Gong reconstructed two rhymes as -ju (R2, R3) and two rhymes as -juu (R6, R7). These rhymes are not in complementary distribution (see part 2 and part 3 of "Nonexistent Tense Twins"). R6 is the rarest of the four:

Rhyme numberGong's reconstructionNumber of tangraphs in Precious Rhymes of the Tangraphic Sea
Level toneRising toneTotal

These numbers alone are not what I would expect if R2 were to R3 what R6 was to R7. R7 outnumbers R6 by a ratio of 9.2 to 1, whereas R3 outnumbers R2 by a ratio of only 1.6 to 1.

The distribution of initials shows four different patterns rather than two (i.e., one for R2/R6 and another for R3/R7):

Rhyme numberHomophones chapter (initial class) and sample initial
R6nokh- onlynoʑ- only

Compare the patterns for R2-R3 with the distribution of initials before modern standard Mandarin -y and -jow:

RhymePinyin spellingsLabialsLabiodentalsDentalsRetroflexesVelarsAlveolarsPalatalsZeroLiquids
-u-u, wuyesnoyes
-y-u, -ü, yunon- onlynoyes
-jow-iu, youm- onlynot-, n- only

MSM palatal -y and -jow appear after palatal initials and only marginally after labials and dental stops but not after retroflexes, velars, or alveolars, whereas MSM non-palatal -u has the opposite distribution. There are historical reasons for this: e.g., earlier velars and alveolars became palatals before palatal rhymes. Similarly, did pre-Tangut velars and alveolars become palatals before R2?

R3, R6, and R7 all can be preceded by velars, unlike R2 or MSM -y, -jow. Moreover, velars and palalals can coexist before R7, unlike R2-R3, R6, or MSM -y, -jow.

Were R2, R3, and R7 at different points of a palatal continuum?

Least palatalMore palatalVery palatal
Can appear after most initials other than palatals (class VII)Can appear after almost any initial except for 'retroflexes' in the rare class IVCan appear after palatals but not after dentals, velars, or alveolars (classes III, V, VI).

If palatal initials required a following -j- glide, was R3 -y whereas R2 was -ju? (This is more or less what Sofronov [1968: 136] proposed, except that his for R3 was not as high as my -y. I do not know why he did not propose a high -y.)

R2-R3 alternations in word families suggest that the two were similar. But an alternation between R2 -ju and R3 -y is difficult to explain unless an earlier affix blocked monophthongization:

pre-Tangut *-ju > R3 -y


pre-Tangut *-ju-C (coda suffix), *-juC (coda part of root) > R2 -ju

The above proposal entails that R2-R3 alternations involve R3 base forms and R2 derivatives. This has yet to be confirmed.

One might then reconstruct R6-R7 as long-vowel counterparts of R2-R3 (-juu and -yy), but there is no reason for long-vowel rhymes to have different initial distribution patterns.

I had hoped that the evidence would point toward two subgrades (IIIa with R2 and R6 and IIIb with R3 and R7) but the four rhymes that Gong reconstructed as -ju(u) seem very different from each other. Although I will continue to use Gong's -ju(u) notation for those rhymes, their phonetic values remain uncertain.

*The reconstruction of the 'retroflex' initials of Homophones chapter IV is controversial, so I have not supplied any sample initial.

**Although Nishida (1964: 43, 82) thought that there were R3 tangraphs in chapter II of Homophones, these tangraphs were listed as R2 words in Precious Rhymes of the Tangraphic Sea (which Nishida was unable to consult). -M-ISSING: WHICH RGYALRONG RHYMES DON'T CORRESPOND TO TANGUT LONG VOWELS?

According to Gong's reconstruction of the rhymes of the Tangraphic Sea, rhyme groups consist of short-vowel rhymes followed by long-vowel rhymes: e.g.,

lax -u rhyme grouprhyme numberrhymegrade
short vowels1-uI
2-juIII (the difference between R2 and R3 is unclear)
4-u ([uʕ]?)I (acc. to Gong; I think it might be II [which I interpret as pharyngealized], though Grade II precedes Grade III in all other rhyme groups)
long vowels5-uuI
6-juuIII (the difference between R6 and R7 is unclear)

(I hesitate to label R2/R3 and R6/R7 as Grade IIIa/b since I have not yet deterimined that R2 was to R3 what R6 was to R7.)

Gong's reconstructed vowel length cannot be confirmed in Chinese or Tibetan transcriptions, and the Sanskrit transcription evidence does not appear compelling (though I have to yet to perform a full-scale study). Tangut does not seem to have been like Thai, whose borrowings from Sanskrit and Pali consistently replicate the original vowel length. (But it's possible that the Tangut were transcribing a Chinese pronunciation of Sanskrit lacking vowel length.)

I am also skeptical of Gong's vowel length because I don't know of any Southeast or East Asian language like Gong's Tangut reconstruction that has 'live syllables'* ending in short vowels bearing a full range of tones**.

I looked at Jacques (2003) to see if I could find any correspondences between Tangut vowel length and gDong-brgyad rGyalrong (DG) rhymes. The following DG rhymes only correspond to Tangut short-vowel rhymes:

-ar (4)-aβ (4)
-ɤɣ (10)-ɤt (6)-ɤs (1)-ɤβ (5)-ɤm (3)
-iɯ (1)
-ɯɣ (2)-ɯm (4)
-oʁ (3)-om (9)
-um (3)

Numbers in parentheses indicate the number of cognates found by Jacques for that rhyme.

All other DG rhymes can correspond to Tangut short- or long-vowel rhymes.

There are not enough Tangut cognates for DG -iɯ and -ɤs words to draw any conclusions about those rhymes.

DG -aʁ (< Proto-rGyalrong *-aq, *-ɐq,*-ɔq)

DG -ɤr (< PGR *-ɐr, *-ɔr)

DG -ɯr (< PGR *-ɯr, *-ur, *-ir)

DG -at (< PGR *-at)

DG -ɯt (< PGR *-ot, *-ut, *-it)

can correspond to Tangut short- or long-vowel rhymes. Although other DG rhymes with the same codas

DG -oʁ (< PGR *-oq)

DG -ar (< PGR *-ar)

DG -ɤt (< PGR *-ɐt, *-ɔt, *-et)

apparently lack long-vowel Tangut cognates, such cognates may be discovered in the future, or simply did not exist due to chance.

I don't know what to make of the absence of long-vowel Tangut cognates for DG words with grave codas:

< PGR *-k

< PGR *-p

-m < PGR *-m

This looks like a pattern, but it has one exception: DG -o < PGR *-aŋ, the only PGR rhyme with a velar nasal corresponds to Tangut -jiij (my -jee) in two cases:

'dream': DG tɯ-jmŋo : TT0038 mjiij 1.39

'wait': DG kɤ-nɤ-jo (< PGR ?*ljaŋ) : TT2266 ljiij 2.35

I can't think of any phenomenon that could occur in all rhymes but those ending in -k, -p, or -m.

Tangut -o partly corresponds to DG -o < PGR *-aŋ. It may be significant that there are no known DG cognates for Tangut roots ending in -oo***.

*A translation of Siamese คำเป็น kham pen. Live syllables end in sonorants, whereas 'dead syllables' (Siamese คำตาย kham taay) end in obstruents.

**1.13.4:31: Even if Siamese syllables ending in short vowel-glottal stop sequences are analyzed as phonemically ending in short vowels with predictable glottal stop codas, such syllables can only have three tones (high, low, or falling), whereas Siamese live syllables can have one of five tones (the aforementioned three plus mid and rising). (If there are any Siamese dead syllables ending in

***1.12.00:06: The only Tangut -oo word in Jacques (2003) is TT0530 ɕjoo 1.53 'kill an animal'. Its -oo is a Tangut suffix:

TT5016 ɕjii 1.14 + -o = TT0530 ɕjoo 1.53

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2008 Amritavision