There is only one (Grade IV) R24 -iaa tangraph in chapter II of Homophones:

TT3018 R24 1.23

I would have expected a chapter II (labiodental) initial before Grade III R21-ɨaa since chapter II initials are associated with Grade III rather than Grade IV rhymes. Why is TT3018 an exception? The archetypal chapter II initial is v-, so I initially hypothesized that it was a tangraph viaa representing Sanskrit vyaa, a syllable that violates a native Tangut phonological constraint against combining v- with -i-/-y-. But in fact TT3018 is ?xwiaa R24 1.23 'mouse; rat' and is obviously cognate to the Grade III word

TT3016 ?xwɨi R10 1.10 'mouse; rat'

I am not sure how to reconstruct the initials of those two words. Both are listed in chapter II of Homophones, yet xw- (which is a carryover from Sofronov and Gong's reconstructions) is obviously not labiodental like v-. I wonder if their initial was a voiceless counterpart of w- or v-: e.g., wh- [ʍ] or f-.

Their fanqie initial spellers belong to Sofronov's (1968 II: 94) glottal (!) fanqie chain 9 which contains a mixture of Homophones Chapter II (labiodental) and VIII (glottal) tangraphs. I will examine this chain next time.

I am not even sure if the chain is relevant to TT3018. Sofronov (1968 II: 129) lists the fanqie of TT3018 as

TT3016 ?xwɨi R10 1.10 + TT5101 miaa R24 1.23

but I have no idea where it is from. The only source of R24 fanqie that I know of is Tangraphic Sea which does not list TT3018. According to Han Xiaomang (2004: 300), TT3018 occurs in four lexicographic sources:

Homophones B 12B62

合编_甲 (Homophones and Tangraphic Sea) Combined A 05.073

同义_甲 Synonyms A 0816/08

切韵_乙 Cut Rhymes B 02B2

I have not seen the last three sources. Is the fanqie for TT3018 from one or more of them?

I have regarded R21 as Grade III and R24 as Grade IV, but am not entirely certain about these assignments because of the distribution of initials before each rhyme.

The archetypal distribution of initials before Grade III and IV can be found before Grade III R10 and Grade IV R11 (Gong, "Phonological Reconstruction of Tangut"):

(Although Homophones chapter IX initials ʒ- and z- are not written as liquids, I suspect they were phonetically similar to r and l: e.g., they might have been [ʐ] and [ɮ]. lh- may have been [ɬ], the voiceless counterpart of [ɮ].)

Here is the distribution of initials before R21 and R24. Exceptions to the expected pattern are in red.

Neither rhyme patterns quite like R10 or R11.

I thought R21 should be classified as Grade III because it has most of the alveopalatals. But dentals, alveolars, and glottals are rare in Grade III yet are more common before Grade III R21 than Grade IV R24.

Next: Examining the exceptions to the pattern.

09.1.8.23:39: DGA > GIAA?

TT3777 giaa R24 2.21 'to rejoice; to enjoy'

Are there any other Tibetan loanwords (as opposed to calques) in Tangut?

Guillaume's etymology appealed to me because it made me wonder if acute preinitials like d- could condition vowel breaking in Tangut:

*gaa > gaa

but

*dgaa > *dgiaa > *giaa (Grade IV)

I reconstructed -i- after g- because it is followed by a Grade IV rhyme. Until now I have mechanically carried over Gong's arrangement of rhymes in Li Fanwen (1997: 2) which treats

R21 -ɨaa as the long vowel counterpart of (my Grade III) R19 -ɨa

R24 -iaa as the long vowel counterpart of (my Grade IV) R20 -ia

(Gong does not distinguish between Grade III and IV and reconstructs a single medial -j- for all four rhymes.)

However, I am now wondering if the grades for R21 and R24 sholuld be reversed on the basis of their initials: e.g., dentals which do not normally precede Grade III rhymes appear before R21. I'll look at the distribution of initials before R21 and R24 in more detail next time.

*Although this is spelled as dgH in Tibetan, the final -H is probably a silent disambiguator needed to differentiate dgH [dga] 'be glad; take pleasure in' from dg [dag] 'pure'. Final -H in other cases may actually have represented a real consonant (Hill 2005).

I was tempted to associate the final -H of dgH with the final consonants some have reconstructed for R24

Sofronov: -i̭aC

Starostin: -i̭aʔ

Arakawa: -a:'

but without further evidence, I will continue to regard -H as a mater lectionis.

09.1.7.23:33: WHO NOSE WHAT'S GOING ON

with these sinographs?

自 GSR1237m MC *dzih 'self' (a drawing of a nose)

鼻 GSR 521c MC *bih, *bit 'nose' (the latter is implied by Amoy pit, Nanchang Gan phit, and the tone of Mandarin bí - but Amoy and Chaozhou phĩ may imply a earlier nasal!)

洎 GSR1237a-b MC *kɨih, *gɨih 'pour out; broth of boiled meat'; *gɨih 'together with' (same as 暨)

John Bentley drew my attention to them because he noticed that 洎 represented the Old Japanese syllable si in Nihon shoki, even though it should represent OJ ki. My guess is that si was based on the erroneous assumption* that 洎 was (nearly) homophonous with 自.

According to Shuowen, 自 is phonetic in 洎. Boodberg ("Proleptical Remarks") and Pulleyblank (1962: 135) both thought 自 was phonetic in 鼻. But as the GSR numbers above indicate, Karlgren did not regard these graphs as belonging to a single phonetic series. (GSR1237 is not a xiesheng series; it is a category for MC *-i graphs whose OC readings Karlgren couldn't confidently reconstruct: e.g., GSR1237c is 彝 MC *ji which obviously doesn't contain 自.)

The simplest OC reconstructions for the above three graphs without any regard for potential xiesheng relationships would be

自 MC *dzih < *dzi(t)s

鼻 MC *bih < OC *bit-s, MC *bit < OC *bit 'nose' (but how can Amoy and Chaozhou phĩ be accounted for? - do they imply OC *bins?)

洎 MC *kɨih < OC *krits and *gɨih < *N-krits; cf. 暨 *grits < ?*N-krit-s; both < *-ps?

Reconciliations

1. Boodberg reconstructed OC *BDZi for both 自 and 洎 and linked them to 面 *mens 'face' (< his *BDZän), 前 *dzen 'fore-part' (< his *BDZ-; no rhyme given), and 息 'breath' (< his *BSEK; Shuowen says自 is a 亦聲 'also phonetic' = 'near-phonetic').

2. Pulleyblank (1962: 135) reconstructed *sb(δ)- (equivalent to *sbl- in my system) for 自. He also included

辠 MC *dzwəjʔ (= 罪 with 非 MC *puj)

氵+辠 MC *tshwəjʔ (氵+罪 = with 非 MC *puj)

臬 MC *ŋet and *kwɨejh (latter in Jiyun)

習 MC *zip

as derivatives of 自, though as far as I know, the early forms of the last two did not include 自. I know of no variant of 習 with 自. Shuowen states that 自 is phonetic in 臬 but not in 習 whose phonetic is listed as 白 MC *bæk (!).

I think 臬 and 習 are unrelated to 自 but am not sure how to deal with the rest of the series. Most members might fit an *(s)P-archetype:

自 MC *dzih < *dzis < *s-bit-s (Pulleyblank compared this to Written Tibetan sbrid-pa 'sneeze' - but there is no trace of an *r in Chinese!)

鼻 MC *bih < OC *bit-s, MC *bit < OC *bit

辠 MC *dzwəjʔ < OC *sbəjʔ

or even OC *N-s-pəjʔ ('crime'; cognate to 非 MC *puj < OC *pəj 'not')

氵+辠 MC *tshwəjʔ < OC *sphəjʔ

The rhymes of the last two don't match the first two, though one could regard *-əj as an *ə-grade expansion of zero-grade *-i (= *-j). But I don't know of any other alternation between *-ʔ and *-t.

Although the rhyme of 洎 belongs to the same OC *-its category as 自 and 鼻 (unless the *-ts of 洎 is from *-ps),, its initial *(N-)kr- has no labial. I have considered reconstructing this series as labiovelar: e.g., 鼻 *b- < *gw-. However, there is no evidence for such a shift in other phonetic series and there is no trace of labiality in 自 or 洎. (1.9.22:18: There is also no trace of rhotacism in 自, 鼻, 辠, or 氵+辠.) So I have no idea what to do with 洎.

*Recently, Guillaume Jacques pointed out to me that the Tangut translation of Leilin contained a word for 'Japanese person'

vɨi R10 2.9 ʒɨĩ R16 1.16

This is obviously a transcription of Chn 倭人 'dwarf (= Japanese) person' implying a Tangut period NW Chinese original like *wi ʒĩ. The trouble is that 倭 'dwarf' should be something like *wɑ or *wɔ in TPNWC, not *wi. Here are three explanations ranging from the least to the most probable:

1. Variant or local reading

倭 'dwarf' may have had an emphatic reading *ʔoj in Old Chinese. ('Dwarf' is not attested in early OC texts according to GSR.) The emphasis pushed down the vowel in Late Old Chinese *ʔwɑj (source of Sino-Korean 왜 wE < *waj), though by the post-Middle Chinese period, the vowel rose to *ɔ.

倭 also has a nonemphatic OC reading meaning 'long and winding'

OC*ʔoj > *ʔwaj > *ʔwɨaj > *ʔwɨej > *ʔwɨe > *ʔwɨj > *wɨj (cf. Tangut vɨi)

homophonous with one reading of its phonetic 委. Perhaps 'dwarf' was homophonous with 'long and winding' in TPNWC. However, it is dangerous to reconstruct a local reading on the basis of one transcription.

2. TPNWC speaker error

In TPNWC, 倭 'dwarf' was misread as 倭 'long and winding' or 委, and this misreading was taught to the Tangut. Is 倭 'dwarf' rare enough to be misread?

3. Tangut speaker error

The Tangut translator of Leilin wrongly assumed that 倭 'dwarf' was read like 倭 'long and winding' or its phonetic 委 and transcribed it accordingly as

vɨi R10 2.9

just as some Nihon shoki scribe wrongly assumed that 洎 was read like its phonetic 自 and used it as a phonogram for OJ si.

09.1.6.23:50: PULLEYBLANK'S IN-*FL-UENCE

In "A plaao-sible Derivation?", I referred to Pulleyblank's 1962 Asia Major article which I didn't have on hand at the time. I often don't have what I need since my library is split between 夏威夷 and 新澤西. When I came back to 新澤西 this morning, I found my copy of Pulleyblank 1962 and saw the passage I was thinking of on p. 121:

[丑] is the second earthly branch and we have Tai loans for it. They seem to imply Primitive Tai *pl- (Li Fang-kuei 1945, p. 338). There is no trace of labials in the [xiesheng] series [for 丑] and we should like to reconstruct *nhl > *θl. Tai *pl- might perhaps represent *fl, as a substitution for *θl.

Then I checked my email and found that Sven Osterkamp had sent me links to both parts of Pulleyblank's article in an online Asia Major archive. Now I'll always have access to that article and many, many others. Thanks, Sven!

Pulleyblank's 1962 *θl- would be mechanically equivalent to *hlr- in my reconstruction. Questions:

1. Is there any language in which nh- became θ-?

2. Would Pulleyblank's Primitive (= Proto-)Tai *fl be a sui generis cluster created by PT speakers to handle Old Chinese *θl-, or can *fl-clusters be reconstructed for native words in PT? Li Fang-keui's PT reconstruction has no *fl-, *fr- and *vl- have only one example each, and *vr- is the most frequent with four examples (LFK 1977: 94).

Here's a fourth solution (the first three are here) leading to the third and fourth questions: 丑 was orignally *r-hnuʔ sharing the *nu-core of its phonetic series. *r-hn- shifted to *hlr- which then hardened to *thr- and then fused into *ʈh-.

3. I would guess that hnr- is a less marked cluster than hlr-, though both are unusual. Why would a less marked cluster become a more marked cluster?

4. OC nonemphatics generally do not harden: e.g., OC *hl- became a Middle Chinese fricative *ɕ-. I would imagine that *hlr- would become a *ɕr- that fused into ʂ-, but this solution requires fortition before fusion: *thr- > *ʈh- (rejected path here*). Such fortition is expected for OC emphatics: cf. 畼 'kill sprouting grass' and 暢 'penetrate everywhere':

OC *r-hlaŋs > *thraŋh > Middle Chinese ʈhaŋh

Is there any reason why *hlr- and *hl- developed in different directions?

*I regard

*hlr- > *hɖr- >*hʈ- > *ʈh-

with a preaspirated retroflex stop phase as improbable. OC *hl is a voiceless [l̥], not [h] + [l], so it would probably not develop into *hɖ, a consonant that may not exist in any language. It's not in UPSID.

09.1.5.17:25: A KAP-ITAL CONUNDRUM

Since the Wa borrowed their names for the twelve branches from Chinese via Tai, I would expect their names for the ten stems to also be Chinese borrowings. But only the first stem is borrowed. Wa kap resembles southern Chinese forms like Cantonese 甲 kaap, Meixian Hakka kap, etc. It must have been borrowed into Tai after the reduction of OC *Cr-clusters. The branch 'rabbit' also postdates *Cr-reduction:

Why weren't stems 2-10 also borrowed from Chinese? What are their etymologies?

None of the Tangut names for the stems seem to be borrowed from Chinese: e.g., stem 1 is

TT0860 nɛ̣j R63 2.53

which can also mean 'shoots; seedlings' - hence the 'wood' radical on top indicating a plant.. The tangraph for stem 2 consists of the bottom of the stem 1 tangraph. It also has a botanical meaning though it lacks the 'wood' radical::

TT3985 liẹ R64 1.61 'sprout; bud; hatch from an egg'

No derivation for the stem 1 tangraph is known, but it appears to be based on the stem 2 tangraph. Why would the stem 2 tangraph have been created first? Did the creator(s) of the tangraphs only have the non-stem meanings in mind?

One might expect the third stem tangraph to also refer to plants, but it has no other meaning:

TT0571 bi R11 1.11

It has one of the two radicals for 'fire' on the left.

Since stems 1 and 2 are associated with the wood element, perhaps the Tangut selected two existing plant words for them. Stem 3 is associated with fire but does not have another fiery gloss. It is homophonous with

TT1648 bi R11 1.11 'light; ray; bright; shining'

(why is jaʳ R87 1.82 'eight' on the left?)

so perhaps they are really one and the same:

'bright' > 'bright like fire' > 'stem associated with fire'

I doubt that the Tangut simply created a word for 'stem 3' ex nihilo. If they did, why didn't they also create arbitrary names for stems 1 and 2?

Similarly, I assume the Wa stem names are also not ex nihilo but recyclings of existing words. For example, is Wa nap 'stem 2' homophonous with a botanical word? I'd check the Wa dictionary but it wasn't working when I was last online and I'm offline in a plane at the moment. Maybe I'll be able to check after I land ...

09.1.4.23:12: TWELVE ANIMALS OF THE TILE PEOPLE (who?*)

On Friday I found a list of Wa names and a Wa dictionary that are currently offline. The list contained the Wa names for the Chinese twelve-animal cycle borrowed through Tai. Not surprisingly, 'ox' has initial pl-:

General note: Breathiness (represented by a subscript umlaut) doesn't seem to correspond to anything in Chinese.

Animal-specific notes:

1. The Wa form implies a southern OC *Cʌ-tsəʔ with an emphatic prefix triggering vowel bending:

*ə > *əɰ ?[ʌɰ]

3. It's not clear whether the southern OC *N- implied by the Wa form was also present in northern OC.

I don't know why ɲi̤ has no -n.

OC *i corresponds to Wa i here and in 'dragon', but to a in 'monkey'. Does this reflect multiple strata of borrowing into Tai?

4. My OC reconstruction predicts a disyllabic Wa form, but mau is monosyllabic.

The absence of -n in si is inexplicable (cf. 'tiger') unless OC *-r went to zero after *i in the source dialect.

8. The Tai source of the Wa form may have reduced OC *-ts to *-t. It is also possible that a southern OC dialect lacked final *-s in this word.

9. Wa -ŋ implies OC *-ŋ. Cf. the graphically similar character 田 OC *Cʌ-liŋ > *lin.

Wa -a- is reminiscent of the lowered vowel in Cantonese san and seems to reflect a southern Chinese-internal innovation. This low vowel and the initial s- (as opposed to a lateral) imply late borrowing into Tai.

The Wa rhyme -au may or may not reflect an OC vowel-lowering emphatic prefix. Perhaps Wa -au reflects a southern Chinese dialect in which *-u had broken to *-əu or the like: cf. Cantonese jaw corresponding to MC *juʔ.

11. Wa m- may reflect Chinese *m- simplified from *sm- or *hm-, or the bare initial of a root without a prefix *s-.

The mid vowel may reflect an OC presyllable *sʌ- that fused with the initial in mainstream OC but triggered emphasis before being lost in southern OC:

mainstream: *sʌ-mit > *smit > *hmit

southern?: *sʌ-mit > *sʌ-mit > *sʌ-meit > *met

12. Wa -aɯ is like pre-MC / late OC -əɰ. Cf. 'rat'. Wa k- probably reflects pre-MC / late OC *g- rather than the *-k- of early OC.

More on Wa names here.

*The sinograph for 'Wa' is 佤: 亻 'person' (semantic) + 瓦 Md wa 'tile' (phonetic).