In Kotaka's sample of Arakawa's reconstructions, what appear to be glottal stops* appear in 12 rhyme classes (21 if level and rising tone rhymes are distinguished):

Arakawa's short vowel rhymesArakawa's long vowel rhymes
(presumably 1.22?)-a'1.23-aa'
no level tone counterpart-ya'no long vowel counterpart
no short vowel counterpart1.40-ee'
no rising tone counterpart
1.52-o'no long vowel counterpart
(why not? why no -' after
all vowels but -oo, -I, -II?)

I don't know why Arakawa reconstructed both 1.39/2.35 and 1.40 as -ee'.

Sanskrit and Tibetan did not have final glottal stops and it is not clear whether the northwestern Chinese dialects of this period had them. Thus I can only test Arakawa's final glottal stops - if that is indeed what they are - indirectly.

If Tangut had final glottal stops in a given rhyme class, the transcriptions of syllables in that rhyme class should not have ended with the Tibetan letter -H which may have been a voiced glottal fricative [H]. (I do not know of any cases in which final glottal stops were transcribed as final fricatives. If such practices exist, then this test may be faulty.)

Here are the rhymes in Tibetan transcription from Nishida (1964) and Sofronov (1968) corresponding to Arakawa's -V(V)' rhymes:

Arakawa's short vowel rhymesArakawa's long vowel rhymes
1.5-u(H)1.7-yu, -u(H)
1.12-i, -e1.14-i, -e
(presumably 1.22?)no transcriptions known1.23-a
no level tone counterpart-arno long vowel counterpart
1.37-uH, -o, -e(H)1.39-e(H), -i
no short vowel counterpart1.40-e, -i
no rising tone counterpart
1.52-o (Sofronov 1968 II: 62)no long vowel counterpart

Assuming that the H-test is valid, at face value the above table suggests that Arakawa's glottal stops did not exist after -u(u) and short -e, and in some, if not all, rhymes with the vowels ee (1.37/2.34, 1.39/2.35) and oo (1.53/2.46).  The final -r of -ar for 2.20 suggests that rhyme did not end in a glottal stop either.  The 'corrected' distribution of glottal stop looks rather implausible:

Arakawa's short vowel rhymesArakawa's long vowel rhymes
no short vowel counterpart1.40-ee'
1.52/2.45-o'no long vowel counterpart

There is nothing about these vowels that make them more 'attractive' to glottal stops than u, uu, e, and oo. Moreover, if a rhyme has transcriptions which never end in -H, that does not tell us that the rhyme had a final glottal stop. It is also possible that the rhyme had no coda at all.  Here is how Gong reconstructed Arakawa's glottal-stop-final rhymes:

Rhyme classArakawaGongRhyme classArakawaGong

Notice that although both reconstructions have long vowels and -y-,

Arakawa's -(y)V' corresponds to Gong's -(i)VV(y)

Arakawa's -VV' corresponds to Gong's -yVV(y) (except in 1.40)

A short vowel followed by a glottal stop (Arakawa) sounds very different from a long vowel (Gong).  One could try to compromise and propose that Arakawa and Gong were correct for different dialects, but then one would have to ask which dialects were recorded in different sources. (I don't doubt that different Tangut books reflect different dialects. The question is whether dialects exhibited the V'VV correspondence pattern.)

It is difficult to test Arakawa's hypothesis with the Chinese transcriptions. If Tangut period NW Chinese had lost all final stops, then Tangut final glottal stop would correspond to Chinese zero.

I initially was going to propose that correspondences like

Arakawa -(V)V' : Chinese -Ṽ / -VN (N = nasal)

would constitute counterevidence. However, if the best match for a Tangut vowel was a Chinese nasal vowel or a Chinese vowel followed by a nasal, then 'Gule Maocai', the author of the Pearl, might have chosen to write -(V)V' rhymes with Chinese -Ṽ / VN rhymes: e.g.,

if the Chinese of that time and place had the rhymes -õ or -ong (but no -o)

and Tangut had -o'

Chn -õ or -ong might have represented Tangut -o' in spite of the coda mismatch

rGyalrong correspondences might provide better counterevidence. rGyalrong sonorant codas presumably correspond to pre-Tangut sonorant codas which would not have become glottal stops in Tangut itself:

rG -r, -m : pre-Tangut *-r, *-m > Tangut zero (but not -'!)

Here are the correspondences I found in Jacques (2003):

Rhyme classArakawarG of gDong-brgyadRhyme classArakawarG of gDong-brgyad
1.5/2.5-u'-U1.7/2.6-uu'no known cognates
1.12/2.11-i'no known cognates1.14/2.12-ii'-a, -i
1.22/2.19-a'-aR < *-aq1.23/2.21-aa'-as, -at
2.20-ya'no known cognates
1.37/2.34-e'no known cognates1.39/2.35-ee'-e, -i, -o
1.52/2.45-o'no known cognates
1.53/2.46-yo'-a (not really**)

Although Arakawa's final glottal stops are not yet known to correspond to gDong-brgyad sonorant codas, their correspondences with open vowels might imply that Tangut preserved glottal stops lost in rGyalrong: e.g.,

Proto-Qiangic *-ii'

> Tangut -ii'

> but gDong-brgyad rGyalrong -i

On the other hand, such correspondences would also be expected if Tangut never had any final glottal stops: e.g.,

Proto-Qiangic *-ii

> Tangut -ii

> gDong-brgyad rGyalrong -i

Arakawa's glottal stop coda corresponds to rGyalrong obstruents in only three cases after -a:

Arakawa -' : gDong-brgyad rGyalrong -R*-q, -t, -s

Although it is tempting to claim that a Tangut final glottal stop developed from earlier final obstruents -

Arakawa -' < Pre-Tangut *-q, *-t, *-s

- most rGyalrong final obstruents correspond to Arakawa's (and others') zero rather than his glottal stop. Why would Pre-Tangut final obstruents survive as glottal stops after -(a)a in only three cases*** but not in all others? And where would Arakawa's glottal stops after non-a vowels come from if they did not originate from earlier obstruents? They might be new default codas of old codaless rhymes, whereas open-vowel rhymes could have originated from earlier consonant-final rhymes:

*-(V)V > -(V)V'

*-(V)VC > -(V)V

Since open syllables are more common than closed syllables, this hypothesis would predict that there would be a large number of Tangut glottal stop-final syllables.  But in fact some of Arakawa's glottal stop-final rhyme classes contain relatively few tangraphs, which is why they have no transcriptional evidence or correspondences:

Rhyme classArakawaNumber of tangraphs in main body of PRTS (level tone + rising tone) not including 'misfilings'Rhyme classArakawaNumber of tangraphs in main body of PRTS (level tone + rising tone) not including 'misfilings'
1.5/2.5-u'22 + 151.7/2.6-uu'25 + 21
1.12/2.11-i'7 + 151.14/2.12-ii'31 + 27
1.22/2.19-a'18 + 81.23/2.21-aa'11 + 7
1.37/2.34-e'7 + 141.39/2.35-ee'32 + 16
1.52/2.45-o'11 +17
1.53/2.46-yo'16 + 8

(I'll explain what I mean by 'misfilings' in a future post.)

Although 18 (< 11 + 7) tangraphs in, for example, 1.23/2.21 may seem like a lot, the number is drastically reduced to 7 (< 6 + 1) if tangraphs representing homophonous syllables are subtracted:

1.23: phaa', baa', gaa', maa', daa', laa'

2.21: gaa' (just one syllable!)

(These reconstructions combine Gong's initials with Arakawa's finals. Gong's and Arakawa's initials should be identical in these cases.)

Gong also reconstructed syllables with dz- in 1.23 and m-, n-, and j- in 2.21.

According to my glottal stop as new default coda hypothesis, these syllables would have come from Pre-Tangut phaa, baa, maa, daa, naa, laa, dzaa, jaa, gaa (ignoring tones).

It is difficult for me to believe that Pre-Tangut lacked paa, taa, tsaa, chaa, kaa, etc.  I suspect that 1.23/2.21 was a more complex rhyme than a simple -aa or -aa' which almost always occurred after voiced initials.  (The sole exception is TT4651 CUT-OFF with ph-.) 1.23/2.21 and another rhyme might have formed a single class in Pre-Tangut which split into two because of different vocalic developments after voiced and voiceless initials: e.g.,

Pre-Tangut p- + a-like rhyme X > Tangut p- + rhyme other than 1.23/2.21

Pre-Tangut b- + a-like rhyme X > Tangut b- + rhyme 1.23/2.21

Cf. the vocalic split in Khmer:

Old Khmer paa > modern Khmer paa

Old Khmer baa > modern Khmer piə

However, unlike Khmer, Tangut apparently did not devoice initial voiced obstruents, judging from these correspondences (examples from Gong 1995: 55 added 3.25.0:00):

TT5764 BEE borw  1.91 :

Written Tibetan bung 'bee'

(but Old Chinese had both 蜂 *phong and *bong < ?*m-phong for 'bee')

TT1856 QUARREL dze 1.8, TT5509 TO-QUARREL dzeey 1.37 :

Written Tibetan Hdzing 'to quarrel', zing-chha 'quarrel'

(but Old Chinese 爭 *rtseng 'quarrel' and Written Burmese chach < *tsik  'war' have ts- instead of dz- or z-!)

If Tangut had devoiced initial voiced obstruents, Written Tibetan b- and (d)z- would correspond to Tangut p(h)- and ts(h)- instead of Tangut b- and dz-.

The case of 1.12/2.11 (Arakawa's -i') is even more extreme. It only occurs after labial, labiodental, dental, and liquid initials.  (Nishida 1964: 45 also reports it after glottals, though I cannot confirm this.) Why wouldn't an -i' occur after palatals or velars?  Is this gap accidental or systematic? I suspect that 1.12/2.11 was some sort of 'anti-back consonant' rhyme instead of a simple -i'. Perhaps it had a velar glide -W- (IPA [ɰ]) that couldn't occur after palatal and velar initials. (It is not obvious to me why Nishida's -wiH, Sofronov's -eC,  Li Fanwen's -iwi, and Gong's -ee could not occur after ch-, k-, etc.)

(3.25.0:05: Huang Zhenhua [1983: 125] reconstructed 1.12 as -ïr with the vowel counterpart of my -W- [ɰ]. He did not provide reconstructions for rising tone rhymes.)

I suspect that all of Arakawa's -(V)V' were in fact complex rhymes ending in vowel or vowel/glide sequences originating from both *-V and *-VC rhymes.

Next: Uh-öh.

*I wonder if Arakawa's vowel + -' symbols are like Vietnamese ư [ï] and ơ [əə]: i.e., representing vowels distinct from those written without the dấu móc rather than vowels followed by glottal stops.  If the apostrophes are simply diacritics meaning 'vowel like ...', then much of this post should be ignored.

**gDong-brgyad rGyalrong kë-ncha 'kill an animal' was listed with two Tangut cognates, TT5016 BUTCHER shii' 1.14 and TT0530 宰殺 SLAUGHTER-ANIMAL shoo' 1.53. (The reconstructions are hybrids of Gong initials and Arakawa finals.)

The latter does not really indicate that rGyalrong -a corresponds to Tangut -oo' because it might be a suffixed form of the former.  This is easy to account for in Gong's reconstruction:


Using Arakawa's finals, it would be difficult to explain how the glottal stop of BUTCHER 'skipped over' the suffix -oo to be at the end of SLAUGHTER-ANIMAL -

BUTCHER shii' + -oo = SLAUGHTER-ANIMAL shoo' (not sh'oo or shoosh'oo)

- unless this were a case of vowel alternation rather than -oo-suffixation:

root sh-' > BUTCHER shii', SLAUGHTER-ANIMAL shoo'

***gDong-brgyad rGyalrong -R*-q, -t, -s usually correspond to Arakawa's zero, -k, or -k' rather than his glottal stop.  See the charts in part 7. WHAT ANIMAL HAS A TEMPERAMENTAL BELLY?

I briefly wondered if

TT2901 (second half of CAMEL) dyï 2.28


TT2955 CAMEL? kyaa 1.21

from "Neither Old nor Distinct" were influenced by the second graph of Chn 駱駝 'camel' (Md luotuo). Although there is a vague similarity between the right-hand elements of TT2901 and TT2955 and the phonetic element 它, the tangraphs do not have any equivalent of the semantic element 馬 'horse' on the left.  I would expect something like

TT5258 耐 ENDURE nya (tone unknown)

which does have the Tangut version of 馬 'horse'.  (But what do horses have to do with endurance?)

The left-hand element of TT2901 was identified by Nishida (1966: 244) as 腹 BELLY (a reference to a camel's hump?).  So does TT2901 represent 'something with a [prominent] belly with a Chinese name written with它'?  That would not explain the vertical line separating BELLY from the 它-like element.  (I have no idea what the vertical line is doing in any tangraph.)

The left-hand element of TT2955 is the right-hand element of

TT4820 等 PLURAL nyï 2.28

See "The Plural Puzzle".

If tangraphy were semantocentric, one might expect all tangraphs for CAMEL to share common elements representing their shared meaning. Initially I wondered ifthe 它-like element were a semantic symbol for 'camel' influenced by Chn 駝 representing the second syllable of 'camel', but it is absent from five other CAMEL tangraphs:

TT5280 2901

lãã 2.14 dyï 2.28 'camel'

TT1171 5519

myïï 2.28 chyu 1.2 'camel'

TT1171 shares the vertical line and right-hand element with TT5280

TT5519 was analyzed as TT2901 + TT2900; i.e., as CAMEL + CAMEL

the element on either side of TT5519 was identified as TEMPERAMENT by Kychanov (Grinstead 1972: 15)

but what's the inverted 六 'six' in the middle doing?

TT2839 活 LIVE xwo 2.42 (borrowing from Chn)

a phonetic character here

TT2900 CAMEL tha 1.17

there's BELLY again on the left (cf. TT2901 CAMEL)

and TEMPERAMENT on the right (cf. TT5519 CAMEL)

surrounding another inverted 六 'six' (cf. TT5519 CAMEL)

was xwotha 'camel' a borrowing from a NW Chn dialect equivalent of Middle Chinese lakda* < a hypothetical Old Chinese *kəlak lay 'camel'?; but why initial x- instead of l-**? Do modern NW dialects have an x-word for 'camel'?

Why did Tangut have so many CAMEL characters?  Were camels important to the Tangut? Are these synonyms or terms for specific types of camels? Is one or more of these words from the 'ritual language' which reportedly had mostly disyllabic nouns?

None of these terms appear to be loans from Tibetan terms for 'camel' with the root rnga (< 'drum'?) or Written Tibetan ku ma naH pa (presumably a loan, but from what? - not Sanskrit or Hindi).  They also have nothing to do with Uighur töge 'camel' or Old Chinese 橐駝 thak lay 'camel' (used in 史記 Shiji [Pulleyblank 1962: 245; *thak is 'sack' and has the phonetic element 石 *dak, so it cannot be related to the first syllable of 駱駝 *kəlak lay 'camel'; *d- and *l- are not interchangeable in phonetic series).

( The second syllable 駝 lay may be the root 馱 lay 'carry on the back'. Hence a 橐駝 *thak lay 'camel' might have been a 'sack-carrier'.).

And is

TT2955 kyaa 1.21

CAMEL at all?  Nishida (1966: 480) defined it as a transcription character (which doesn't necessarily meant that it was meaningless).  Grinstead (1972: 188) listed it as a transcription character for (Sanskrit?) ga. Li Fanwen (1986: 332) defined it as 根 ROOT. Shi et al. (1983: 436 similarly defined it as 根腳 FOUNDATION, lit. 'root-foot'.)

*Middle Chinese *lakda 'camel' is the source of Vietnamese lạc đà 'camel', Korean 낙타 naktha 'camel', and Japanese rakuda 'camel'.

As far as I know, Korean 약대 yaktE 'camel' has no Chinese characters associated with it, but it might be from a hypothetical Late Old Chinese *lïakdai 'camel'.  (Korean could not allow l- initially, so it was dropped.)

**Was the *x-variant of 'camel' from a hypothetical Old Chinese variant *hrak lay, which would have become Middle Chinese *xak da and later something like *xO thO in the northwest?

Middle Chinese *x- ~ *l- < Old Chinese *hr- ~ *r- alternation can be found in 虎 MC *xo' < OC *hra' 'tiger' which shares a phonetic component 虍 with MC l-graphs such as the surname 盧 MC *lo < OC *ra. NEITHER OLD NOR DISTINCT

refers to the Tangraphic Sea's analysis of

TT1465 NEW syiw 1.46

the first tangraph in the original title of Le nouveau recueil sur l’amour parental et la piété filiale:


left of TT1491 NOT myi 1.11

right of TT3846 OLD kywi 1.11

right of TT3819 DISCRIMINATE / 分明 DISTINCT sywiy 1.36

刂 NOT is a well-attested semantic element.

One might think that the right side of OLD means OLD, but Nishida (1966: 244) glossed it as the radical having to do with 傾く 'incline' (his #230), and it appears in tangraphs having nothing (obvious) to do with OLD such as

TT4711 WRITE ryar 1.82

TT4720 PESTLE tsorw 1.91

TT4723 PLEASED lyïy 2.55


TT4727 GOOD ngạ 2.56 (well, yes, there is Eng good old)

TT4729 CHAPTER tyiy 1.36

with INCLINED doubled around a vertical line

TT4731 CRAWL byu 1.2

TT4734 FOREHEAD lyạ 1.64

TT4735 ARTERY zorw 1.91 (phonetically close to PESTLE above)

TT4736 HAVE-LEISURE Gyã 1.25

TT4742 THIGH dyụ 1.59

Nishida's radical 176 昇る 'climb' on right

TT4743 DO-NOT tyi 1.11

HAND on the right

TT4744 羞愧 ASHAMED bow 2.47

TT4745 STIRRUP mya 1.20

TT4746 CURTAIN lyạ 1.64


as well as tangraphs which might have something to do with OLD: e.g.,

TT1319 BEFORE ma 2.14

Nishida's radical 138 以前 BEFORE on left

TT2012 AGE/TEETH shywi 1.10

'horned hat' atop following tangraph

TT3548 YEAR kyiw 1.45

TT4713 TRACE kywi 1.11 (homophonous, poss. cogn. with OLD?)

TT4724 TIME zyịy 1.61

TT4730 TRANSFORM dyi 2.12

Nishida's radical 230 does not appear in other OLD-tangraphs:

TT0224 OLD-MAN 'o 1.49


TT3839 ANCIENT dzyo 2.44

PERSON plus next tangraph

TT4612 OLD dzyụ 1.59

WORDS on left

Nishida's radical 138 以前 BEFORE on right

same root as TT3839?

TT4906 OLD wə̣ 1.68

TT5437 GROW-OLD shywïí 2.32

TT5438 OLD nar 2.73

TT5699 OLD wyị̈ 1.6

The Tangraphic Sea analyzed TT3846 OLD kywi 1.11 as

< +

bottom [left] of TT4559 舊 OLD lhywi 2.10 +

what is the function of 그?

center of TT2779 安/置 TO-PLACE tyị 1.67

with HAND on left + INCLINE + unknown but common right-hand element

One could claim that

'not incline' > 'not bent over from old age' > 'not old' > 'new'

but that seems like a stretch to me, and it still doesn't account for the right side of DISTINCT (last seen in the analysis of ONE), unless one regards NEW as


rather than as


Where does the right side of TT3819 DISCRIMINATE / 分明 DISTINCT sywiy 1.36 come from? Here's its Tangraphic Sea analysis:

< +

left of TT3823 DARKNESS kow 1.54 +

right of TT2246 照 ILLUMINATE swew 1.43

And the Tangraphic Sea tells us that the right of ILLUMINATE comes from ... the right of DISTINCT.

That right-hand element is also in non-light tangraphs such as

TT0141 SISTER nyiy 2.33

TT1283 FOX tyi 1.11


NOT + TT5087 below

TT2901 (second half of CAMEL) dyï 2.28

TT2955 CAMEL? kyaa 1.21


unknown top element + TT5087 below

this is the Categories of Mixed Categories of the Tangraphic Sea

TT4608 FEW nyiy 2.33

WORDS on left

TT5087 SAME ləw 2.38 (part of the title of Homophones)

Next: A sywiy-mpler explanation?

(Question added 20:36.  Posted along with above post so that the following L&P post could be on top all day.) SHIOO1 NYIY21 LA1 MYIIY2 L...?

is Gong's reconstruction of "the title of the Tangut book that made my year". (Yes, I know there are still nine months left, but it'll be hard to top this.)  The tangraphs are

TT1465 4482 4652 4095 2746


TT0288 2769

TAIL (also 下/後/末/終 LAST / END [Nevsky 1960 II.598]) SCROLL*

This book has been translated into French as Le nouveau recueil sur l’amour parental et la piété filiale by - who else? -

TT5559 0834 2968


xyow1 zhyiw1 bo1

a.k.a. Guillaume Jacques** (official site). I think it's his first book (and the first Tangutological book in French in about a century). Congratulations!  This is the first of a series, despite its Tangut title.  May there be many more volumes of Textes tangoutes!

David Boxenhorn (who could use a Tangut name) guessed that was the book I was referring to as "L&P" (for Love and Piety). I wish I could give him a copy as a prize, but alas ...

Lincom Europa is selling L&P online for 64 Euros or $76.80 US. I don't want to hurt sales, so I can't give away everything on this blog. But I can outline what Guillaume's edition of L&P consists of:

- a short but sweet seven-page introduction on

- the Tangut language

- tangraphy

- the L&P text

- the Tangut verbal system and its

- directional prefixes

- agreement suffixes

- vocalic alternations

(all topics for future posts; I know I've been neglecting Tangut grammar, but I think the structure of tangraphy is even more neglected, and it is my main interest, so ...)

- a bibliography

- the L&P text itself

- in tangraphy (5,919 tangraphs total)

- in Li Fanwen / Mojikyo numbers

- in Gong's reconstruction

- with Chinese glosses of each tangraph whenever possible

- and a French translation

- plus Chinese originals of the stories retold in L&P

- a tangraphic index of all 1,054 different tangraphs (about one-sixth of the total number in existence) in L&P

- in Li Fanwen / Mojikyo number order

- in Gong's reconstruction

- with Chinese and French glosses

- and citations

If I may coin an English-Chinese-Tangut calque, I'd call L&P a

TT5486 0226

niaa 2.20 kiẹ 1.66

MINE GOLD 'gold mine'***.

It's a shame that the first volume didn't survive. But this has enough to keep me occupied for a very long time. Many thanks to Guillaume for bringing this work to light. I'd like to write more about it after I tie up some loose ends from previous posts.

*The reading of TT2769 SCROLL is obscure. It is an isolated tangraph in the liquid chapter of Homophones (54B31). It is used in the fanqie for an l-initial tangraph (TT4971 揉 RUB lyụ 1.59) so I presume it had initial l-. Its rhyme is unknown. I cannot find any tangraphs with the same right side with an initial l- which might have been near-homophones.

**The tangraphs roughly correspond to Guillaume's Chinese name 向柏霖  (Md Xiang Bolin).

PERFUME xyow1 is a Tangut approximation of the Tangut period northwestern Chinese pronunciation of 向, a near-homophone of 香 'perfume' (the source of Tangut PERFUME).

Tangut CYPRESS WOODS zhyiw1 bo1 corresponds to 柏 'cypress' and the phonetic 林 'woods' of 霖 'extended rain' (which has 雨 'rain' on top).

***The Homophones contains a phrase

TT5486 2599

niaa 2.20 shyow 1.56

MINE IRON = 'iron mine'

on 14B33. (IRON is used as a right subscript clarifier for MINE.)

'Gold' and 'iron' are adjectives in the phrases MINE GOLD and MINE IRON and hence follow the nouns that they modify.  Perhaps more accurate - but also more awkward - translations would be 'golden mine' and 'ferrous mine'. THE PLURAL PUZZLE

Last night, I tried to make sense out of the tangraphic descendants of

TT2501 獨/一 ALONE/ONE lheew 2.41

Tonight, I'll look at the graphic offspring of

TT4820 等 PLURAL nyï 2.28

which appears in such phrases as

TT1999 0245 4820

wya 1.19 mya 1.20 nyï 2.28

FATHER MOTHER PLURAL = 'parents' (L&P* 3.8, 4.3)

TT0589 4820

zyị̈ 2.61 nyï 2.28

BOTH PLURAL = 'the two' [women] (L&P 5.3)

TT4130 0369 4130 4171 4820 0589 1649

biaa 2.20 gyi 1.16 biaa 2.20 tụ̃ 1.96 nyï 2.28 zyị̈ 2.61 myï 2.28


= '[his] two [nephews], 馬嚴 Ma Yan and 馬敦 Ma Dun' (L&P 6.5)

TT0591 4820

nyi 2.20 nyï 2.28

YOU PLURAL = 'you (plural)' (L&P 6.6, 7.3, 7.4, 7.7)

TT3811 4105 4820

dzey 1.33 dzywo 2.44 nyï 2.28

PRIVATE PERSON PLURAL = '[his] own people' (L&P 16.5)

(despite Grinstead's gloss, PRIVATE does not follow like a normal adjective, so I suspect it's a noun meaning 'one's own'.)

PLURAL appears to be an abstract suffix. How can such a thing be drawn?

The Mandarin solution is 們, a combination of the semantic element 亻 'person' plus the phonetic element 門 Md men 'gate' which is nearly homophonous with the Md plural suffix -men. (The suffix is atonal, whereas 'gate' has a high rising tone.)

Li Fanwen and Guillaume Jacques gloss PLURAL in Chinese as 等, a combination of the semantic element 竹 'bamboo' and the phonetic element 寺 Md si 'temple'. Neither seem appropriate for Md deng 'etc.', but in Old Chinese, 等 originally meant 'step of a stair' (made of 竹 bamboo?) and came to mean 'degree', 'rank', and 'class' by extension. 'etc.' refers to the members of a class. 等 was pronounced something like təng' which at least shared a stop initial and a schwa vowel with 寺 OC sdəs 'eunuch'*** (later 'temple'), unlike modern deng and si.

But what elements combined to form the tangraph PLURAL?

Nishida (1966: 365, 480) did not gloss either of its components. (They are his 59th and 246th 'radicals'.)

Its left side appears in that position in only eight other tangraphs which have no single phonetic or semantic common denominator:

TT4817 羞愧 ASHAMED lyiy 1.36 (Nishida: 密集 'swarm'?)

all but the right side is said to be from

TT4815 NOON khywï 1.30

whose left-hand component has one less stroke

and whose right-hand component has one more stroke!

TT4818 羞愧 ASHAMED lyiy 2.33

probably the same root as TT4817 羞愧 ASHAMED lyiy 1.36

but which is the base and which is derived?

and what do the different right sides signify?

TT4819 LION ka 2.14

TT4821 (see below) Gyow 1.56

TT4822 MAESTRO lụ 1.58 (Nishida: 君子 'gentleman')

TT4823 (transcription graph) byụ 1.59

analyzed as left of TT4824 below

+ left of TT5502 pyu 2.3

(syllable of a name, written as FIRE + PERSON)

TT4824 (a wild animal) byị̈ 1.69

WAIST/BIRD is on the right

TT4825 RUN lwəy 1.40

I cannot find that element on the right side of any tangraph in Grinstead's (1972) index.

The right side of PLURAL is somewhat more common. Here are glosses and Gong's reconstructions for some tangraphs containing it

on the right side (a partial list from Grinstead 1972: 127-128):


TT1278 EXCELLENT tyiy 2.33

TT1745 WEAKEN dzəy 1.40

TT1890 BEAUTIFUL zerw 1.87

TT1937 STRIPED phə 2.25

TT1890 BEAUTIFUL with a 'horned hat' on top

TT2506 DRIP gyirw 2.79 (from last night)

TT2525 SHORT wyịy 1.61

TT3667 RISE (of sun) to 2.42

TT3671 REACH nyï 2.28

TT4632 INFORM nyï 2.28

on the left side (a complete list?; see Nishida 1966: 480-481):

TT2952 耀く SHINE nywï 2.28

TT2953 (transcription graph) nụ 1.58

to transcribe what? Tibetan, Chinese, and Sanskrit didn't have tense vowels like - was a non-tense vowel perceived to be like Tangut ụ, and if so, why?

TT2954 SKILL thu 1.1

TT2955 CAMEL kyaa 1.21

TT2956 (transcription graph) thwey 1.33

TT2957 MOTHER-IN-LAW nyï .30

TT2958 祖父母 GRANDPARENTS (on mother's side) ba 2.14

TT2959 (see below) nyï 1.30

TT2960 ESTABLISH thu 1.1

TT2961 (surname) thẽ 2.13

TT2962 (Sanskrit transcription graph) kyaa 1.21

TT2963 惜しむ HOLD-DEAR? ba 2.14

Some are (nearly) homophonous with PLURAL, so perhaps the right side of PLURAL is a phonetic symbol for nyï (and a distortion of Chn 女 'woman', pronounced *ñji in pre-Tangut period NW Chn?) But then what is its left side?

No Tangraphic Sea analysis has survived for PLURAL, but the extant analyses of other tangraphs do contain it. Can you guess the meanings of these four tangraphs containing parts of PLURAL?


TT2959 ? nyï 1.30 <

right of TT4820 等 PLURAL nyï 2.28 (phonetic) +

bottom [left] of TT2030 MOTHER-IN-LAW wyï 1.29

i.e., PERSON, which is in hundreds of other tangraphs


TT2649 ? chyow 1.56 (the mirror image of TT2959!) <

left of TT3890 SURNAME/CLAN 2.25 +

right of TT4820 等 PLURAL nyï 2.28


TT4821 ? Gyow 1.56 <

right of TT4820 等 PLURAL nyï 2.28 +

right of TT4477 PILLOW wọ 1.70


TT4573 ? dyụ 1.59 <

left of TT4566 SPRING / 疲れ FATIGUE dyụ 1.59 (phonetic) +

right of TT4820 等 PLURAL nyï 2.28

Scroll over the blank lines below to see the answers.


TT2649 = (a surname)

the right-hand element PLURAL might have been meant to evoke Chn 眾 'crowd' (pre-Tangut period NW Chn *chung?) which sounded like chyow 1.56

so the graph could be 'the surname that sounds like [Chn] CROWD'

TT4821 = SOURCE (why?)

TT4573 = CROSSROADS (why?)

*L&P is short for the title of the Tangut book that made my year. Details next time.


TT0589 BOTH zyị̈ 2.61

looks like a combination of


TT0590 TWO nyïï 1.32 +

TT0006 ALL zyi 2.1

ALL is phonetic and possibly even semantic in BOTH.

Was BOTH zyị̈ 2.61 borrowed from pre-Tangut period NW Chn 二 *zhi 'two'? The mismatch in initials makes me skeptical.

***Is 寺 OC sdəs 'eunuch' from s-N-təng'-s 'those who wait' with a root təng' 'wait' (written with the same graph as its homophone 等 'stairs') also found in 待 də' < N-təng' 'wait'? Could the initial s- and/or final -s be deriving a noun from the root 'wait'? (But what would the function of N- be?) FROM SEVEN TO EIGHT TO ONE

Instead of part 9 of "Riddle of the Rhymes", I'm going to post something I planned on the 11th.

In the Tangraphic Sea, EIGHT is analyzed as SEVEN minus its 'head':


TT1632 EIGHT 'yar 1.82

< all of TT1986 SEVEN shyạ 1.64 minus its head

The function of the 'head' of SEVEN is unknown. It occurs in no other tangraph. Although the top three strokes look like the common element that I've called the 'horned hat' (Nishida's 7th 'radical'), I have never seen the bottom half of the 'head' (从) in any other tangraphs.

Why would SEVEN be written as EIGHT with an additional element (or two)? Does this suggest that 'seven' and 'eight' rhymed? They almost rhyme in various reconstructions (e.g., Gong's -yạ : -yar), so for years I wondered if the analysis could be read as

EIGHT'yar is SEVEN shyạ without its sh-

However, if the 'head' of SEVEN were a phonogram for sh-, why wasn't it used in tangraphs for other sh-initial syllables?

If tangraphy was semantocentric, the structure of SEVEN suggests some cultural association between SEVEN and EIGHT that I'm not seeing.

But SEVEN is not EIGHT's only graphic relative. EIGHT forms the bottom of

TT2501 獨/一 ALONE/ONE* lheew 2.41

which was analyzed as


left of TT2506 DRIP gyirw 2.79 +\

bottom [left] of TT2507 ONE/ALONE tyịy 1.61

which obviously contain it.

What does EIGHT have to do with being ALONE? It's eight times ONE! And why is ALONE on the left side of DRIP or the surname

TT2505 zhyịy 2.54

which doesn't sound like ALONE? (Was that clan associated with isolation?)

TT2505 forms part of the title of the ninth chapter of Homophones. It is the Tangut counterpart of the Chinese phonological term 日 (zhir in pre-Tangut period northwestern Chinese) representing r- and zh-like initials. (The idea from this post originated from my reference to this tangraph all the way back in "The Tibetan Test".)

The functions of the right-hand components of TT2506 DRIP and TT2505 ALONE are unknown.

TT2501 ALONE lheew 2.41 might be phonetic in

TT2502 CARVE lyuu 1.7

said to be from


left of TT2507 ONE/ALONE tyịy 1.61 +

center of TT3981 DIG luu 1.5

(which lacks the horizontal line atop the right side of CARVE!)

ALONE might be semantic in

TT2503 完全 COMPLETE Gwẽ 1.15

analyzed as


left of TT2507 ONE/ALONE tyịy 1.61 +

right of TT0457 COMPLETE ngorw 2.82**

i.e., 'to be as one', 'to be whole'?

Notice that TT0457 COMPLETE has an EIGHT-like element on its left side! No analysis of TT0457 COMPLETE is avaialbe, but that left-hand element looks like the surname

TT0452 guu 2.5

(which might be phonetic in TT0457 COMPLETE ngorw 2.82?)

Is that surname just an arbitrary set of lines that happen to contain EIGHT beneath a horizontal line, or does it have some semantic (or even Tangut B phonetic?) relationship to EIGHT?

ALONE must be semantic in

TT2504 獨 ALONE tywar 1.82

Nishida (1966: 245) identified the right-hand element as SURPASS, his 241st 'radical'. (Was it based on 走, the left-hand element of 越 'surpass'?) So was this 'very lonely'? Its Tangraphic Sea analysis derives the right side from

TT4475 CROSS dzyịy 1.61

and 'to cross' is yet another meaning of 越 'surpass'.

Lest you think that ALONE was the semantic element for 'alone', it doesn't even appear in

TT3728 獨自 BY-ONESELF kywị̈y 2.55

which consists of PERSON plus what might be the right side of

TT2507 ONE/ALONE tyịy 1.61

or tangraphs such as (Grinstead 1972: 111)

TT1721 ENTRUST neey 2.34

TT2263 EXCEED gyiiy 1.39

TT2424 SKILLFUL khioo 2.43

TT2748 ADORN tyạ 1.68

(HAND on left)

TT3169 FAULT ngyạ 2.57

(EVIL on left)

TT3618 CALCULATE ngerw 1.87

(PERSON on left)

TT3753 億 ONE-HUNDRED-MILLION ryir 2.72

(with a mere two PEOPLE on the left)

(practically the opposite of ALONE!)

TT3838 ABLE nywi 2.10

(PERSON on left)

TT5298 GOVERN myii 2.12

TT5662 THOUGHT syiiy 2.35

(with BIRD or WAIST on the left!)

How would anyone know that

TT3728 獨自 BY-ONESELF kywị̈y 2.55

did not mean, say,











instead of BY-ONESELF?

Next: A Tangut B-based explanation for this tangraphic family.

And PLURAL, the polar opposite of ONE! ADDENDUM: I thought this post covered every tangraph containing

TT2501 獨/一 ALONE/ONE lheew 2.41

i.e., TT2502-2507:

but I found one tangraph with TT2501 in the middle

TT0514 tsew 2.38

which looks like MOUTH + ONE/ALONE. Of course, it means ... 齋 FAST!? (Fortunately, this is FAST in the sense of 'abstain from food', not 'speedy', though the latter wouldn't have surprised me!)

* I just realized that

TT2501 獨/一 ALONE/ONE lheew 2.41

TT1075 ONE lew 1.43

TT3603 ONE/ONLY lew 1.43

are probably cognates. lew is the bare root 'one' spelled two different ways. lheew might be from *s-l-ee-w with a prefix and vowel lengthening.

Notice that the two tangraphs for ONE have nothing in common but PERSON (their left elements). The second tangraph has a Tangraphic Sea analysis which makes no semantic sense:


TT3603 lew ONE/ONLY 1.43 <

left of TT3819 DISCRIMINATE sywiy 1.36 +

(how would anyone know that PERSON was from this particular tangraph and not from the 604 other tangraphs with PERSON on the left, or other tangraphs with PERSON in other positions?)

left of TT2146 (place name) thyï 1.30 +

(not an exact match for the middle element of TT3603 ONE/ONLY)

left of TT5607 EXPLAIN/CLEAR jyu 1.2

Is Tangut B the ONLY way out of this graphic dilemma?


TT0457 COMPLETE ngorw 2.82 and its homophone

TT0004 ALL ngorw 2.82

are probably the same word with two different spellings.

The structure of TT0004 is not known to me, but its resemblance to

TT0006 ALL zyi 2.1

is probably not coincidental. What might the different lower right components signify? Neither appears independently or has any known meaning. RIDDLE OF THE RHYMES (PART 8)

In part 7, I laid out arguments against a final -k in Tangut.

The transcriptive evidence for -k is nonexistent and the comparative evidence is extremely weak. The best case involves Arakawa's -ak which corresponds to gDong-brgyad rGyalrong (DG) -aq (< Proto-rGyalrong *-aq, *-ɐq, *-Oq [Jacques 2004: 266]) in eight cases but also corresponds once to

DG -aV (< Proto-rGyalrong *-ap)

and once to

DG -ër (< Proto-rGyalrong *-ɐr, *-Or)

This suggests that Arakawa's -ak might have been an -a which resulted from the merger of two or more earlier rhymes: *-aq?, *-ap?, *-a- + other consonants other than -r?, *-a? (Pre-Tangut *-ar would have become Tangut -ar with a retroflex vowel.)

Although Tangut must have had a -k at some time during its history, that period was prior to the compilation of the Tangraphic Sea. (It is, of course, possible that conservative Tangut dialects preserved one or more final stops, but there are no records indicating that such dialects existed.)

Pre-Tangut must also have had a final -ng. The Tibetan transcriptions do contain a final -ng and the Chinese transcription material is full of sinographs that either ended in -ng or in nasal vowels. This suggests that Arakawa's -ng might be more viable than its non-nasal counterpart -k.

Arakawa reconstructed -ng in four rhyme classes in the material I have on hand:

Arakawa rhymeTangut rhyme numberTibetan transcriptions from Nishida (1964) and Sofronov (1968 II)Chinese transcriptions from Nishida (1964) and Sofronov (1968 II); diacritics removed
non-nasal finalnasal-final
-eng1.61-e(H), -ye, -(y)i危 蛇; *-k: [移+責*]命名永榮迎經貞成城令[力+頃]領瓶便
-yeng (some 2.55 have -yeng2)1.62none奈碎靴; *-k: 局洫?兵丙丁頂甯兄
2.55-eH, -eng
(and 2.63?)nonenone?
-yong1.72-o多; *-k: [泥+托?]亡蹤象張浪[泥 +浪][移將?]

Apart from a single -eng, the Tibetan material does not support Arakawa's -ng at all in these rhymes. All other instances of -ng in the Tibetan transcriptions in Nevsky (1926) do not correspond to his -ng:

Arakawa rhymeTangut rhyme numberTangut TelecodeGrinstead's (1972) glossTibetan transcriptionsNevsky (1926) #
-yen [yẽ]?1.425676TIMEHdzeH, dzeng187
2.371666CORNERHdzeH, dzeng, Htse (sic)268
-on [õ]?2.470944TAKEgzong148
-yon [yõ]?2.485718CROWyong184
-yeng(2) 2.553597MIRRORgteng313
-ik'1.683797COLORrtsi, rtsiH, rtsing132
-ok1.701718THREEgso, gsoH, gswong269

In half of these cases, one could defend Arakawa and say that Tibetan -ng corresponded to his nasalization in Tangut vowels (if that is what his -n signifies). Yet it would still be curious that Tibetan -ng- almost never corresponds to his -ng!

The Chinese material is ambiguous. None of the Tangut rhymes correspond exclusively to Chinese rhymes with or without nasals. Was the Tangut author of the Pearl incapable of distinguishing between Chinese oral and nasal vowels because his own language lacked nasal vowels? If his Tangut dialect had nasal vowels, he would surely tend to transcribe those Tangut nasal vowels with Chinese nasal vowels (or Chinese vowel-nasal sequences).

It seems that if Tangut had an -ng, it was in rhymes other than those which Arakawa reconstructed with -ng.

Next: Did Tangut have a final glottal stop? Does Arakawa's -k correspond to rGyalrong -ng?

Although gDong-brgyad rGyalrong does have a final -ng (Jacques 2004: 205), this does not correspond to Proto-rGyalrong *-ng. PG *-ng occurred only in the PG rhyme *-ang which became -o in gDong-brgyad rGyalrong (Jacques 2004: 266). This -o never corresponds to any of Arakawa's -ng rhymes.

What do Arakawa's -ng rhymes correspond to?

Arakawa rhymeTangut rhyme numberCorresponding gDong-brgyad rGyalrong rhyme
(vowel other than -o < *-ang)-m
-eng1.61-e x 3, -i x 1-Um x 1
2.54no data
-yeng (some 2.55 have -yeng2)1.62
2.55-Um x 1
-ong1.71no data
-ong?2.63-om x 1
-yong1.72-a x 2, -i x 1
2.64no data

If anything, some of the rGyalrong rhymes seem to hint at a pre-Tangut *-m rather than a pre-Tangut *-ng. RIDDLE OF THE RHYMES (PART 7)

(3.19.2:36: Added gDong-brgyad rGyalrong comparative tables and commentary.)

荒川慎太郎 Arakawa Shintarou is the only Tangutologist whose reconstruction* contains

- a final -ng

- a final -k

- a final -k' (What is this phonetically? A palatalized k? It can't be a k-glottal stop cluster. Arakawa's reconstruction has an initial q, so presumably k' is a different sound.)

- what appears to be a final glottal stop (-')

- a final -" (What is this phonetically?)

in addition to

-n which I presume indicates nasalization, since it has no stop counterpart -t

but then again, Tangut could be like standard Mandarin which has -n but no -t

but then again, standard Mandarin has -ng but no -k - or any final stops at all, whereas Arakawa's Tangut has two orthographic final nasals -ng and -n and three orthographic final stops -k, -k', -' (and maybe even -"?)

-r which might indicate a retroflex vowel rather than an actual coda -r

Judging from these one thousand sample tangraph readings in Arakawa's reconstruction, these final consonants (if that's what all of them are) have an unusual distribution:

-n (Ṽ?)-an-aan-in-In-en-on-oon
-r (Vr?)-ar-aar-ir-iir-Ir-IIr-ur-er-or-oor
-r' (Vr'?)-aar'

1. I would expect a, the least marked vowel of all, to occur with the largest number of codas. Yet neither -a nor -aa occur with -ng, -k', or -". In fact, the vowel that does occur with every coda is o.

2. Why does -ng only appear after the short mid vowels e and o?

3. Why do u and uu never occur with any nasal element (-ng or -n?)

4. Why does -k' only occur with i(i), e, and o, which have nothing in common apart from being nonlow? (Although i and e are front vowels, o is a back vowel.)

5. Why does -" only occur with o in rhyme 1.57? (If o" represents [ö], then this vowel occurs in only one rhyme in the entire language. This is possible but unlikely.)

6. Why does -r' only occur after aa? (If r represents retroflexion, why does a glottal stop follow only one retroflex vowel?)

Arakawa's codas pose other problems:

7. There is little or no transcriptional evidence for them.

a. I know of no Tibetan transcriptions with final -g (the closest Tibetan equivalent to -k). One could argue that -H represented a lenited descendant of Arakawa's -k(') (perhaps [x]?), but -H also appears in transcriptions of rhymes other than those reconstructed by Arakawa with final -k(').

b. It is not clear whether the northwestern Chinese of this period still had an 'entering tone' (i.e., final stops).

If NWC had an 'entering tone', it did not correlate to Arakawa's final stops in the Pearl. Details here.**

If NWC lacked an 'entering tone', some other device (e.g., [subscript] kV sinographs) could have been used to indicate final -k('), but no such sinographs exist in the Pearl. Cf. the use of 克 ke to indicate a final -k in modern Mandarin: e.g., 巴拉克 Balake 'Barack (Obama)'.

c. I am not aware of any use of the -k(') tangraphs to transcribe Sanskrit syllables ending in -k.

One could argue that all transcriptions date after the loss of final stops. If so, then what evidence is there for final stops?

Comparative evidence indicates that Tangut must have had final stops at some point. Some of its relatives have final stops, and it is most likely that Tangut lost them. (The alternative - that they all independently 'grew' final stops in the same words - is highly unlikely.) The question is when Tangut lost them - before the Tangraphic Sea period or after it. In either case, the comparative evidence is sometimes at odds with Arakawa's reconstruction (points 7 and 8 below).

8. Arakawa's -k(') appears where it is not expected.

In one case, the Tibetan and Chinese transcriptions contradict his reconstruction:

TT1718 THREE 1.70

reconstructed by Gong as sọ

reconstructed by Arakawa as sok

transcribed in Tibetan as gso, gsoH, gswong (Nevsky 1926 #269)

(but not gs(w)og)

transcribed in Chinese as 桑 (pronounced something like sÕ?; has never had a final -k at any stage of Chinese history)

the rhyme class of pre-Tangut period 桑 was transcribed in Tibetan as -ang and -o

The Tibetan -ng and the Chinese nasalization are hard to reconcile with Arakawa's -k.

Moreover, I know of no Sino-Tibetan language with a final stop in its word for 'three'. The probable gDong-brgyad rGyalrong cognate of Tangut THREE is fsum (in sqa fsum 'thirteen'; xsum 'three' was borrowed from Tibetan gsum 'three' [Jacques 2003: 24]).

In another case, a -k' appears where other Sino-Tibetan languages have -ng or zero:

TT5233 HORSE 1.74

reconstructed by Gong as ryiry

reconstructed by Arakawa as ryek'2 (I don't know what the -2 means.)

no Tibetan transcription known

transcribed in Chinese as (Pearl 062) and 領 (no diacritic in eds. of the Pearl reproduced in Kwanten 1982, contra Nishida 1964); 領 was something like lẽ

領 was transcribed in Tibetan in the pre-Tangut period as leng; has a nasal vowel in most modern NW dialects

presumably related to

Old Chinese 馬 mra' 'horse'

Old Tibetan rmang 'horse'

Written Burmese mrangH 'horse'

gDong-brgyad rGyalrong mbro 'horse' (Jacques 2003: 22)

Here are the numbers of correspondences between Arakawa's -k(') and the ten final consonants of gDong-brgyad rGyalrong in the cognate sets of Jacques (2003):

TangutCorresponding gDong-brgyad rGyalrong coda
Arakawa codaArakawa rhymeRhyme number-R-G-t-V-r-l-ng-n-m-s
-ek22.53no examples
2.57no examples
2.58no examples
2.59no examples
-yek21.73no examples

Unless Arakawa's -k(') was a suffix added to non-k rhymes, I cannot explain why it always corresponds to zero or gDong-brgyad rGyalrong consonants other than -G (< *-k). Although *-t > -k is plausible (it has a precedent in Vietnamese), -k is most likely not from *-p (the source of gDong-brgyad rGyalrong -V), *-r, or *-s.

Only Arakawa's -ak for 1.63/2.56 strongly correlates with gDong-brgyad rGyalrong -R which is from Proto-rGyalrong *-q (Jacques 2004: 262).

9. Conversely, Arakawa's final stops do not appear where they are expected:

TT5745 (second half of 'Tangut') 2.18

reconstructed by Gong as nyaa

might be reconstructed by Arakawa as nyaa(2)

my guess based on his n- corresponding to Gong's n-

and his -yaa(2) in his reconstructions of other 2.18 syllables

(I don't know what -2 represents.)

corresponding to Tibetan ñag in Mi-ñag 'Tangut'

TT3448 SIX 1.46

reconstructed by Gong as chhyiw

reconstructed by Arakawa as chheu:

(does eu: represent a single long vowel [öö]/[üü] or a diphthong [euu]?)

transcribed in Tibetan as chhi

transcribed in Chinese as 抽 (Pearl 064); something like chhew

抽 was transcribed in Tibetan in the pre-Tangut period as chheHu

presumably related to

Old Chinese 六 (mə)ruk 'six'

Written Tibetan drug 'six'

Written Burmese khrauk 'six'

gDong-brgyad rGyalrong ChëG 'six' (Jacques 2003: 13)

all sharing a common root *ruk 'six'

Here are the numbers of correspondences between gDong-brgyad rGyalrong -R, -G, -t, -V (< Proto-rGyalrong *-q, *-k, *-t, *-p) and Arakawa's -k('):

gDong-brgyad rGyalrongArakawa -k rhymeArakawa -k' rhymeOther Arakawa rhymesTotal
-R (< *-q)8 (all 1.63/2.56)01523
-G (< *-k)0!1212
-V (< *-p)301215

Ideally, Arakawa's -k(') should correspond to Proto-rGyalrong final stops, but that only occurs in a quarter (14/59) of the cognate sets of Jacques (2003). And if one subtracts the correspondences with -t and -V, the figure goes down to 13.6% and consists solely of the eight cases of

Arakawa -ak 1.63/2.56 : gDong-brgyad rGyalrong -aR (< Proto-rGyalrong *-aq, *-ɐq, *-Oq [Jacques 2004: 266])

But note that in the majority of cases (65.2% = 15/23), gDong-brgyad rGyalrong -R does not correspond to Arakawa's -k(').

10. One huge obstacle to reconstructing stops in the level or rising tones is the fact that stop-final syllables belong to the 'entering tone' in the Chinese phonological tradition.

The Precious Rhymes of the Tangraphic Sea did have a very short entering tone section with eleven legible entries following the rising tone section. Gong reconstructed four of them with the rising tone. The tones of the other seven are unknown. Arakawa would reconstruct final stops for at least two of them (TT3522 and TT3553) but would not reconstruct final stops for another two (TT2109 and 4387).

PRTS entry numberTangut TelecodeGlosses from Grinstead (1972) and Shi et al. (2000: 317)Gong's reconstructionGong's rhymeArakawa's reconstruction of Gong's rhyme
2.18.13023522共 食物 SHARE-FOODkạ2.56-ak
2.18.13033553圍 繞 REVOLVE-AROUNDkhya2.21-a'
2.18.13042109姻 親 RELATION-BY-MARRIAGE2.25-I
2.18.13053842DISCOURSE or 才 ABILITYnwə1.27, 2.25, 1.68, 1.84, or 2.76?-I, -I, -ik', -Ir, or -Ir?
2.18.13062608印 STAMPnwə
2.18.13071837牧 PASTUREnwə
2.18.14025625逼 PRESS-AN-ENEMYtha1.17, 2.14, 1.63, 2.56, 1.80, or 2.73?-a, -a, -ak, -ak, -ar, or -ar?
2.18.14031603迫 PRESS-UPONtha
2.18.14044692YAWN or 斜 SLANTED; Li Fanwen (1986: 426): 歪斜 SLANTED(laryngeal initial)??
2.18.14054387輕 LIGHTshyï2.27-II
2.18.14060289愚 FOOLISH1.27, 2.25, 1.68, 1.84, or 2.76?-I, -I, -ik', -Ir, or -Ir?

Ideally, all of Arakawa's stop-final syllables should have been listed in the 'entering tone' section, and there should be no 'entering tone' syllables such as TT2109 and TT4387 that Arakawa would reconstruct without final stops.

I don't know why Arakawa reconstructed final stops where he did, but their distribution among the rhymes is not random:

- The later rhyme cycles were reserved for syllables with final consonants (including his -r, if it did not indicate a retroflex vowel)

- Exceptions were rhymes ending in a final glottal stop which were in the first rhyme cycle; maybe the glottal stop was 'not consonantal enough' to be grouped with rhymes ending in -k(')

One would assume that the transcriptional and external evidence indicated final consonants in the later rhyme cycles as opposed to the first rhyme cycle, but that is not the case (see points 7-9 above). There is some transcriptional and external evidence for retroflexion in the third rhyme cycle, but what distinguished the second and fourth rhyme cycles from the first is still not clear to me. The tense vowels that others propose for those rhymes are not directly reflected in any transcription because Tibetan, Chinese, and Sanskrit had no tense vowels.

I conclude that Tangut probably did not have a -k during the imperial period. The case for its nasal counterpart -ng is stronger, since at least -ng does appear in the Tibetan transcriptions. But how much stronger?

Next: Nasal or not?

*I have not yet seen an overt presentation of Arakawa's reconstruction. My knowledge of it is based solely on the initial consonant chart, the one thousand sample tangraph readings, and the list of Tangut equivalents of Chinese surnames on 小高祐次 Kotaka Yuuji's site.

**Ideally, Chinese 'entering tone' should correspond to Arakawa's stop-final rhymes, and Chinese 'level', 'rising', and 'departing tones' should correspond to Arakawa's other rhymes. But one can find 'entering tone' corresponding to Arakawa's non-stop-final rhymes:

Tangut rhymeSinographs in the Pearl from Nishida (1964) which had the 'entering tone' in the pre-Tangut periodArakawa's reconstruction of the Tangut rhyme
1.1/2.1*-k: 沐, 北-u
1.2/2.2*-k: 莫, 芍, 叔, 竹, 粥, 藥, 六-yu
1.3/2.3*-k: 玉, 局, 菊, 俗, 宿, 續, 雀, 鵲, 藥, 六-yu
1.8/2.7*-k: 北, 墨, 賊, 勒-i
1.9/2.8*-k: 百, 木+百, 客, 額, 枙, 隔-yi
1.10/2.9*-k: 食, 植, 石, 尺-ii
1.11/2.10*-k: 戚, 寂, 夕, 息, 迹, 脊, 積-ii
1.27/2.25*-k: 特, 各, 屋-I
1.29/2.27*-k: 釋, 赤, 勒; *-p: 笠, 立-II
1.30/2.28*-k: 特, 刻, 尅, 錫, 則, 賊, 驛, 勒; *-p:-II
1.32/2.29*-k: 客, 國, 色; *-p:-yIn
1.43/2.38*-k: 瀆, 則, 賊, 勒-eu
1.49/2.42*-k: 諾, 穀, 閣-o

I have omitted pre-Tangut period *-r (< earlier *-t) sinographs since it is not clear whether final liquid rhymes still belonged to the 'entering tone'. Pre-Tangut NW Chinese*-r rhymes correspond to many of Arakawa's open-syllable rhymes: e.g., 1.1/2.1 -u, 1.8/2.7 -i, 1.17/2.14 -a, etc.

In Arakawa's defense, it is not clear whether the pre-Tangut period 'entering tone' still existed when the Pearl was written. It is probable that *-k and *-p had lenited to fricatives, had been reduced to a final glottal stop, or had been lost altogether. (If *-k and *-p still existed, it would be hard to explain how both *-k and *-p sinographs could correspond to a single Tangut rhyme class [e.g., 1.29/2.27].)

It's also possible that Arakawa's final stops had also disappeared by the late 12th century.

One can also find non-'entering tone' rhymes corresponding to Arakawa's stop-final rhymes:

Tangut rhymeSinographs in the Pearl from Nishida (1964)Arakawa's reconstruction of the Tangut rhyme
1.58/2.51open syllable: 吳, 多, 都, 奴, 怒, 古, 姑, 枯, 刳, 午, 五, 魯, 盧 outnumber*-k: 郭, 落, and possibly 惡 (had both open-syllable and *-k readings)-uk
1.63/2.56all syllables without *-k: open syllable: 那, 訶; *-r: 末, 斡, 割; *-p: 納, 臈; *-m: 罨; *-ng:-ak
1.64/2.57almost all syllables without *-k: open syllable: 巴, 麻, 賈, 我; *-y: 味, 載; *-r: 怛, 捺, 辣; sole exception is *-k: -aak
1.67/2.60open syllable: 爲, 胃, 知, 离, 梨, (合), 日知 and *-y: 底, 埿, 西,齊, 劑, 累 together outnumber *-p: 協 and *-k: 力靂-iik
1.70/2.62*-ng: 郎 (only sinograph corresponding to this Tangut rhyme; as a single example, it could be aberrant)-ok
1.73/2.66*-y: 嵬, 乃; *-ng: -yek'2
1.94open syllable: 果 (only sinograph corresponding to this Tangut rhyme; as a single example, it could be aberrant)-wok

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