished back and non-back low vowels.

If 古 was uvular, then 鋸 'saw' must have also been uvular. However, in LOC, 鋸 'saw' was *[kɯɑh] with a velar initial. I suspect that this velar initial originated from a presyllable with a high vowel preceding an original uvular:

EOC*Cɯ-qɑs > *Cɯ-kɯɑh > LOC *kɯɑh

These northern Qiang forms may confirm an original uvular in 'saw':

Mawo səqqɑr

Ronghong səqa

Could these be borrowed from an Old Chinese *sɯqɑ (or just *qɑ, if the first part is of Qiang origin)?

I cannot explain the retroflexion of the Mawo vowel, unless it is a trace of an earlier Qiang morpheme with *r, or added by analogy with a semantically related word (cf. Ronghong khər 'to saw' - is the Mawo verb for sawing similar?).

The Taoping word for 'saw' is kie (mid tone) ʐi (high tone).

kie could be a borrowing from a Middle Chinese 鋸 *küö.

-ʐi superficially resembles -zi [tsɿ], the noun suffix in Mandarin 鋸子 juzi 'saw'. But -ʐi is retroflex, whereas -子

- is alveolar

- was borrowed into Taoping as -tsɿ

Moreover, unlike -tsɿ, -ʐi appears in native words:

ma (low falling) ʐi (high) 'rain'

cf. Mawo məʁer, Ronghong meer

χta (low falling) ʐi (high) 'axe'

cf. Mawo staar, Ronghong stei

kie (low falling) ʐi (high) 'near'

cf. Mawo guərn, Ronghong gue-ɲi

the first part coincidentally resembles 近 EOC *gərʔ 'near', but is probably 'road' (Mawo gurβa, Ronghong gueer) and is also in Mawo guərχe 'far' and Ronghong guaha 'far'

the second part is probably cognate to 邇 EOC*neyʔ 'near'

In all four cases, Taoping -ʐi corresponds to retroflexion in Mawo, and in two cases corresponds to -i in Ronghong. This implies that ʐi came from an earlier *r(n)i: cf. this correspondence:

'rich': Taoping ʐi : Mawo ɹi

A root-final *-r may have moved over to a following syllable:

*mVr-i > maʐi 'rain' (no trace of *-i in Mawo, Ronghong)

*staar-i > χtaʐi 'axe' (no trace of *-i in Mawo)

*gVr-ni > kieʐi 'near' (no trace of *-i in Mawo)

There is no evidence for a final *-r in Chn 鋸 'axe'. Therefore Taoping kieʐi (< ?*kVr-i) 'saw' and Ronghong khər 'to saw' could be from a native root ending in *-r. QIANG χ < *CV-Q-?

In Mawo Qiang, aspirated and voiced grave stop initials of roots become fricatives when preceded by a vowel in compounds and inflected forms (Sun 1981: 43-44, 54-55). The rhymes -u, -ər, ur also devoice and the achromatic vowels ə and ɤ delete in past tense forms (Sun 1981: 54-55):

ʁdʐə 'star' + phu 'flee' = ʁdʐəɸu 'meteor'

ma- 'not' + bu 'high' = maβu 'not high'

'rain' + khər 'fall (of frost)' = xə̥r 'frost'

the analysis is based on LaPolla and Huang's analysis of Ronghong Qiang merx 'frost (n.)'

the full word for 'rain' is məʁer

cf. Ronghong Qiang meer 'rain', TT1273 SKY r 1.27

TT4279 RAIN wer 2.71 is probably from *r-w- (cf. Proto-Tibeto-Burman *r-wa 'rain' [Matisoff 2003: 618]) rather than *C-m-

da- + khər 'to saw' = daxə̥r 'sawed'

tɕi 'house' + gər '' = tɕiɣə̥r 'rent'

da- + g 'wear clothing' = daɣ 'wore clothing'

dɑ- + qh 'to cover' = dɑχ 'covered'

Affricates lose their stop element in past forms (Sun 1981: 55):

dɑ- + tshɤ '' = s 'select and appoint'

da- + dzə 'to eat' = daz 'ate'

reminiscent of*tsh ~ *s alternations within phonetic series in Middle Chinese; MC *dz ~ *z alternations are less common

Note, however, that weakens to r in compounds and phrases:

The Ronghong equivalent of this r < is a "somewhat retroflexed" l (close to [ɭ]?; LaPolla and Huang 2003: 31, 423).

weakens to the expected ʐ in reduplicated forms (see below):

Fricatives and fricative-initial clusters don't change in that position:

ɲu 'sheep' + hu 'hair' = ɲuh 'hair of a sheep'

dɑ- + stɤ 'to bubble' = st 'bubbled'

da- + xtʂə 'to block' = daxtʂ 'blocked'

If a cluster begins with an aspirated or voiced stop, that stop is lenited (Sun 1981: 46-48):

'earth' + khsə 'god' = xs 'earth god'

21:54: the full word for 'god' is khsi; does khsə with a schwa appear in initial position in other compounds?

guar 'army' + gzə 'senior officer' = guarɣz 'army officer'

rmi 'fertilizer' + qhsɑ 'mud' = rmiχsɑ 'diluted fertilizer'

Lenition also occurs in rapid speech (Sun 1981: 44, 48):

'sweep': dʐɤɑ (slow speech) ~ dʐɤʐɑ (rapid speech)

'roll': luqhur (slow speech) ~ luχur (rapid speech)

'to fool': qhsəqhsa (slow speech) ~ qhsəχsa (rapid speech)

In Ronghong, lenited preinitials are even found in word-initial position (LaPolla and Huang 2003: 32):

'god': Ronghong xsi : Mawo khsi

(but 'earth god' above seems to be derived from khsə)

'govt. official': Ronghong ɣzə : Mawo gzə

'manure': Ronghong χʂə : Mawo qhʂə

Unlike Mawo, Ronghong no longer allows stop-fricative clusters like ks, gz, qhʂ. (Ronghong ts, tsh, dz, etc. are affricates and not clusters.)

Is it possible that earlier Qiang already had a similar process of lenition that changed uvular stops into uvular fricatives? LaPolla and Huang (2003: 333, 345) list two Ronghong χ-words which they derive from Proto-Tibeto-Burman forms with *k-

'needle': *CV-q(h)- > Mawo χe, Ronghong χa, Taoping χe (high tone)

Proto-Tibeto-Burman *kap

cf. Written Tibetan khab < *kap, Old Chinese *kim (< ?*kyəm < ?*Ci-qəp-NV)

Could the glottal stop of Written Burmese 'ap be from *q-? (Matisoff 2003: 300 notes that WB "zero initial" ('-) can correspond to "velars elsewhere".)

Could Proto-Lolo-Burmese *r- in *rap (Matisoff 2003: 517) be from an earlier *ʁ-, a lenited form of an even earlier *q-?

The Zbu rGyalrong word for 'needle' is tɐ-ʁâv (Jacques 2004: 241) with a voiced uvular corresponding to the voiceless uvular of Japhug ta-qaβ (but Somang has a velar in ta-káp, because Proto-rGyalrong *q- and *k- merged as k- in eastern rGyalrong [Jacques 2004: 305, 309] - 22:48)

The initial of Lahu ɣòʔ 'needle' corresponding to my proposed *q- reminds me of the Iranian pronunciation of Arabic q as [ɣ].

'bronze': *CV-q(h)- > Mawo χɑɸi, Ronghong χɑ, Taoping χa

Proto-Tibeto-Burman *kar

cf. Written Tibetan Hkhar-ba ~ mkhar-ba 'bronze', Written Burmese kaiH 'lead' (instead of the expected kanH)

Mawo -ɸi looks like a lenited form of phi 'white', but bronze isn't white!

Why doesn't the Taoping vowel match the other two?

Ronghong does have synchronic uvular q ~ velar x (not uvular χ!) variation, but the only example given in LaPolla and Huang (2003: 37) is in final position in 'black': ɲiq ~ ɲix (cf. Mawo ɲiq; Taoping ɲi [high] ɲi [low falling] has lost the coda)

LaPolla and Huang (2003: 375) believed that the coda of this form "is secondary". Therefore it should not correspond to the codas of WT nag or WB nak.

Japhug rGyalrong kɯ-ɲaʁ and Zbu rGyalrong kə-ɲɐ̂χ have uvular codas like Ronghong and Mawo. (Jacques 2003: 262 does not list a Somang or Cogtse form.) Jacques (2004: 265) reconstructed Proto-rGyalrong *-aq for this Japhug : Zbu correspondence. PR *-ak would have resulted in Japhug -a and possibly Zbu (Jacques 2004: 212).

Ronghong and Mawo -iq looks like a 'brightened' counterpart of this PR*-aq. Could these Qiang and PR uvular codas preserve a Proto-Sino-Tibetan uvular coda? QIANG χ < *C-Q-?

The cognates of TT2597 NEEDLE Ga 1.17 in three varieties of Qiang (two northern and one southern) have a voiceless uvular fricative initial χ- corresponding to the voiced velar fricative initial of Tangut and the voiceless uvular stop q- of gDong-brgyad rGyalrong ta qaβ:

Mawo χe

Ronghong χa

Taoping χe (high tone)

(Sun 1981: 204; LaPolla and Huang 2003: 345)

Since Mawo, Ronghong, and Taoping all have an initial q-, this χ- probably is not from an earlier simple *q-.

I suspect that Qiang, like Tangut, developed fricatives from earlier clusters (which in this case consisted of prefixes before a root initial *q-):

*C-q- > Tangut G-, Qiang χ-

Tangut G- might have been uvular [ʁ]. This would explain why Tangut G- was classified as a laryngeal rather than as a velar in Homophones.

11.22.00:55: But it's more likely that the Tangut were just copying the Chinese phonological tradition which classified back fricatives (velar and/or glottal [and even uvular?]) as laryngeals.

11.22.1:06: Sofronov reconstructed a glottal stop initial for TT2597 NEEDLE: 'a 1.17. The initial fanqie speller for NEEDLE is TT4725 NORTHERN-BARBARIANS (Gong: Gu 1.4, Sofronov 'u 1.4). If this is a borrowing from Middle Chinese 胡 *Gu 'barbarian', then both it and NEEDLE had a fricative onset, and I cannot use NEEDLE as evidence for *q- becoming a glottal stop in Tangut.

11.22.2:55: Mawo and Taoping -e is a partly 'brightened' *-a. The environment *q-p is hardly where I would expect a low vowel to rise and front.

I do not understand why some *a remain low, others raise to mid, and still others raise to high in Qiangic languages. (See Matisoff 2004 for examples.) I would rather not reconstruct several types of *a in Proto-Qiangic.

Maybe the variation in raising reflects variation in prefix vowels. In Old Chinese, some *a became opening diphthongs and others didn't, and I partly account for this with different prefix vowel heights:

High prefix vowel: Early Old Chinese *Cɯ-Ca > Late Old Chinese *Cïa

No prefix or low prefix vowel: Early Old Chinese *(Cʌ-)Ca > Late Old Chinese *Ca

Although Mawo and Taoping happen to have raised *a to the same degree in 'needle', there are several types of height mismatch which might have involved 'brightening':

Pattern 1a: Mawo a : Taoping e

'eight': Mawo khar, Taoping Che (mid tone) (cf. Ronghong kher which looks like a blend of the two)

'low': Mawo ba, Taoping be (mid) (no Ronghong cognate)

Pattern 1b: Mawo a : Taoping ie

'tail': Mawo suka, Taoping sua (high) kie (mid) (cf. Ronghong sikue)

'neck': Mawo tʃəka, Taoping Shɿ (low falling) kie (mid) (cf. Ronghong Chuəke ~ Chuki)

do both these words have a *-kV suffix?

'easy': Mawo za, Taoping zie (low falling) (cf. Ronghong ze)

'point, tip': Mawo tsa, Taoping qə (high) tsie (mid) (can't find a Ronghong cognate)

Pattern 1c: Mawo ɑ : Taoping ie

'root': Mawo kɑpɑ, Taoping kie (mid) (cf. Ronghong kopu)

Pattern 1d: Mawo (u)ɑ : Taoping ue

'city': Mawo ʁuɑɑr, Taoping ʁue (low rising-falling) (maybe not from *a-?; cf. Ronghong ʁu ~ ʁo and cases where *-a- seems to have become Tangut -u)

'late': Mawo ʁlɑ, Taoping ʁue (mid) (cf. Ronghong ʁlɑ)

Pattern 2: Mawo i : Taoping a

'wine': Mawo shi, Taoping chha (mid) (cf. Ronghong shi)

maybe these words aren't cognate

'brain': Mawo qañi, Taoping (low falling) χña (mid)

Pattern 3: Mawo u : Taoping ɑ

'gold' and 'difficult' (homophones): Mawo Shqu, Taoping χqɑ (mid) (cf. Ronghong χqu ~ Shqu)

'waist' thɑsu, Taoping χtɑ (mid) (mid) (is Ronghong Shəq cognate?)

This is not a complete list. There may be other patterns. TANGUT LH < *KL-?

If Tangut G- corresponds to gDong-brgyad rGyalrong velars and uvulars, what does GBR G- correspond to in Tangut?

I could not find any instances of a simple GBR root-inital G- in Jacques (2003).

In general, GBR G comes from Proto-rGyalrong *w (Jacques 2004: 331-332). So it's not surprising that it corresponds to Tangut w- and -w-:

'tooth': GBR tɯ shGa : TT4679 shywi 1.10

'rust': GBR sGa : TT2585 wẹ 1.65

Here, it corresponds to Tangut p-:

'white': ku Grum (< PRG *pr-) : TT5461 phiow 1.55

In seven cases, GBR -G- corresponds to zero in Tangut:

'sickle': GBR tɯ ñcGa : TT2617 jyị 1.67

cf. Somang tə-ntuâ and Zbu tə-ñcGwe', implying a PRG *-w-

although Tangut had the syllable jywi, it lacked jywị; did pre-Tangut *sjw- simplify to j- rather than jw-, or is this just a case of irregular medial loss in Tangut? (The prefix *s- is the source of the tenseness of the vowel written as a subscript dot.)

'cold': kɯ GnJo (is this G- < *k-?) : TT5497 jyiy 1.36

'uncle': GBR ɯ βGo (< PRG *kp-) : TT3336 po 1.49

'bird': GBR pGa (not from *pw-) : TT2483 we 1.8

'sun': ʁmbGi (not from *-mbw-) : TT5728 be 2.7

'deaf': tɤ mbGo : TT5037 ba 1.17

'flee': kɤ phGo : TT5730 bọ 2.62

'ice': tɤ jpGom : TT0600 'wọ 1.70

GBR -G- in 'deaf', 'flee', and 'ice' (and probably 'bird' and 'sun' as well) reflects a PRG velarized vowel (Sun in Jacques 2004: 221; 275). Jacques (2004: 331) did not reconstruct any PRG *labial-w- sequences.

GBR G- plus a consonant sometimes corresponds to nothing in Tangut:

GBR tɯ Gmas 'wound' : TT2293 myaa 1.23 'scar'

'manure': GBR tɯ Gli : TT1313 lyï 1.29

GBR yɯ Gmɯr 'tonight' : TT2247 myïï 1.32 'night'

'mill': GBR kɤ Gnjɯr (< PRG *knj-) : TT4933 jwï r 1.86

Tangut -w- should correspond to a PRG medial -w-, though I cannot find any examples of PRG (n)jw- in Jacques 2004

However, GBR Gɤ l-, Gɤ y-, and, in one case, Gl- correspond to Tangut lh:

'humid': GBR kɤ Gɤ la : TT4960 lhyi 2.10

'manure': GBR tɯ Gli : TT1313 lhyị 2.60

GBR kɤ Gɤ yɯ (< *ly-) 'add' : TT2712 lhu 1.1, TT5373 lhu (tone unknown), TT4643 lhyu 1.13, TT3814 lhyï 1.30, all 'increase'

GBR Gl- comes from PRG *kl- (Jacques 2004: 332). Could pre-Tangut *kl- have become Tangut *lh-?

I doubt that GBR reflects an ancient presyllable, so it seems that the lh- of the Tangut words for 'humid' and 'increase' reflects a prefix absent in GBR.

If the Tibetan transcriptions kli, gli, gliH for TT1075 ONE reflected a Tangut dialect preserving initial clusters, one might expect ONE to have an initial lh- in the standard dialect. Yet ONE had a simple l-: lew 1.43. This casts doubt on the *kl-hypothesis.

11.21.1:41: Jacques (2003: 22) compared GBR zGom bu 'broom' to TT0878 BROOM lhyow 1.56, but this seems unlikely to me because zG is from PRG *sb- (Jacques 2004: 329). *MYI-KUM

(a combination of myi 2.10 'Tangut' and Proto-Tibeto-Burman *m-kum 'block/pillow')

The Tangut words for 'pillow'* and 'needle' had the same initial:

TangraphTangut TelecodeGlossGong's reconstructiongDong-brgyad rGyalrong cognate (Jacques 2003)
1920PILLOWGyïy 1.42tɤ mkɯm 'pillow'
4821PILLOW (from Jacques [2003: 19]; but Grinstead [1972: 114] defined this as SOURCE, as did Nevsky [1960 II: 282]: '源; исток, источник')Gyow 1.56
2597NEEDLEGa 1.17ta qaβ 'needle'

Tangut *G- corresponds to the following gDong-brgyad rGyalrong onsets (Jacques 2003):

unaspiratedaspiratedwith preinitial
Velarsk-kh-fk- (< Proto-rGyalrong *pə-k), mk-

I suspect that Tangut *G- is a 'softened' version of pre-Tangut *k(h)- and *q-:

*Ck(h)-, *Cq- > *G-

just as Vietnamese g is a 'softened' version of earlier Vietnamese *k-: gà 'chicken' < *Cka (cf. Arem lakææ, Ruc rəkạa1 [source: Ilya Peiros' database]).

The pre-Tangut word for 'needle' may have been *t(V)-qap if it had the same dental prefix as gDong-brgyad rGyalrong (and there's no guarantee that it did).

Next: Why doesn't gDong-brgyad rGyalrong G correspond to Tangut G?

*11.20.2:05: I am ignoring

TT4477 PILLOW wo 1.70

which Jacques (2003) did not list as a cognate of gDong-brgyad tɤ mkɯm 'pillow'.

According to the Tangraphic Sea, the right side of TT4477 is the source of the right side of

TT1920 PILLOW Gyïy 1.42

despite having four instead of three strokes over the bottom left hook 乚.

It seems that the Mojikyo glyph for TT4477 has one extra stroke over the hook. The element with four strokes over the hook is associated with evil, whereas the element with only three strokes presumably has a less negative connotation. Nevsky (1960) did not list TT4477, but he handwrote TT4821 with three strokes over the hook (II: 262). Sofronov (1968 II: 376, 383) listed both TT4477 and TT4821 with three strokes over the hook. STUMBLING BLOCK

In Matisoff (2003: 308), Proto-Tibeto-Burman *m-kum ~ *m-kim 'block/pillow' was compared to

656g 枕 Middle Chinese *chim' 'pillow', *chimh 'use as a pillow'

as well as

658f 椹 Late Old Chinese *Tïm 'chopping block'

枕 MC *chim'/h could come from Early Old Chinese *kim-'(-s), *t(l)im-'(-s), or *t(l)əm-'(-s). *kim obviously matches PTB *kim. Moreover, 658 could have been a velar series.

The trouble is that 656 probably wasn't a velar series. It had no velars in Middle Chinese:

656a 冘

MC *yim < EOC *(tə-)ləm < ?*-lum

MC *yu < EOC *lu (Jiyun; open syllable!)

11.19.00:30: is this simply due to confusion with 尤 MC *yu?

or did this reflect a lost final labial: e.g., Pulleyblank's (1962: 235) *-v (or *-b or *-gw, which have known voiceless counterparts *-p and *-kw unlike *-v which lacks a counterpart *-f)

Jiyun also lists the alternate spelling 彳+占 with the dental phonetic 占 implying a lost dental-initial presyllable. This alternate spelling had three other pronunciations. The last two were strange amalgams of MC *y- < EOC nonemphatic *l with rhymes from EOC emphatic syllables:

MC *chiemh < EOC *t-lam-s

MC *yEm < EOC ?*r-tʌ-ləm

MC *yam < EOC ?*-lam

656b-d 沈 (is this really absent from Jiyun?)

MC *Dim(h) < EOC *rləm(-s)

MC *shim' < EOC *hləm'

MC *chim' < EOC *tləm

656e 酖 MC *Dim < EOC *rləm

656f 鴆 MC *Dim < EOC *rləm

656h 忱 MC *jim < EOC *Ntləm

656i 抌 MC *təm' < EOC *tləm'

656j 眈 MC *təm(-') < EOC *tləm(-')

656k 紞 MC *tam' < EOC *tlam'

656l 耽 MC *təm < EOC *tləm

656m 髧 MC *dəm' < EOC *ləm'

656n 黕 MC *təm' < EOC *tləm'

656o 醓 MC *thəm' < EOC *hləm'

Although it would be possible to derive all of the above MC initials from EOC velar presyllables and clusters with velars, it's more likely that 656 was a lateral series. 656a has the Middle Chinese initial *y- which came from Old Chinese *l-.

11.19.1:28: Even if one reconstructed 656g 枕 with a velar initial as EOC *kləm, the medial *-l- would correspond to nothing in PTB *m-kum ~ *m-kim. The syllable *klum was possible in PTB (Matisoff 2003: 596), so one could not claim that PTB lost medial *-l- between *k and *-um. Could the root of PTB *m-kum ~ *m-kim reflect a loan from an Old Chinese form like *kəm which had lost medial *l-?*

(The PTB prefix *m- may be an innovation absent in Chinese. If there ever was a nasal prefix in 656g 枕, it was a loosely attached *mə- that left no trace in Middle Chinese. A *m- might have voiced the initial: EOC *m-kim > MC *jim.)

11.19.1:09: It's also remotely possible that 656 was a 'yodic' series with a phonotype *yəm. Most possible EOC sources of the MC initials are 656 are listed below. (I have excluded dental/velar clusters: e.g., the rather unlikely *rNtkly-, which would have become MC *D-.)


MC *y- < EOC *(tə-)(l)y-

MC *sh- < EOC *h(l)y-

MC *ch- < EOC *t(l)y-, *ky-

MC *j- < EOC *d(l)y-, *Nt(h)(l)y-, *gy-, *Nk(h)y-

EOC *kly- and *gly- would not have palatalized to MC *ch- and *j-


MC *D- < EOC *rly-, *rd(l)y-, *rNt(h)(l)y-


MC *t- < EOC *t(l)y-

MC *th- < EOC *(t)h(l)y-

MC *d- < EOC *(l)y-, *d(l)y-, *Nt(l)y-

I have been assuming that EOC *y and *l may have had similar MC reflexes, but this may be incorrect. Perhaps EOC *y- became MC *G-, unlike EOC *l- which became MC *d-. Emphatic yod needs much more investigation.

According to Youssef (2006: 13), Cairene Arabic has a nonphonemic emphatic [jʕ] which would be equivalent to my EOC *y. All Cairene Arabic consonants other than [q] have emphatic counterparts. CA [q] behaves like an emphatic consonant.

Could PTB *m-kum ~ *m-kim and 656g *kyəm-'(-s) descend from a Proto-Sino-Tibetan ?*kyum?

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2007 Amritavision