(Either is meant to be pronounced with [ay] in the title.  I don't know whether I tend to use the [ay] pronunciation or the [i] pronunciation.)

I'm out of time tonight, so I'll just point out one important fact:

Tibetans generally transcribed pre-Tangut period northwestern Chinese -Vw and -Vy rhymes as -VHu and -VHi: e.g.,

SinographTibetan transcriptions of pre-Tangut NWCTT for Tangut transcription of Tangut period NWCRhymeNishidaSofronovGongTibetan transcriptions of rhyme of Tangut transcriptionModern NWC dialects
keHu35551.43kəwkeɯkew-i, -iH; -e (only once) (never -eHu!)ko, kO, kau (no i or e!)
gwe, Hgwe, HgoHi53921.40vǐewaiwəy-i, -e (no -aHi, -eHi, or -oHi)uE, vE

(9.15.13:55: Less frequent Tibetan transcriptions of pre-Tangut NWC rhymes ending in glides included -VHï (ï = reversed i), -VHe and the very rare -VHo [in tsaHo for 糟; no other examples].)

I don't know of any Tibetan transcriptions of Tangut period NWC, but I do know that the Tibetan transcriptions of Tangut that I have seen lack -VHu and -VHi.  Does this mean that Tangut lacked -Vw and -Vy rhymes?  If Tangut had no syllables ending in glides, then how could it have 105 rhymes without any final consonants at all? NO -AW AT ALL?

(In my dialect, all is [ɔɰ] which is close to -aw.)

Nishida (1966: 417) reconstructed

TT4205 (a surname written as WATER + SWORD; last seen here)

as taw, which sounds like 刀 Tangut period northwestern Chinese ?*taw 'sword' but conflicts with

how TT4205 was used to transcribe 段 TPNWC *thwã

how others reconstructed TT4205:

Sofronov (1968 II: 370): thwan [thwã] 2.22

Li Fanwen (1986: 287): thuã 2.22

Gong (in Li 1997): thwã 2.22

the classification of TT4205 in Precious Rhymes of the Tangraphic Sea as a rhyme class 2.22 syllable. Nishida reconstructed 2.22 as (not unlike the of others' reconstructions). -aw is his reconstruction for rhyme class 2.20.

Even if TT4205 did belong to rhyme class 2.20, others don't agree that it ended in -w:

Sofronov (1968 I: 137): -âɯ < earlier *-âC (C = consonant)

Li Fanwen (1986: 188): -ǐa̯ (if ǐ = [y], what is a̯? An ultrashort [a]?)

Gong (in Li 1997: 2): -iaa

In fact, none of these three reconstructions have a rhyme -aw (or -au). Moreover, neither Li (1986) nor Gong (1997) have a rhyme -ay (or -ai).

Why couldn't the most common vowel a be followed by either -w or -y? Gong's reconstruction has seven vowels (ignoring length, tenseness, and retroflexion). Out of those seven, three combine with w and four combine with y:

glide codafrontcentralback

Is the absence of -aw and -ay purely by chance? I don't think so. I wonder if pre-Tangut *a assimilated to a following glide:

*-aw > -ow (*a became labial to match labial -w)

*-ay > -ey (*a became palatal to match palatal -y)

Perhaps some -ow and -ey are primary (i.e., original) rather than secondary:

-ow < *-ow (primary), *-aw (secondary)

-ey < *-ey (primary), *-ay (secondary)

Next: Does other evidence support this hypothesis? SPADA D'ACQUA

A near-homophone of ?SANDALWOOD (see here; actually ?BEACH/RAPIDS) was the surname

TT4205 thwã 2.22

Unlike many surname tangraphs, it lacks the right side of

TT3890 SURNAME 2.25

I don't know why some surnames have the right side of SURNAME (Nishida's radical 164 CLAN) and others don't. Did CLAN represent a common morpheme in some but not all Tangut B versions of surnames* (e.g., like Mc-, O'-, or son?).

TT4205 appears to consist of WATER + SWORD:


TT4205 thwã 2.22 =

?left of TT4282 WATER zyïïr 2.85 +

?all of TT2570 SWORD byïr 1.86

(TT4205 has no surviving analysis, so the above is my guess.)

Obviously, neither is phonetic (in Tangut A). One might guess that the thwã were a clan associated with water and swords (why?).

Could TT4205 be a calque of a sinograph? I can't think of any Chinese surname with WATER + SWORD.

Nishida (1966: 417) reconstructed TT4205 as taw, which is identical to what the Tangut period northwestern Chinese word for 'sword' might have been: 刀 ?*taw. So was SWORD being used as a Chinese-based phonetic?

Probably not, because taw doesn't match the rhyme class of TT4205. Nishida (1964: 48) reconstructed rhyme class 2.22 as which isn't too different from Gong's -ã. Nishida's (1964: 95) grid of dental-initial syllables in Homophones implies twà which is close to Gong's thwã. It's not clear why Nishida reconstructed an unaspirated initial t-, or why he changed the rhyme to -aw (his reconstruction for 2.20, not 2.24). The use of TT4205 to transcribe 段 Tangut period NW Chinese ?*thwã indicates that Gong's reconstruction is correct, unless the tangraph had two readings, thwã and taw.

Next: No -aw at all?

*If Tangut B existed, its names may not have been Tangut A names filtered through a different phonology (or vice versa). Name 'equivalents' can be noncognate and even mistranslated: e.g.,

Irish Ó hÉanacháin or Ó hÉinigh > English Bird (pseudo-translation from Irish éan 'bird')

Irish Ó Fathaigh > English Green(e) (pseudo-translation from Irish faiche 'lawn') MORE WOODLESS SANDALWOOD

(The original "WOODless SANDALWOOD" post is here.)

Apparently the actual Tangut word for 'sandalwood' was written as


TT0674 SANDALWOOD byịy 1.61 =

(top of) TT0673 佳 GOOD zur 2.69 +

(left of) TT1070 金剛 VAJRA Gyị̈ 1.69

right of TT2129 喜愛 TO-LIKE tshyï 1.30

There are two curious things about SANDALWOOD.

First, it doesn't contain the WOOD element at all.

Although both VAJRA and SANDALWOOD have Buddhist associations, I don't see how they relate to each other. (Did the Tangut make vajras out of sandalwood?)

GOOD and TO-LIKE seem rather generic as semantic elements. The defining characteristic of sandalwood is its fragrance (hence one of its Chinese names is 檀香[木] 'sandalwood-fragrance-[wood]'). Why wasn't SANDALWOOD written as, say, WOOD + FRAGRANT?

Second, sandalwood is not native to the Tangut Empire. So I would expect it to have a borrowed name.

檀 'sandalwood' was pronounced as ?*thã in Tangut period northwestern Chinese. This ?*thã came from an earlier *dan which in turn was an abbreviation of 旃檀娜 ?*chïan dan na, a Sinified version of Sanskrit chandana 'sandalwood' (< chand 'gladden' + noun suffix -ana).

The sandal of sandalwood is also ultimately from chandana. (Why did -n- change to -l somewhere along the way?)

byịy 1.61 obviously does not look like chandana, its Sinified form, or its Tibetanized form tsan dan. How did byịy 1.61 come to mean SANDALWOOD? I can only assume that byịy 1.61 originally meant something else - some other kind of wood, or some attribute associated with sandalwood. byịy 1.61 had six homophones:

TT1232 昇 ELEVATE byịy 1.61

TT1233 富 RICH byịy 1.61

TT1426 LOOK (仰視 LOOK-UPWARDS) byịy 1.61

TT1438 FOOLISH (丸 PILL) byịy 1.61

TT4884 危 DANGEROUS byịy 1.61

TT5582 DAMP (烘烤 TOAST) byịy 1.61

(Alternate glosses in parentheses are from Li Fanwen [1986: 205].)

None look like probable sources of SANDALWOOD byịy 1.61. WOODLESS SANDALWOOD

Nishida (1966: 418) and Grinstead (1972: 106) glossed

TT4204 thã 1.24 (last seen here)

as SANDALWOOD. Yet it doesn't contain the tangraphic element WOOD. WOOD is not in its Tangraphic Sea analysis:


TT4204 thã 1.24 =

(left of) TT4296 SURROUND 'wiọ 1.71 (i.e., WATER) +

(all of) TT1821 BROAD low 2.47 +

right of TT5716 EARTH tser 1.77 (i.e., an E-shaped element of unknown function)

How could WATER + BROAD + ? convey SANDALWOOD? If the E-shaped element were from, say,

TT0881 WOOD syi 1.11

the Tangraphic Sea should have pointed this out.

The answer is simple: TT4204 may not have meant SANDALWOOD. Shi et al. (1983: 440), Li Fanwen (1986: 289), and Shi et al. (2000: 76) defined it as 灘 ?BEACH/RAPIDS. Its clarifier tangraphs in Homophones are

TT4282 WATER zyïïr 2.85

TT5232 INSIDE gu 2.1

WATER INSIDE is also the definition of TT4204 in Tangraphic Sea (F31A33).

So why did Nishida and Grinstead interpret TT4204 as SANDALWOOD? Gong's (1991) index of Tangut transcriptions of Chinese in The Forest of Categories offers a clue: TT4204 was used to transcribe 檀 'sandalwood', which would have been pronounced as ?*thã in Tangut period northwestern Chinese. 檀 can also be a surname. Without seeing the Tangut text, I cannot determine whether TT4204 represented 'sandalwood' or a surname. In either case, the tangraph was presumably originally supposed to represent thã 1.24 ?BEACH/RAPIDS, a loan from TPNWC ?*thã 'id.' (The landlocked Tangut couldn't see any beaches, so they would have lacked a native word for the concept.)


9.11.23:26: Gong's (1981) article on Chinese loanwords in Tangut does not list TT4204 as a Chinese loanword. *dOn BUY THIS ETYMOLOGY

Last night, I wondered if

TT3228 PIECE/LUMP (Nishida [1966: 330]: 半分に切る CUT-IN-HALF, Li Fanwen [1986: 273]: 斷 CUT-OFF, Jin et al. [2000: 161] 段 SECTION) thọ 1.70

was a loan from Chinese. thọ looks like a blend of

*dOn, an Early Mandarin reading for 斷 and 段

and the devoicing *d > *th characteristic of the Tangut period northwestern Chinese dialect

Was the TPNWC reading of 斷 and 段 more like ?*thÕ than the ?*thwã that I reconstucted last night?

First, was TT3228 really thọ? Although its initial (th) and its rhyme category (tone 1, rhyme 70) are secure, the actual phonetic shape of its rhyme is open to question. I know of no Tibetan transcriptions for 1.70 syllables but Sofronov (1968 II: 44) lists transcriptions ending in -o and -oH for syllables in rhyme category 2.62, the 'rising tone' counterpart of 1.70. Thus TT3228 was pronounced something like tho, though whether it had a tense vowel (indicated by the subscript dot) as opposed to some other peculiarity is uncertain.

Second, did 斷 and 段 have an *o-like vowel in TPNWC? Gong (1991, rep. 2002: 438) found that TPNWC 段 was transcribed in Tangut as

TT4204 SANDALWOOD thã 1.24 (< loan from 檀 TPNWC ?*thã)

(Guess what semantic element is missing. Answer next time.)

TT4205 (a surname) thwã 2.22

which Gong reconstructed with the vowel a, not o. Even if one rejects Gong's reconstruction, one must acknowledge that rhyme classes 1.24 and 2.22 were transcribed in Tibetan as -a, -ang, -an (Nishida [1964: 48], Sofronov [1968 II: 19]). This implies that the TPNWC reading of 段 had*ã, not an*o-like vowel. So my reconstruction of 段 (and its homophone 斷) as ?*thwã is probably not far off the mark. In modern NWC dialects, syllables rhyming with 段 and 斷 end in -wã, -wæ̃, -wẼ (Coblin 1994: 310).

The mismatch of vowels (Tangut : TPNWC *ã) forces me to regard thọ and TPNWC ?*thwã as unrelated words sharing nothing more than an initial. I had wondered if TPNWC might have undergone a *-wã > *-wÕ > -wã shift influenced by the *-wan > *-On > -wan shift in the east, but nothing of the sort had ever happened. -wã might have remained more or less unchanged in the northwest from the Tangut period up to the present.

9.11.00:19: Gong's (1981) article on Chinese loanwords in Tangut does not list TT3228 as a Chinese loanword.

9.11.1:01: Gong (1991) showed that very few tangraphs in the first and second small rhyme cycles were used to transcribe TPNWC. This indicates that the small rhyme cycles had some characteristic(s) (generally assumed to be tenseness and retroflexion) that made them inappropriate for approximating TPNWC rhymes. I am fairly certain about retroflexion but am not so sure about tenseness.

9.11.1:21: How can I be so sure that TPNWC had nasal vowels instead of *-VN vowel-nasal sequences? Pre-Tangut period Tibetan transcriptions sometimes have open vowels corresponding to earlier *-VN and to modern -Ṽ (though not in the rhyme class of 段 and 斷.) Since it is unlikely that *-VN became non-nasal *-V and then regained its nasality in modern times, it is simpler to interpret Tibetan -V as a transcription of pre-Tangut NWC *-Ṽ. (The Tibetan alphabet has no letters or diacritics for nasal vowels.) It seems that the shift toward nasal vowels had already begun during the pre-Tangut period. I suspect that it was completed by the Tangut period (under Tangut influence? - but then why are there no Tangut loanwords in NWC? - did Tangut nasalize under Chinese influence?). MORE WOODEN GRASS

TT0848 PLUM ner 2.71 (analysis unknown)

contains both WOOD (on the top) and GRASS (on the bottom left). Did the Tangut think of PLUMs as WOODen GRASS? No, for GRASS turns out to be the left half of the phonetic of PLUM,

TT3277 萎 WITHER ner 2.71

The structure of PLUM is similar to sinographs such as


which combine 艹 GRASS (semantic) with 木 WOOD within a phonetic:

Tangut: (卅 WOOD on top (phonetic: GRASS + something else))

Chinese: (艹 GRASS on top (phonetic: 木 WOOD + something else))

There are five other tangraphs with WOOD over GRASS:

TT0845 SHORT (Nishida [1966: 330]: 短い[稻] SHORT-[RICE-PLANT], Li Fanwen [1986: 363]: 棘 BRAMBLE, Shi et al. [2000: 89]: 刺 ?THORN) tsorw 1.91

TT0846 GOURD lu 2.1

TT0847 BAD-SMELL dəə 1.31

TT0849 PLANT (Li Fanwen [1986: 347]: 柴 ?FIREWOOD; Shi et al. [2000: 90]: 草 GRASS) syï 1.30

TT0850 (a kind of tree) rur 1.75

Do all of these share the WOOD-over-GRASSy phonetic structure of PLUM?

TT0846 and TT0850 clearly do:

TT0846 GOURD lu 2.1

TT3223 (a kind of grass) lu 2.1

TT0850 (a kind of tree) rur 1.75

TT3289 吃草 EAT-GRASS rur 1.75

(< GRASS + TT5233 HORSE ryiry 1.74)

The other three don't:

TT0845 SHORT (Nishida [1966: 330]: 短い[稻] SHORT-[RICE-PLANT], Li Fanwen [1986: 363]: 棘 BRAMBLE, Shi et al. [2000: 89]: 刺 ?THORN) tsorw 1.91

TT3228 PIECE/LUMP (Nishida [1966: 330]: 半分に切る CUT-IN-HALF, Li Fanwen [1986: 273]: 斷 CUT-OFF, Shi et al. [2000: 161] 段 SECTION) thọ 1.70

(unless the phonetic 'fit' is very loose)

TT0847 BAD-SMELL dəə 1.31

TT3270 DEMONS 鬼魅 (Grinstead [1972] and Li Fanwen [1986: 442] had no def.) lew 2.38

looks like GRASS + DEMON

TT0849 PLANT (Li Fanwen [1986: 347]: 柴 ?FIREWOOD; Shi et al. [2000: 90]: 草 GRASS) syï 1.30

TT3273 (syllable of surname) pạ 2.56

looks like GRASS + SURNAME

The Tangraphic Sea analyses of BAD-SMELL and PLANT reveal that their bottom components are amalgams rather than unitary phonetics:


TT0847 BAD-SMELL dəə 1.31=

top of TT0681 VILLAGE gyiy 2.33 +

right of TT2305 PUTRID tshie 2.7



TT0849 PLANT (Li Fanwen [1986: 347]: 柴 ?FIREWOOD; Jin et al. [2000: 90]: 草 GRASS) syï 1.30 =

top of TT0881 WOOD syi 1.11 +

right of TT5516 GRASS shyï 2.60 (is this phonetic?) +

right of TT4484 ORIGIN r 2.76

But the analysis of


TT0845 SHORT (Nishida [1966: 330]: 短い[稻] SHORT-[RICE-PLANT], Li Fanwen [1986: 363]: 棘 BRAMBLE, Jin et al. [2000: 89]: 刺 ?THORN) tsorw 1.91 =

top of TT0881 WOOD syi 1.11 +

all of TT3228 PIECE/LUMP (Nishida [1966: 330]: 半分に切る CUT-IN-HALF, Li Fanwen [1986: 273]: 斷 CUT-OFF, Shi et al. [2000: 161] 段 SECTION) thọ 1.70

implies that TT3228 is either phonetic or semantic in TT0845, even though the two don't sound much like each other, and the semantic link between SHORT/THORN and PIECE/LUMP/CUT-IN-HALF/CUT-OFF/SECTION is not clear. Were the two (nearly) homophonous in Tangut B?

One need not resort to Tangut B to devise a solution involving homophony. In Tangut period northwestern Chinese, 短 'short' was ?*twã and 斷 'cut off', and 段 'section' were ?*thwã. Thus the structure of TT0845 could have been:

WOOD (semantic? - what does this have to do with SHORT?) +

TT3228 (semantic via near-homophony of its TPNWC translations 斷 'cut off' or 段 'section' with TPNWC 短 'short')

9.10.00:51: Obviously, this analysis only works if Nishida and Grinstead's translation of TT0845 as SHORT (contra Li Fanwen and Shi et al.) and Li Fanwen or Shi et al.'s translations of TT3228 (contra Nishida and Grinstead) are correct. It is difficult to test semantically-based hypotheses of tangraphic origins because of the lack of consensus on translations.

9.10.1:01: TT3228 was analyzed as


TT3228 PIECE/LUMP (Nishida [1966: 330]: 半分に切る CUT-IN-HALF, Li Fanwen [1986: 273]: 斷 CUT-OFF, Jin et al. [2000: 161] 段 SECTION) thọ 1.70 =

[top and] left of TT0845 SHORT (Nishida [1966: 330]: 短い[稻] SHORT-[RICE-PLANT], Li Fanwen [1986: 363]: 棘 BRAMBLE, Jin et al. [2000: 89]: 刺 ?THORN) tsorw 1.91 +

right of TT3819 DISCRIMINATE sywiy 1.36

I would have expected TT3228 = all of TT0845 minus the top of TT0881 WOOD.

The function of the right side of DISCRIMINATE is obscure in this tangraph and in others such as

TT1465 NEW syiw 1.46 (with NOT + the right of OLD)

TT4250 LATER mya 1.20 (with WATER + PERSON)

TT4822 MAESTRO (Shi et al. [2000: 136]: 男人 MAN) lụ 1.58 (with ? + PERSON)

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