65: Two tangraphs in this line have 'horned hats' (bio). Why do some surname tangraphs have them instead of dex 'person' or pux 'surname'? Do these common 'onomatangraphic' elements signify different types of families: e.g., do nobles get to write their name with one radical absent from the names of lower classes?

Tangraph number 321 322 323 324 325
Li Fanwen number 0570 0794 4867 3229 4938
My reconstructed pronunciation 1dʒɔ 2roʳ 1ʃɨəʳ 1ŋwəə 1ɣa
Tangraph gloss first half of the surname Joror second half of the surname Joror the surname Shyr to chant an incantation; first syllable of Tangut surnames the surname Gha
Word the surname Joror Ngwygha
Translation Joror, Shyr, Ngwygha,

321: 0570 has a phonetic + semantic structure:


0570 1dʒɔ 'first half of the surname Joror' (wiadex; wia = puebee) =

1218 1tʃhɔ 'the surname Chho' (puebee; phonetic; see line 66) +

2888 2mə 'surname' (dexpux; semantic)

The fanqie for 0570 in Mixed Categories of the Tangraphic Sea is a bizarre combination of the Grade I/IV initial dz- with the Grade II rhyme -ɔ:


4497 1dzaa + 4507 2ʃɔ

However, 0570 is in the alveopalatal sections of both Tangraphic Sea and Homophones, and alveopalatals are Grade II/III initials that can combine with -ɔ. So the fanqie in Tangraphic Sea could

a. be an error

b. reflect an irregular variant reading with a rare grade mismatch

c. be interpreted as evidence that the author regarded [dʒ] as an allophone of /dz/ before

322: 0794 (porzorcin; analysis unknown) is probably derived from its homophone

2757 2roʳ 'circle' (dexzorcin)

The function of the por replacing dex 'person' is unknown.

323: 4867 was analyzed in terms of nearby tangraphs:


4867 1ʃɨəʳ 'the surname Shyr' (biodexheu) =

4940 2ʔiə 'the surname Y' (biotumbee; see next line) +

2854 0dzị 'the surname Dzi' (dexheu; see line 64)

Were the Shyr related to the Y and the Dzi?

324: Is 3229 in surnames a syllable that happens to be homophonous with 'chant an incantation' or does it refer to 'chanting' branches of families: e.g., is the Ngwygha the chanting branch of the Gha?

The analysis derives the right side tus from a combination of radicals.


3229 1ŋwəə 'to chant an incantation' (dextus) =

3190 1lhwia 'tongue' (dexcuo) +

3228 1thioo 'beautiful, wonderful' (dexjeu; what is jeu 'eight'? a reference to the Noble Eightfold Path?)

The source tangraphs are in normal Tangut noun-adjective order and could be a phrase.

Why is dex 'person' in all three tangraphs ('chant', 'tongue', 'beautiful')?

325: 4938 has a 'horned hat' from the same source as 4867. Do all the surname tangraphs with hats (or at least those with hats from 4940) represent related families?


4938 1ɣa 'the surname Gha' (biogoljas) =

4940 2ʔiə 'the surname Y' (biotumbee; see next line) +

4938 1ɣa 'bag' (goljas; phonetic; why is 1lew 'one' on the left?)

66: This line contains two tangraphs we've already seen in the analyses for line 65. 4940 and 1218.

Tangraph number 326 327 328 329 330
Li Fanwen number 4940 2714 1218 4562 4832
My reconstructed pronunciation 2ʔiə 1riuʳ 1tʃhɔ 2zəʳ 1tshie
Tangraph gloss the surname Y the surname Rur the surname Chho the surname Zyr the surname Tshe
Word Yrur Zyrtshe
Translation Yrur, Chho, Zyrtshe.

326: 4940 probably has 1278 as phonetic beneath a horned hat (bio):


4940 2ʔiə 'the surname Y' < 1278 2ʔiə 'to say'

The function of bio is unknown.

1278 also has a level tone cognate with a similar tangraph which has tun 'skin' (why?) on the left:


1279 1ʔiə 'to say' (tunbee) =

0795 2riʳ 'perfective prefix' (tunduucin) +

0677 2ʃwɨu 'to flatter' (beedexgoi)

I doubt that 0677 'flatter' was devised before the far more common 1279 'say'.

327: Were the Rur related to the Me (from "Me-aty Monarchs"?)


2714 1riuʳ 'the surname Rur' (baegemhincin) =

2919 2mie 'the surname Me' (baedexfak) +

2795 1riuʳ 'to herd; herbage' (gemhincin)

2795 (which looks like gem 'grass' + hin 'horse') has a circular analysis:


2795 1riuʳ 'to herd; herbage' (gemhincin) =

3719 'to herd, tend' (gemtae) +

2714 1riuʳ 'the surname Rur' (baegemhincin)

328: Were the Chho known as gatherers?


1218 1tʃhɔ 'the surname Chho' (puebee; phonetic) =

1535 1loo 'to gather' (pueqaldex; qaldex = 'person') +

1259 1tʃɔ̃ 'to gather (beebee; the Tangraphic Sea specifies the right side)

329: The unknown analysis of 4562 might be


4562 2zəʳ 'the surname Zyr' (jeucirdex) =

4607 2zəʳ 'dew' (jeucen; phonetic)

2888 2mə 'surname' (dexpux; semantic)

Note that cen is the right-hand variant of cir 'water'.

The unknown analysis of 4607 might be


4607 2zəʳ 'dew' (jeucen; phonetic) =

3228 1thioo 'beautiful, wonderful' (dexjeu; semantic) +

3058 2ziəəʳ 'water' (cirzaa; semantic)

330: Are the Tshe more relatives of the Y?

If 4653 is phonetic in 4832, the fit is a little loose:


4832 1tshie 'the surname Tshe' (biobukquu) =

4940 2ʔiə 'the surname Y' (biotumbee) +

4653 1tshiaʳ 'to instigate' (bukquu; phonetic) THE GOLDEN GUIDE: LINE 64: TANGRAPHS 316-320

64: It took me a week to finish this line, even though it's very simple:

Tangraph number 316 317 318 319 320
Li Fanwen number 4561 2284 5405 2607 2854
My reconstructed pronunciation 2bia 2dʊ 2ma 1kwiə̣ 0dzị
Tangraph gloss the surname Ba the surname Du the surname Ma the surname Kwy to cover; the surname Dzi
Word the surname Badu the surname Kwydzi
Translation Badu, Ma, Kwydzi.

316: Were the Ba related to the Lu and Me?


4561 2bia 'the surname Ba' (viadex) =

4579 2lɨu 'the Chinese surname 呂 Lü' (viacin; via = buk 'mouth' doubled parallels the structure of Chinese 呂 which can also be written as two mouths 吕) +

2259 2mɛ 'the Chinese surname 孟 Meng' (dextiibuk)

Unlike Lu and Me, Ba is not a surname of Chinese origin.

317: Although Kotaka's online version of the Golden Guide has 2290 2lõ 'round', I suspect that is an error for the similar tangraph 2284 2dʊ:


2290 (dexfol) <> 2284 (dexfoi)

I don't know if this error is in the original GG.

2284 is a straightforward semantic-phonetic compound:


2284 2dʊ 'the surname Du' (dexfoi) =

2888 2mə 'surname' (dexpux; semantic) +

1051 2dəəu (foijeu; phonetic; why is jeu 'eight' on the right?)

I still do not understand why d- precedes rhyme 4 unlike other dental initials which precede rhyme 1 -əu.

I assume that Ba and Du can be independent surnames on the basis of annotations in the D version of Homophones which gloss both as 'surname'. In any case, there is no doubt that they represent a disyllabic surname Badu.

318: 5405 2ma 'the surname Ma' (analysis unknown) is probably phonetic in 5406 which has an extra vertical line (bae) and an improbable, circular analysis:


5406 1maaʳ 'wonderful, beautiful' (boxdumbaejox) =

5657 1nwiə 'luxuriant, majestic' (dumher) +

4378 1maaʳ 'a kind of tree' (boxdumbaejox)

4378 must postdate 5406.

319: Were the 2607 Kwy related to the Tsi and even to the 2606 Kwy family?


2607 1kwiə̣ 'the surname Kwy' (dexcirqos) =

3120 2tsi 'the surname Tsi' (dexcircor; is cir 'water' symbolic?) +

2606 1kwə̣ 'the surname Kwy' (dexqos)

Do the different grades of the two Kwy names (Grade IV 2607 and Grade I 2606) indicate a proto-name with different prefixation conditioning different rhymes?

Zero or high vowel presyllable > Grade IV

*s(ɯ)-kwə̣ > 2607 1kwiə̣

Low vowel presyllable > Grade I

*sʌ-kwə̣ > 2606 1kwə̣

320: Why was 'cover' recycled for surnames? Why not create a tangraph 0dzị without any semantic baggage?

Do covered things reside quietly, or is this analysis suspect?


2854 0dzị 'to cover' (dexheu) =

0522 1dziẽ 'quiet' (qisdex) +

5464 2ʒɛʳ 'to reside' (tiiheu)

7.24.1:24: The tone of 2854 may have varied depending on idiolect or dialect.

In Homophones, 2854 is grouped with rising tone tangraphs (29A17-29A28).

But in Tangraphic Sea, 2854 is grouped with level tone tangraphs sharing the fanqie


1941 2dziə̣ + 5767 1s

Hence I have written the tone of 2854 as 0 for uncertain.. A SM-AW PROBLEM

Earlier this week, I Tangutized the 小 'little, small' in Bruce Lee's Chinese name (李小龍) as

3799 2sew 'small' (gesgux) < borrowed from Chn 小

written like the tangraph for its native equivalent

3798 1tsẽ 'small' (dexgux)

but with ges 'earth' instead of dex 'person' on the left added to gux 'small' on the right.

2sew is a Tangut Grade I syllable. If Gong (1995) is correct and if Tangut grades more or less correspond to Chinese grades (suggesting that the Tangut borrowed the concept from Chinese), then we would expect the original Chinese word 小 to also be Grade I. But in fact, 小 had Grade IV *-iew with *-i- according to the Late Middle Chinese rhyme table tradition. Moreover, Gong found that Chninese words which once had

LMC Grade I *-aw

LMC Grade I *-əw

LMC Grade III *-ɨw

as well as LMC Grade IV *-iew were all borrowed into Tangut with Grade I rhyme 44 -ew (西夏語中的漢語借詞, pp. 772-773).

Words which once had LMC Grade III *-ɨw could also be borrowed into Tangut with the expected Grade III/IV rhymes *-ɨew and *-iew. (Tangut initials determined whether a following high nonback vowel was central or front.)

Tangut transcriptions of Chinese have similarly mismatched grades (Gong 2002: 444, 456): e.g.,

Chinese My Tangut grade Rhyme of Tangut transcription Tibetan transcriptions of Tangut rhyme (frequency in parentheses from Tai 2008: 216-217)
Graph LMC rhyme LMC grade This site Sofronov 1963 Nishida 1964 Hashimoto 1965 Sofronov 1968 Li Fanwen 1986 Gong 1997 Arakawa 1999
曹騷高澡 *-aw I I -ew -ы̂ -əw -ääw -eɯ -əu -ew -eu -i (16), -iH (13), -e (1), -a (1)
侯歐陋后口 *-əw
譙鷦蕭 *-iew IV
秋就修璓秀繡 *-iw
*-aw I IV -iew -i̭ы̂ -ǐəw -jäw -i̭eɯ -ǐəu -jiw -eu: -iH (5), -i (2), -ing (1), -eH (1), -uH (1)
堯姚繇 *-iew IV

(7.23.0:31: Added reconstructions other than mine and Tibetan transcriptions.)

Are the grades of LMC rhyme tables irrelevant to Tangut period northwestern Chinese? Is Gong's Tangut grade hypothesis wrong? Or is something else going on?

Here's what I think right now:

- The LMC reconstructions are not relevant because they lack innovations hinted at by both pre-Tangut period Tibetan transcriptions and the Tangut transcriptions. For example, LMC *-aw apparently shifted to *-ew since it was transcribed in Tibetan as -eHu: e.g.,

高 Tib keHu (cf. its Tangut transcription 1kew)

(But that Chinese rhyme was also transcribed in Tibetan as -aHu ~ -aHo and -oHu. Perhaps -a/eHo/u transcribed *-ɛw or *-æw and -oHu reflected some other northwestern Chinese dialect.)

- The grade system was originally designed for the dialect in the LMC rhyme tables and may not have been applicable to Tangut period northwestern Chinese.

- The Tangut grade system may have been absent (lost?) in the dialect(s) of the transcribers. A distinction in the presumably standard dialect recorded in Tangraphic Sea did not necessarily exist in all other Tangut dialects. The extent of diversity within spoken Tangut in the 11th and 12th centuries is unknown, but may be inferred from the many mismatches between the various bodies of evidence for Tangut phonology. The phonology of the transcribers may be very different from the phonology in native Tangut dictionaries, just as the phonology of Cantonese-based transcriptions of non-Chinese words is very different from the phonology in standard Mandarin dictionaries.

- The Tangut reconstructions are probably wrong. Although my simplified transcription merges Grade I -ew and Grade IV -iew as -ew, ignoring grades entirely, I am not even certain that -ew is their lowest common denominator, since these rhymes were transcribed as -i in Tibetan rather than as the expected -eHu. But perhaps the Tibetan transcriptions represent a Tangut dialect with final *-w loss. ME-ATY MONARCHS

Were the Me a family of prominent butchers?


2919 2mie 'the surname Me' (baedexfak) =

3465 1tʃɨi 'meat' (baedexgus; gus = 'meat' < Chn 肉 'meat'?) +

3150 2thiə 'emperor' (dexgiifak < 'person' + 'sun' + ? [fak is only in right-hand position])

The name Me came up in last night's analysis of the tangraph for the surname Bu (relatives of the Me?).

I'm surprised there is no tangraph dexgus (3465 minus bae). I wonder if baedex is a unit, just as ger = dexbae is a unit.

7.22.0:10: baedex appears to be a unit in this analysis:


2839 ko1 'the surname Ko' (baedexpux) =

2946 ko1 'to put the palms together' (baedexbee; phonetic) +

2888 2mə 'surname' (dexpux; semantic)

One might think baedex is a phonetic for (1)ko in other tangraphs, but this shape-sound correspondence is unique to 2839 and 2946.

Moreover, baedex has two parts if this analysis can be believed:


2946 ko1 'to put the palms together' (baedexbee) =

3485 1lạ 'hand' (baepik; semantic)

3013 2reʳ 'canal; ditch' (cirdexdek; why?; don't confuse with 3121 1ɣiu 'the surname Ghu' or 3000 1ɣiu 'ditch'; see line 63)

1751 1ʃwɨa 'hand' (beepik; semantic)

3013 is a very unlikely source for dex in 2946. LI SEW LON (1940-1973)

Even if you don't know any Tangut, you've heard of

Li Sew Lon 'Li little dragon'

who passed away 37 years ago yesterday. Who was he?

His personal name could also be Tangutized as

Vi Tsen 'dragon little' =

'little dragon' with Tangut noun-adjective order

Notice how similar the characters for 'little' are. Only a dot distinguishes them:


tsen 'little' (native) <> sew 'little' (borrowed from Chinese 小)

The characters for 'dragon' are also similar


vi 'dragon' (native) <> lon 'dragon' (borrowed from Chinese 龍)

though if you look closely, they only share their bottom left and right components (干 and ヒ).

So who was Li Sew Lon / Li Vi Tsen? Select the blank space below for the answer:

Bruce Lee / 李小龍 Cantonese Lei Siu Lung 'Lee little dragon' THE GOLDEN GUIDE: LINE 63: TANGRAPHS 311-315

63: I actually started this post last Friday, but didn't finish it until tonight.

Tangraph number 311 312 313 314 315
Li Fanwen number 1925 3121 3823 5031 0547
My reconstructed pronunciation 1bɨu 1ɣiu 2ne 1bə 2nie
Tangraph gloss the surname Bu the surname Ghu the surname Ne second half of Lyby, ancestor of the black-headed Tangut, transcription of Sanskrit b- the surname Ne
Word the surname Bughu the surname Byne
Translation Bughu, Ne, Byne,

311: Were the Bu an artistic family related to the Me, and were they and the Ghu related to the Bughu?


1925 1bɨu 'the surname Bu' (baedumcok) =

2919 2mie 'the surname Me' (baedexfak) +

2150 2kiuʳ 'skill, artistry' (dexdumcok)

Although tangraphy has a reputation for complexity, 2150 is simpler than its Chinese translation equivalent 藝 (though more complex than the modern simplifications of 藝: 艺 in the PRC and 芸 in Japan.)

7.21.1:02: Could the vertical line bae be sufficient to evoke 2919? I doubt it.

The analysis of 2150 is unknown. It looks like dex 'person' plus

5165 1twəụ 'each; place'

but that can't be its analysis, can it? The left two-thirds of 2150 look ike

3639 1rieʳ 'the surname Rer'

but I doubt there's a connection, unless the Rer were a family of artisans.

312: 3121 has a straightforward semantic + phonetic structure - and an unpleasant homophone:


3121 1ɣiu 'the surname Ghu' (dexcirdek) =

2888 2mə 'surname' (dexpux; semantic) +

3000 1ɣiu 'ditch' (cirdek)

313: 3823 (analysis unknown) at first looks like pik 'hand' plus cue, similar in shape to Chinese 广. But this analysis suggests that cue might actually be two radicals, bas + bae:


2644 1ʒɨiw 'the surname Zhiw' (dexvoi; voi = basdex) =

2888 2mə 'surname' (dexpux; semantic) +

3823 2ne 'the surname Ne' (cuepik = basbaedex; semantic?)

Oddly, 3823 2ne is a phonetic for 1ʔiew in

3939 1ʔiew 'the surname Ew' (with 'surname' on right)

3954 1ʔiew 'oil, fat, grease' (with 'water' on right)

3937 1ʔiew 'disease; illness' (with 'give birth to' on right)

Perhaps the Ew family was related to the Ne.

314: Why is 'scheme' in 5031?


5031 1bə 'second half of Lyby' (biobalbaeduucin) =

5042 1bə 'soft' (biobalbaebil) +

0797 1phi 'scheme, idea' (beeduucin)

The analysis of 5042 is circular and dubious, but at least it hints that bil and bim are equivalent (unless 'hair' was arbitrarily chosen as a 'soft' source tangraph for 'soft'):


5042 1bə 'soft' (biobalbaebil) =

5031 1bə 'second half of Lyby' (biobalbaeduucin) +

2061 2pɛ̃ 'hair' (bimbun)

315: The analysis of 0547 is unknown but is probably something like


0547 1nie 'the surname Ne' (qiedex) =

1671 1nie 'red' (qiepak; phonetic) +

2888 2mə 'surname' (dexpux; semantic)

The surname Ne may be the same word as 1nie 'red': cf. surnames like the rare Chinese surname 赤 'red', English Reid, Read (< 'red'), Dutch Rood(t) (< rood 'red'), etc. RAPID DUPLICATES

The fanqie for two of the three rhyme 105 tangraphs


4689/5014 1ʔwa(ʳ) = 2298 1ʔwiəəʳ + 5099 1ʃwa(ʳ)


2298 'fast'

which has an exact lookalike 2299:

2298 occurs in the level tone volume of the Tangraphic Sea with the definition

5014 1ʔwa(ʳ) 'fast'

whereas 2299 doesn't seem to occur anywhere. Sofronov listed two readings for 2298/2299:

Sofronov # Sofronov reading Li Fanwen 1997 # Li Fanwen 1997 reading Li Fanwen 1997 gloss Li Fanwen 2008 # Li Fanwen 2008 reading Li Fanwen 2008 gloss
3095 2·i̭ẹ 2298 1lhjwɨr very fast, 急速 2298 1·jwɨr fast, rapid, 急速
3096 1lhi̭ẹ 2299 2jɨr none 2299 none none

Sofronov listed no glosses and mixed up the tones, whereas Li Fanwen mixed up the initials.

2298 had a level tone, for its fanqie was


2298 1ʔwiəəʳ 'fast' = 3673 1ɣiu + 3674 1kwiəəʳ

with a level tone final speller. (Sofronov [1968 II: 272] accidentally listed this as the fanqie for his 1lhi̭ẹ in spite of the initial.)

I don't know why Gong's reconstructions in the different editions of Li Fanwen have (·)j- instead of ɣ- like 3673. Perhaps (·)j- was chosen on the basis of the Chinese transcriptive evidence (月/越; see this post), but I would favor the internal evidence.

3673 belongs to glottal fanqie chain 1 with an initial ɣ- that can be confirmed by its Tibetan transcriptions: g(h)- and k(-h)- (k- with subscript h) (Tai 2008: 195, 197).

Sofronov's reconstruction had a glottal stop instead of ɣ- in its consonant inventory, and its glottal stop cannot be reconciled with the Tibetan transcriptions.

All this leads me to reconstruct ɣ- as the initial of 2298, 5014, and 4698. Li Fanwen (1986: 429) reconstructed 2298 'fast' and 5014 'fast' as ɣǐəə and ɣǐua (sic; should be ɣǐuạ since he reconstructed R105 as -ǐuạ elsewhere). I suspect 2298 and 5014 are ablaut variants of the same root.1ɣw-ʳ:

2298 1ɣwiəəʳ < *prɯ-Kəə

5014 1ɣyaʳ < *prɯ-Ka

The features of the *prɯ-prefix (or a reduction of a first syllable of a disyllabic root?) were shifted into the following syllable:

- The labiality of *p- conditioned -w- and -y-

- The *-r- conditioned

- The *-ɯ- conditioned -i- and -y-, and *-K- intervocalically lenited to -ɣ-

Could the other native R105 word 4689 1ɣyaʳ 'bright, glittering' share the same root as 2298 and 5014 'fast'? Watkins (2000: 11) derived Old English bregdan 'to move quickly' from Proto-Indo-European *bherək- 'to shine', setting a precedent for a link between speed and brightness.

The only similar word I can think of in another language is Old Chinese 急 *kəp 'urgent'. I have wondered if Tangut vowel length orignated from lost codas, so perhaps *prɯ-kəə < *prɯ-kəp, but *prɯ-Ka must have had a short vowel because a descendant of its rhyme transcribed Skt śva with a short vowel.

Could Tangut ɣ- be from an intervocalically lenited *-x-? If so, then maybe these words for 'fast'/'bright' may be cognate to Tibeto-Burman words that Matisoff (2003: 428-430) regarded as descendants of his Proto-Tibeto-Burman *hwaC 'white, yellow, bright, shine'. THE SOUND OF MEAT

I've been writing about the pronunciation of 5099 (here and here), but not its structure. 5099 looks like dox 'meat' plus bos 'sound' but its analysis is slightly different:


5099 1ʃwa(ʳ) 'Sanskrit transcription tangraph' =

5701 1ʃwɨaa 'upper part of the body' (doxpik; pik = 'arm'; phonetic)

1586 1ɣɛ̣ 'sound' (bostal; semantic)

There is no doubt that 5099 has the same initial as 5701 since both belong to alveopalatal fanqie chain 8 whose initial was transcribed in Tibetan as sh- or zh- (Tai 2008: 192).

5701 is presumably phonetic in 5099 since there is no semantic reason for 'upper part of the body' to be a an abbreviated component in a transcription tangraph. So 5099 must sound something like 5701. The apparent use of 5099 to transcribe Skt śva confirms the similarity of the readings of 5099 and 5701.

The fanqie of 5099 is


5099 1ʃwa(ʳ) '= 4216 1ʃwɨi + 4689 (sounding like 月/越; see this post)

5099 in turn is in the fanqie for the other two rhyme 105 tangraphs:


4689/5014 1ʔwa(ʳ) = 2298 1ʔwiəəʳ + 5099 1ʃwa(ʳ)

The problem with 1ʔwa(ʳ) is that it should have been listed under R17 -a or R85 -aʳ. Why was it listed under R105? How did R105 differ from R17 and R85?

Now I'm cycling back to my earlier reconstruction of R105 as -ya (or -ɥa) with a palatal (semi)vowel. Why a palatal?

- Perhaps /ʃ/ + /ya/ = [ɕwa] with a palatal [ɕ] that was a better match for Sanskrit śva [ɕʋɐ]. (I suspect Tangut /ʃ/ and other alveopalatals were really retroflexes: [ʂ], etc.)

- If 4689/5014 1ʔwa(ʳ) really had no palatal, why would it be transcribed as 月 *jwa with a palatal? Why would 越 *jwa with a palatal be transcribed as 4689 1ʔwa(ʳ) without an palatal?

Maybe earlier Tangut had a lot of *-y- that merged with -w(ɨ/i)- except in two archaisms:

4689 1ʔya? 'bright, glittering'

5014 1ʔya? 'to go; fast, quick' (the name of R105 in Tangraphic Sea)

Conversely, these two words might have been the first to undergo a shift of -w(ɨ/i)- > -y-.

Unlike those two native words, the syllable 5099 1ʃya was concocted specifically to transcribe Skt śva. If Skt śva deserved a special Tangut syllable, why didn't other Skt śv- or even ś- syllables get their own special Tangut syllables? TO THE TREASURE LORD

I thought I'd get to this in my last post, but I ran out of time.

The only Sanskrit transcription containing the rhyme 105 tangraph 5099 that I know of is

4342 0724 5099 5523 (5032) 5314

(5032 'long' indicates a Sanskrit long vowel; its unknown Tangut reading is irrelevant.)

Li Fanwen's identification: 檀泥說羅也 = Skt dhaneśvarāya 'to the treasure lord'

Simplified romanization: da ny shwa(r) rar (long) ya

This site: 2dia 2niə 1ʃwa(ʳ) 1riaʳ (long) 2ʔia

Gong: 2dja 2njɨ 1ɕjwa(r) 1rjar (long) 2·ja

Arakawa: 2da: 2nI: 1shua 1ra:r (long) 2ya:

Hashimoto: 2-ääN 2-jaaj 1-joj(N) 1-jaa (long) 2-ääN

(Hashimoto did not publish a reconstruction of Tangut consonants. I have converted his idiosyncratic long vowel notation into double vowels.)

in Li Fanwen (2008: 806).

All these reconstructions have issues. Let's look at 5099 first.

My 1ʃwa(ʳ) / simplified shwa(r) is a guess based on Skt śva(r). In my reconstruction, ʃ- can normally precede Grade II R18 -wæ(ʳ) or Grade III R19 -wɨa(ʳ). ʃwa(ʳ) is an unusual combination of a Grade II/III initial with the Grade I final -wa(ʳ) that was closer to Skt -va(r) than those normally permissible finals. This reconstruction of R105 is identical to Grade I R17 -wa and R85 -waʳ except for its ability to follow Grade II/III initials. Would that unusual combination justify setting up a new rhyme? The other two R105 syllables present other problems for my reconstruction which I'll examine later.

Gong already reconstructed -jwa as R19-20 and -jwar as R87, so what distinguishes his R105 -jwa(r) from those rhymes?

If Arakawa's 1shua were correct, why would the Tangut transcribe Skt -va as a diphthong -ua instead of as a glide-vowel sequence -wa?

Hashimoto's -joj(N) for R105 bears no resemblance to Skt -va.

Now let's look at the transcriptions as a whole. In theory, I would expect the following reconstructions for a Tangut transcription of dhaneśvarāya [dɦɐneeɕʋɐraajɐ]. Tones are omitted since the logic of choosing tones for Sanskrit loans is unknown. Differences from the actual reconstructions are in bold.

Simplified romanization: da ne shwa(r) raar ya

This site: da nee ʃwa(ʳ) raaʳ ʔia

Gong: da nee 1ɕjwa(r) raar ·ja

Arakawa: da ne: shwa ra:r ya (nothing but mismatches!)

Hashimoto: -aa -e -aa -aa -ja or -ja -e -ja -aa -ja (nothing but mismatches!)

(Hashimoto's system has no short -a or long -e; short -a must be preceded by -j-.)

The mismatches between Sanskrit and the actual reconstructions* are of several types:

1. The actual reconstructions have many palatal vowels and glides absent from Sanskrit. Why would the Tangut transcribe CV as CiV or CjV? My gradeless simplified transcription lacks these unnecessary palatal segments and may be closer to the real Tangut readings.

2. None of the actual reconstructions has a ne-like syllable corresponding to Skt ne [nee]. They all have some non-e vowel (ə, ɨ, I:, aa). This rhyme (R31/2.28) was transcribed with all five Tibetan vowels (Tai 2008: 212) and therefore was probably a schwa-like vowel rather than -e.

3. Length in the actual reconstructions corresponds poorly to length in Sanskrit:

Sanskrit sylalble dha ne śva ya
Sanskrit length short long short long short
My length and Gong's length short short short short + (long) short
Arakawa's length long long short long + (long) long
Hashimoto's length long long short long + (long) long

If Tangut really had distinctive length, why was it necessary to indicate the length of the fourth syllable with 5023 'long'? Wouldn't a tangraph for raaʳ with a long vowel be sufficient to transcribe Skt rā?

Length in reconstructions may correspond to some other distinction in Tangut.

Lest one think that dhaneśvarāya is an isolated example of length mismatch, the Tangut transcription of Śākyamuni

is 2shI: 1ka: 2mo 2ni: in Arakawa's reconstruction (1999: 67) and 2ʃɨi 1kia 2mo 2ni in mine. (Unexpected lengths are in bold. Note also that the vowels don't quite match either.)

4. Hashimoto has final consonants corresponding to zero in Sanskrit. Although the -N of his first syllable could simply reflect the n- of the second Skt syllable, there is no excuse for the other codas because his system does have open rhymes which are better matches.

These mismatches lead me to conclude

A. The Tangut were not precise phoneticians and transcribed Sanskrit loosely. I find this difficult to believe given the extreme precision of the native Tangut works on phonology. Has any other people ever written such works shortly after creating a new and highly complex writing system?

B. The Tangut did not hear correct Sanskrit but transcribed some distorted pronunciation of it. Sanskrit borrowed through Chinese should be distorted, but Sanskrit borrowed through Tibetan shouldn't, since Tibetanized Sanskrit is very accurate.

C. The reconstructions are unreliable.

D. Some combination of all of the above.

*I don't count the retroflex vowels in most reconstructions as mismatches. Such vowels are obligatory after r-, so their presence is predictable. Apparently the Tangut thought it was OK to transcribe Skt rV as rVʳ, though in theory they could have created special rV syllables purely for Sanskrit transcription. The only nonretroflex r-syllable in my reconstruction is 2riẽ (= Gong's 2rjɨj), presumably a simplification of an earlier *2riẽʳ (= Gong's 2rjɨjʳ) with an unusual nasal retroflex vowel.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2010 Amritavision