If I can figure out what the Geng and Zeng rhyme groups were like in the northwest, I might be able to propose better reconstructions of these Tangut rhymes:

Rhymes of transcription tangraphs in the Tangut translation of the Forest of Categories
(based on Gong, 類林西夏文譯本漢夏對音字研究)
Used to transcribe Geng syllables Used to transcribe Zeng syllables
Rhyme Gong AMR Rhyme Gong AMR
R8 -e -əi R8 -e -əi
R9 -ie ɣ
R10 -ji -ɨi
R11 -ji -i R11 -i
R12 -ee -əii
R15 -ẽ -əĩ
R29 -iə ɣɨ
R31 -jɨ -iɨ
R34 -ej -e
R35 -iej ɣ
R36 -jij -ɨe
R37 -jij -ie
R42 -iəj ɣj
R43 -jɨj -iej
R51 -o -o
R56 -ow -ow
R58 -jow -uow
R62 -jụ -ɨụ
R104 -ũ -əũ

It's hard to figure out what distinguished Geng from Zeng syllables purely on the basis of Tangut transcriptions because there is some overlap between the two categories:

Main vowel in my reconstruction
(disregarding medials and warping)
Used to transcribe category
i Geng and Zeng
ɨ, e Geng only
o, u Zeng only

This implies that Geng and Zeng were partly merged in the dialect of northwestern Chinese known to the Tangut. This is not surprising, since the two categories have partly merged in modern Chinese languages. Hence their standard Mandarin names both end in -eng. WHAT'S GENG ON? (PART 2)

In EG Pulleyblank's (1991) Late Middle Chinese reconstruction, all the rhymes of the Geng group end in -ajŋ or -ajk. They differ only in terms of medial vowels (none, -i-, or -y-) and length (short a or long aa). Such a simple system seems to conflict with the actual Sinoxenic data presented in part 1:

- On the one hand, Sino-Vietnamese Geng group syllables almost always end in palatals (in northern pronunciation which preserves earlier codas). This seems to validate Pulleyblank's LMC reconstruction. (Some analysts even interpret Vietnamese final palatals as -j- + velar clusters!)

- On the other hand, Sino-Korean and the Kan-on stratum of Sino-Japanese seem to indicate that some Geng syllables ended in palatalized velars whereas others ended in plain velars:

Sino-Korean Kan-on Implies LMC
-jŋ -i *-jŋ
-jk -ki *-jk
-u *-ŋ
-k -ku *-k

- However, SK and Kan-on often imply different LMC rhymes!

(SK forms are premodern spellings from 金敏洙 Kim Minju's [1961] 새字典 sEjajOn. -ajŋ and-ajk were not used in Middle Korean and have been archaized to -ʌjŋ and -ʌjk.)

LMC grade Sinograph Pulleyblank's LMC SK Implies LMC Kan-on Implies LMC
II *kjaajŋ kjəŋ *-ŋ kau *-ŋ
kʌjŋ *-jŋ
*kjaajk kjək *-k kaku *-k
*khjaajk kʌjk *-jk
*xɦwaajŋ kojŋ *-jŋ kwau *-ŋ
*kwaajk kojk *-jk kwaku *-k
III *kiaŋ kjəŋ *-ŋ kei *-jŋ
*kiajk kɯk *-k keki *-jk
IV *jiajŋ jəŋ *-ŋ ei *-jŋ
*jiajk jək *-k eki *-jk

That's where I left off yesterday. Looking outside the Geng group, I found one case of an SK reading implying a palatalized velar in the 曾 Zeng group that Pulleyblank reconstructed with plain velars:

LMC grade Sinograph Pulleyblank's LMC SK Implies LMC Kan-on Implies LMC
I *kuə̆k kujk *-jk koku *-k

(Although the reading kujk for 國 is not in sEjajOn, it is in the title of 東 國正韻 Tongguk chOng'un [1448], which was spelled as 동귁졍운 toŋkujk cjəŋŋun in early han'gUl. The current SK reading of 國 is kuk, which might be reduced from kujk.)

The Vietnamese reading of 國 has a final velar: quốc.

All of this data indicates that the three varieties of Sinoxenic cannot be easily derived from a single generic variety of LMC. I think they were based on three different LMC dialects: southern, northeastern, and northwestern:

Rhyme group LMC grade Pulleyblank's generic LMC SV S LMC SK NE LMC Kan-on NW LMC
Zeng (國 only) I *-k -c *-k -jk *-jk -ku *-k
Geng II *-jŋ -nh *-jŋ -(j)ŋ *-(j)ŋ -u *-ŋ
*-jk -ch *-jk -(j)k *-(j)k -ku *-k
III and IV *-jŋ -nh *-jŋ *-ŋ -i *-jŋ
*-jk -ch *-jk -k *-k -ki *-jk

NW Chinese is of particular interest to me because (1) Tangut and pre-Tangut borrowed many words from it and (2) the Chinese transcription of Tangut in Pearl in the Palm is based on a 12th century NW Chinese dialect.

Next: The Tibetan transcription evidence for the Geng and Zeng groups in the northwest. WHAT'S GENG ON? (PART 1)

(The pun only works - barely - if Md geng [kəŋ] is mispronounced as [gɛŋ].)

A reader and I have been talking about final palatal(ized) nasal and stop codas in Middle Chinese. In EG Pulleyblank's (1991) reconstruction of Late Middle Chinese, all rhymes ending in palatalized velars (*-jŋ, *-jk in his notation) belonged to the 梗 Geng rhyme group.

In the stratum of Sino-Vietnamese borrowed from LMC, Geng syllables end in the palatals* -nh [ɲ] and -ch [c].

However, the situation is more complex in Sino-Korean and the Kan-on stratum of Sino-Japanese. (The representations of both are historical and do not necessarily reflect current pronunciation. The SK may contain errors, since I can't check the early SK rhyme dictionary 東國正韻 Tongguk chOng'un [Correct Rhymes of the Eastern Nation] and I haven't seen it in almost a decade. Nonetheless, the general pattern below is still valid for modern SK.)

MC grade Pulleyblank's LMC reconstruction Sino-Vietnamese Sino-Korean Kan-on
II *-aajŋ -anh -ʌjŋ ~ -jəŋ -au ~ -ei
*-aajk -ach -ʌjk ~ -jək -aku ~ -eki
III/IV *-iajŋ -inh ~ -anh -jəŋ -ei
*-iajk -ich (-ach is marginal) -jək -eki
*-yajŋ -uynh -jwəŋ -ei
*-yajk -ich (-uych is rare) -jwək -eki

SK -jŋ/k and Kan-on -(k)i are unique to the Geng group and presumably reflect LMC *-jŋ/k. But SK -ŋ/k and Kan-on -(k)u are also found in the 宕 Dang, Zeng,Jiang, and 通 Tong rhyme groups which Pulleyblank reconstructed with final plain velars (*-ŋ/k) or final labialized velars (*-wŋ/k). Moreover, SK -jŋ/k generally corresponds to Kan-on -(k)u rather than Kan-on -(k)i, which corresponds to SK -ŋ/k. (The common graph 生 SK sʌjŋ : Kan-on sei violates my generalization.)

Next: As if that weren't confu-Zeng enough ...

(That's another bad pun, since Zeng is really [tsəŋ].)

*I can't think of any true counterexamples ending in velars. Apparent counterexamples are either analogical readings (based on other graphs with the same phonetic) or borrowings from earlier stages of Chinese which lacked palatal(ized) nasal and stop codas.

Strictly speaking, these 'palatals' are actually dentals in central and southern Vietnamese. Final dentals merged with velars, leaving a gap that was filled when the palatals became new dentals:

> -n >


Until yesterday morning, I had no idea that Guillaume Jacques was a member of this team:

JACQUES contributes a broader historical perspective on the Qiangic subgroup within the Sino-Tibetan language family as well as his knowledge of rGyalrong (Japhug) and Tangut languages, which are indispensable for historical reconstructions within this subgroup due to their well-preserved archaic phonological and morphological features.

I look forward to their results. Understanding modern Qiangic languages can only improve my understanding of Tangut.

*A literal translation of 木里 Muli, the Mandarin name of the autonomous county that Guillaume's team is studying. I presume Muli is a Mandarinized form of སྨི་ལི་ Smili, the Written Tibetan name of the former kingdom. I don't know why the name wasn't Mandarinized as Mili. Is the local pronunciation of Smi something like [mɨ] or [mə]?

I've long been puzzled by how Minyag ended up as 木雅 Md Muya 'wooden elegance' with -u-.

Minyag is a northern Qiangic language. Minyag is also the Written Tibetan name for the Tangut. The name ultimately comes from an early form of the Tangut autonym

mjɨ 2.28 njaa 2.18

with a final *-k.

Aha, the Minyag call their own language məɲɑsu! I guess 木 Md Mu is the best possible match after all. (Graphs for the Md syllable me [mə] aren't used in any transcriptions that I've ever seen.)

Guillaume's site has sound samples of Minyag that he recorded in 2003.*Q-UESTIONING THE UVULAR TO GLOTTAL HYPOTHESIS

In "Unusually Great Possibilities", I expressed "reservations" about deriving Late Old Chinese glottal stops and fricatives from Early Old Chinese uvular initials that weren't 'shielded' by presyllables:

EOC*q- > LOC *ʔ-

EOC*qh- > LOC *h-


EOC *CV-q(h)- > LOC *k-

If one assumes that a phonetic series mostly contains syllables without presyllables, one would predict that uvular phonetic series should have more glottal-initial members than velar members. Although I don't have the time to check all potential uvular series, my general impression is that velar initials outnumber glottal initials. Here's what I found when I looked at eight potential uvular series in Karlgren's (1957) Grammata Serica Recensa. Numbers in bold represent readings which had been emphatic in earlier OC.

GSR number Phonetic LOC glottal initials LOC velar initials LOC retroflex (!) initials LOC alveolar (!) initials
*ʔ- *h- *k- *kh- *khj- *g- *ŋ- *ʈh- < ?*rɯ-tɯ-q- *tsh- < ?*sɯ-qh- *s- < ?*sɯ-χ- *z- < ?*sɯ-ɢ-
1 3+8 2+0 6+8 1+4 - 6+7 0+2 - - - -
43 1+0 0+1 - 10+0 - 4+0 - - - - -
122 9+4 - - 2+6 0+1 - - 0+1 - - -
313 4+3 1+2 3+2 4+8 - 5+7 - - - - -
629 1+0 - 1+0 - - 3+0 - - - - -
651 1+7 - 0+2 1+2 - 3+9 2+3 0+1 0+1 - -
652 - 0+1 0+2 0+3 - 0+1 1+2 - - - -
1173 1+0 - 1+0 - - - - - - 0+1 0+4

(652 is actually a subset of 651. I have included 1190 松 in 1173.)

(Am I missing any other *ʔ-/k-mixed series? I can't think of any others off the top of my head.)

Glottal initials are outnumbered by nonglottals:

GSR number Phonetic Glottals Non-glottals Total
1 13 34 47
43 2 14 16
122 13 10 23
313 10 29 39
629 1 4 5
651 8 24 32
652 1 9 10
1173 1 6 7
Total - 49 130 179
Percentages - 27.4% 72.6% 100%

Does this mean that simple EOC uvular initials became LOC velars, and that the LOC glottals came from complex initials (or presyllable-initial sequences)?

Conversely, does this mean that simple EOC initials were uncommon, and that a normal EOC word consisted of a core syllable preceded by one or more presyllables? UNUSUALLY GREAT POSSIBILITIES (logonote*)

I opened EG Pulleyblank's (1994) Lexicon of Reconstructed Pronunciation at random and found two words that I never noticed before (LOC = Late Old Chinese):

觭 LOC *khɨaj 'one horn turned up, one turned down'

踦 LOC *khɨaj 'lame; one-legged, standing on one leg'

which has alternate readings

LOC *kɨajʔ 'leg, shin' (not in Karlgren 1957; is this a late meaning?)

cf. 掎 LOC *kɨajʔ 'pull by one leg'

LOC *kɨajʔ 'stand around close to the door' (cf. 'lean' below)

LOC *ŋɨajʔ 'knock against'

The graphs for these words share a phonetic 奇 which has two readings:

LOC *kɨaj 'odd number'

the same word as 畸 *kɨaj 'odd; left over; remainder; irregular; lopsided; wing of an army'

LOC *gɨaj 'strange, unusual'

could 琦 LOC *gɨaj 'fine jade; rare, beautiful' < 'something unusual'?

that meaning is late; the earliest meaning is 'handle'

All of the above might be cognate to each other:

'leg' > 'one-legged' > 'odd' (number/unusual; English odd also has both meanings)

So far, I have been citing LOC forms because they are more certain than their OC sources. Ideally, I'd like to derive all of these words from a single OC root plus various suffixes.

The phonetic 可 LOC *khɑjʔ 'can, able, may' of 奇 LOC *kɨaj/*gɨaj 'odd' may give us a clue about the root of the 'odd' word family. *khɑjʔ comes from an earlier emphatic *khalʔ [qhɑʕlʔ]. This in turn could come from an even earlier *qhalʔ or *khalʔ, since I hypothesize that *a became emphatic and shifted a preceding velar to uvular unless it was preceded by a high-vowelled presyllable. I favor *qhalʔ since 可 is also phonetic in graphs for glottal-stop initial words:

倚 LOC *ʔɨajʔ < *ʔal lean on, rely on; leaning to one side'; phonetic loan for 奇 'strange')

also written 猗 LOC *ʔɑjʔ < *ʔalʔ

cognate to

阿 LOC *ʔɑj < *ʔal 'slope; to slope, lean towards'

攲 LOC *kɨaj < *kal 'slanting'

If glottal stops partly originated from uvulars (as hypothesized by Sagart, Baxter, and others), then

*q- > *ʔ- + emphasis

*qh- > *h- (phonetically *[χ]?) + emphasis

presyllables added to the resulting *ʔ- and *h- could change them to nonemphatic:

*Cɯ-ʔ- > *ʔ- + nonemphasis

*Cɯ-h- > *h- (phonetically *[x]?) + nonemphasis

*Cʌ-q(h)- > *q(h)- + emphasis

*Cɯ-q(h)- > *k(h)- + nonemphasis

*H- = 'aspirator' (probably *s-, possibly *k- and/or *x-)

*Hʌ-q(h)- > *Hq(h)- > *qh- + emphasis

*Hɯ-q(h)- > *Hɯ-k(h)- > *Hk(h)- > *kh- + nonemphasis

*N- = unknown nasal (*ŋ-, *n-, *m-)

*Nʌ-q(h)- > *ɴq(h)- > *ɴɢ(h)- > *ɢ- + emphasis

*Nɯ-q(h)- > *Nɯ-k(h)- > *ŋk(h)- > *ŋg- > *g- + nonemphasis

*Cʌ-Nʌ-q(h)- > *Cʌ-ɴ- > - + emphasis

*Cɯ-Nɯ-q(h)- > *Cɯ-ŋ- > - + nonemphasis

or in table format (with emphatics in bold),

Core initial No presyllable; backed Presyllables added after backing Presyllables shielding uvulars from backing
*Cʌ- *Cɯ- *Cʌ- *Cɯ- *Hʌ- *Hɯ- *Nʌ- *Nɯ- *Cʌ-Nʌ- *Cɯ-Nɯ-
*q- *ʔ- *ʔ- *q- *k- *qh- *kh- *ɢ- *g- - -
*qh- *h- *h- *qh- *kh-

I have reservations about this scheme, but assuming that it is correct, then the above words might have developed in this fashion:

Sinograph OC before backing of uvulars without presyllables Backing of uvulars; emphatic harmony; apocope of second prefixes Late prefixation; loss or fusion of earlier prefixes Late Old Chinese with vowel bending; uvulars front to backed velars?
*qal *ʔal *ʔɑj
*qal-ʔ *ʔalʔ *ʔɑjʔ
*ʔalʔ *Cɯ-ʔalʔ *ʔɨajʔ
奇畸攲 *Cɯ-qal *Cɯ-kal *(Cɯ-)kal *kɨaj
踦 掎 *Cɯ-qal-ʔ *Cɯ-kalʔ *(Cɯ-)kalʔ *kɨajʔ
*Hʌ-qalʔ or *Cʌ-qhalʔ *Hʌ-qalʔ or *Cʌ-qhalʔ *qhalʔ or *(-)qhalʔ *khɑjʔ
觭踦 *Hɯ-qal *Hɯ-kal *khal *khɨaj
奇琦 *Nɯ-qal *Nɯ-kal *gal *gɨaj
*Cɯ-Nɯ-qalʔ *Cɯ-ŋ-kalʔ *ŋalʔ *ŋɨajʔ

(琦 'handle' and 踦 'knock against' could also have been *Cɯ-ɢal and *Cɯ-Nɯ-ɢalʔ with voiced-initial roots in early OC.)

There seem to have been at least three early roots (excluding 'knock against' and 'handle'):

*qal 'one-legged' (踦 > 奇畸攲掎觭)

*qal 'slope/slant/lean on' (阿猗倚)

*qalʔ or *qhalʔ 'possible' (可)

Schuessler (2007: 275) glossed this as 'to bear' and linked it to 何荷 OC *galʔ < ?*Nʌ-q(h)alʔ 'to carry' and Written Tibetan khal 'load', Hgel-ba 'to load'.

Schuessler also mentioned Bodman's (1980: 138) comparison with Chepang khaaj 'be able'. However, the Chepang comparisons in Matisoff (2003) seem to indicate that Chepang -j is from *-j and not *-l, which Chepang generally preserves. (Matisoff gave isolated examples of *-l becoming -r and -n, but not -j.)

The first two roots may really be one if 'one-legged' is from 'slanting' or 'leaning' (since a one-legged person would not be able to stand up straight).

*The title refers to 奇 'unusual' and its components, 大 'big' and 可 'possible'.

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