In "Beyond D-bris", I proposed that Gong's reconstructed Tangut consonant G- may have originated from an earlier velar (e.g., k-) that lenited after a consonant. In "Y Gamma", I proposed that pre-Tangut Gi and yi merged into yi since there was no Gi in Tangut. These two proposals can be linked:

*C-ki > *Gi > yi

The Tangut genitive/dative marker

TT5283 'yiy 1.36

might confirm this hypothetical development. It was transcribed in Tibetan as g-ye, ye, yi. The g-y- transcription might have represented a Tangut Gy-. Perhaps

*C-kyiy > Gyiy (which could further weaken to 'yiy?)

(But where'd the glottal stop in 'yiy come from?*)

I doubt that a high-frequency grammatical morpheme had a cluster initial *C-k-. So I suspect that the *C- before *-kyiy could have been a preceding syllable's final consonant that was later lost:

*CVC kyiy > CV Gyiy ~ CV 'yiy

In Gong's reconstruction, no Tangut syllable can end in a consonant other than a glide: i.e., -w, -y.

*C-kyiy, Gyiy, and 'yiy resemble the variant forms (allomorphs) of the Written Tibetan (WT) genitive:

-kyi after nonvelar obstruents (-b, -d, -s)

-gyi after nonvelar sonorants (-l, -m, -n)

-gi after velars (-g, -ng)

-Hi after vowels

I assume that the original genitive suffiix was -kyi whose initial assimilated to match the voicing of following sonorants and lenited to H (phonetically zero?) after vowels. (I don't understand why the postvelar allomorphs aren't -kyi after the obstruent -g and -gyi after the sonorant -ng.)

Was the initial of the Tangut genitive suffix (G- or 'y-) similarly determined by the preceding syllable?

(I thought Nevsky was the first to link the Tangut and Tibetan genitives, but I can't confirm that. Surely I can't be the first!)

The proposed earlier Tangut and Tibetan genitive suffixes in turn resemble Mawo Qiang -k. The equivalent Ronghong Qiang suffix is the genitive/nominalizing** suffix -ch, which might have come from an earlier *-ki (or *-kyi?) that is even more like the pproposed earlier Tangut and Tibetan genitive suffixes. See LaPolla and Huang (2003: 223).

This chart ties all of the above speculations together while glossing over details and adding rGyalrong to the mix:

Common ancestor of Tangut, Qiang, and TibetanEarly stageLater stage
*-kyi? Proto-Qiangic *-kyi?Tangut -Gyiy? ~ -'yiy
Mawo Qiang -k, Ronghong Qiang -ch
Proto-rGyalrong *-k > Japhug rGyalrong -G
pre-WT *-kyi?WT -kyi, -gyi, -gi, -Hi

It may not be necessary to reconstruct medial *-y- at deeper levels. Perhaps it is secondary in both Tangut and WT.

I am assuming that Tangut was a Qiangic language (see Matisoff 2004).

It's a shame that the Old Chinese genitive particle 之 has nothing to do with this genitive suffix, but related languages need not share every bit of their morphology: e.g. (Beekes 1995: 191-192),

Proto-Indo-European *-os (genitive singular ending for -o-stems):

Sanskrit -asya

Hittite -as

Homeric Greek -oio, Attic Greek -ou (< *-osio)

but Latin -i, which cannot be derived from -os at all!

(The ending -osio found in early Latin [or some close relative of it] has not survived in later Latin.)

Next: Y is y a 'gullet sound'?

* Generic Middle Chinese MC distinguished between y- and 'y-. Ideally one would expect TT5283 'yiy 1.36 to be transcribed in Tangut-period northwestern Chinese (TNC) with sinographs which had the MC initial 'y-. However, the TNC transcriptions for TT5283 GENITIVE/DATIVE 'yiy 1.36 in the Pearl are

Line 296: 耶

MC *ya

Transcribed in Tibetan and Khotanese Brahmi prior to Tangut period as ya

Modern NW Chn dialects: Lanzhou ie, Pingliang iə, Xi'an iE (latter two as heard by Karlgren almost a century ago)

Line 344: 盈

MC *yeyng

No pre-Tangut period transcription data available; the probable near-homophone纓 was transcribed in Tibetan as e

Modern NW Chn dialects: Xi'an ing (literary; cf. standard Mandarin ying), iã (colloquial)

As always, their TNC pronunciations are unknown (耶 = ye?, 盈= yẽ?), and it is not clear whether TNC distinguished between 'y- and y- like MC or the slightly later eastern Chinese koine recorded in the Hphags-pa alphabet (Coblin 2006: 43). If TNC only had y- (which is more likely), TNC y- would be the closest equivalent to both Tangut y- and 'y-. So the fact that TT5283 GENITIVE/DATIVE was not transcribed with Generic Middle Chinese 'y-sinographs does not tell us whether it had an initial glottal stop in Tangut or not.

Maybe TT5283 GENITIVE/DATIVE did not have an initial cluster 'y- after all. Its fanqie is

TT5283 'yiy 1.36 = TT3119 'yi 1.11 + TT5792 kyiy 1.36

and the fanqie for its initial speller is

TT3119 'yi 1.11 = TT5050 yï 1.30 + TT2644 kyi 1.11

with a y- (not 'y-!) initial speller. TT5050 also serves as initial speller for 'y-syllables. But its fanqie has a y-initial speller:

TT5050 yï 1.30 = TT3142 yu 1.2 + TT5480 syï 1.30

TT3142 serves as an initial speller for both y- and 'y-syllables and has a fanqie with a 'y-initial speller:

TT3142 yu 1.2 = TT4879 'yi 1.10 + TT0962 shyu 1.2

This interchangeability between y- and 'y- may suggest that there was only one y-like initial (y- or 'y-) rather than two (y- and 'y-) in the dialect underlying the fanqie (and perhaps in Tangut as a whole). Several scholars have taken the one-initial position, though there is disagreement about what that single initial was. Prior to adopting Gong's reconstruction, Li Fanwen (1986: 127, 407) reconstructed y- for that initial and reconstructed TT5283 GENITIVE/DATIVE as yẽ. If TT5283 GENITIVE/DATIVE had no glottal stop, then I could derive it as follows (CV[C] = syllable of a preceding word):

*CVC kyiy > CV Gyiy ~ CV yiy

**Cf. Japanese -no which not only marks the genitive after nouns but also can serve as a nominalizer after verbs. Y GAMMA?

In "Y-Clusters? Y Not?" I asked,

What Greek letter does y resemble?

γ, obviously*. Gong reconstructed IPA gamma - G in my notation - as the initial consonant of Tangut syllables transcribed in Tangut period northwestern Chinese (TNC) as 夷 'barbarian' followed by velar-initial sinographs (格 隔 耿 皆 客). TNC 夷 'barbarian' was also the transcription of three Tangut syllables that Gong reconstructed as 'yi, 'yï, yï with (')y-. (See the lists of 夷X and 夷 transcriptions along with their reconstructions here.) One might then expect 夷X to represent Tangut syllables with initial yk-, though I just argued against that! How does one get from a transcription implying yk- to G-?

A clue might lie in the Tibetan transcriptions of Tangut. Nevsky (1926 #274) lists five different Tibetan transcriptions of

TT4979 UNQUOTE 2.28

Tib. tr. g-yiH, g-yi, g-yyi (sic!), yiH, yi

(g-y- = ga + nonsubscript ya in Tibetan spelling, not gy- = ga + subscript medial -y-; -yy- is ya with subscript medial -y-)

and five different sinographs (依 醫 已 夷 姨) which sounded like it. Unfortunately, the TNC pronunciations of those characters are unknown. In modern northwestern languages, 依 and 已 are generally i (but Xining has vowelless y!) and I expect the others to also be i. (Coblin 1994 does not list modern NW readings for 醫 夷 姨.) Pre-Tangut Tibetan and Khotanese Brahmi transcriptions indicate that 依 and 已 once had distinct initials (glottal stop and y-):

依 Tib 'i, 'ï; Kbr ii

已 Tib yi, ye, yï, Kbr yii

Whether they still had distinct initials in the Tangut period is unknown. (And if they did, could the Tangut distinguish between TNC 'i and 'yi - and did they have a similar distinction in their own language?)

Nishida (1964: 127-128) lists other cases of g-y- ~ y- alternation in Tibetan transcriptions of Tangut:

Homophones Chapter VIII homophone group numberRhymeTibetan transcriptions
41.36g-ye, ye
82.2ywe, g-yu, yu
571.39g-ye, ye

Unfortunately, he does not specify the tangraphs represented by the transcriptions. (However, I think I know which rhyme 1.36 word he's referring to. I'll get to it in a future post.)

For a long time I assumed that Tibetan transcriptions with g-y- represented Tangut gə- or kə- presyllables which were 'unstable': cf. nonobligatory (initial / pre-)syllables in the unrelated Austronesian language Phanrang Cham and the unrelated Austroasiatic language Ruc (Alieva 1994 and Solntsev 1996 in Sagart 1999: 15):

PC tahla' ~ thla' ~ hla' 'I'

Ruc kU-lE ~ glE ~ lE 'chopstick', tu-mo, t-mo, mo 'entrance'

Could UNQUOTE have been gə-yï ~ yï? (The significance of the inconsistent final -H in the Tibetan transcriptions is unknown.)

Recently I've wondered if g-y- was [y] as in modern Lhasa Tibetan in the unknown Tibetan dialect(s) of the transcriber(s). But then why would anyone go to the trouble of writing [y] as g-y- when a simple y- could - and did - suffice? In modern Lhasa Tibetan, syllables written with y- and with g-y- have different tones. Was that already the case in the unknown Tibetan dialect(s) of the transcriber(s) c. 900 years ago? Was Tibetan g- intended to indicate a Tangut tone? I doubt that, since g- occurs in Tibetan transcriptions of both 'level' and 'rising tone' Tangut syllables.

Now I wonder if g-y- and y- were attempts to write a Tangut Gy-. If so, then:

- UNQUOTE was Gyï 2.28

- UNQUOTE sounded like the NTC pronunciation of 夷 'barbarian' to

kwə1 le2 ryi r2 phuu2

author of the Pearl. Perhaps he pronounced 夷 'barbarian' with a Tangut accent as Gyi.

- Thus he used 夷X to represent Gy-initial Tangut syllables.
In fact, according to Gong's reconstruction, all the Tangut syllables transcribed as 夷X began with Gi-, not Gy-:

Level tone: Gie 1.9, Giey 1.34, Giew 1.44, Gier 1.78

Rising tone: Giẹy 2.53, Giẹ 2.59, Giẹe (sic) 2.59

So perhaps 夷 'barbarian' stood for Gi-, assuming that Gong's reconstruction is correct.

In Guillaume Jacques' index of Gong's reconstructions from Li Fanwen (1997), there are no Gi-s, but there are three yi-s. Were these yi-s [Gi]-s, at least for some speakers (e.g., the author of the Pearl)? Were Gi and yi in complementary distribution? Did one only occur where the other did not, and vice versa? Here's a list of all the Gi- and yi-syllable types in the index:

RhymeGi-Rhyme numbersyi-Rhyme numbers
-iG- did not occur before rhymes with i-type monophthongsyi2.9 (yi 1.30 error for 1.30?)
-iyyiy1.36 (yiy 1.39 error for yiiy 1.39?)
-iryyiry1.74, 2.68
-iãGiã1.25, 2.23y- did not occur before rhymes with i-diphthongs (i.e., sequences of i with other vowels)
-ieGie1,9, 2.8
-iẹGiẹ1.59 (sic for 2.59?) 1.66, 2.59
-iẹe?Giẹe2.59 (sic; prob. Giẹ)
-iwəyGiwəy1.41, 2.36

Apparently there was no distinction between G- and y- before i. It's possible that there were 'Gi-speakers' without yi, 'yi-speakers' without Gi, and 'split pattern speakers' who had y- before i-type monophthongs but had G- before i-diphthongs. The author of the Pearl could have been a Gi-speaker.

This nonrandom pattern of distribution - assuming it is real - may imply that pre-Tangut Gi- and yi- merged. (It does not necessarily imply a merger of G- and y- before other vowels!) I hope it's not just an artifact of Gong's reconstruction.

Next: G-enitive guidance?

*Greek γ even sounds like y before front vowels in modern Greek pronunciation. Joseph (in Comrie 1987: 416) describes its value in that environment as [y], but Wikipedia describes it as a voiced palatal fricative. Maybe Tangut G was a voiced palatal fricative before front vowels (and -y-) and a voiced velar fricative elsewhere: cf. how Greek γ is y-like before front vowels but [G] elsewhere. Y-CLUSTERS? Y NOT?

I've heard of rGyalrong for years but I knew almost nothing about it until I examined the work of Guillaume Jacques. Guillaume did fieldwork on Japhug rGyalrong (JG) and wrote his doctoral dissertation on its phonology. One characteristic of JG that jumped out at me was its y-clusters (j-clusters in IPA; Jacques 2003: 43, 56):


(Note that not all JG consonants are attested with y-: e.g., there are no syllables with yb-, yth-, ydz-, ykh-, yg-.)

Until last night, all the examples I had ever seen of such clusters in JG had followed prefixes. In Guillaume's JG-Tangut rhymes paper, I found:

ypa 'snow' (cf. Tangut wyị [< -a] 1.67 'snow')

ywaR 'leaf' (cf. Tangut bạ 2.56 'leaf')

yme 'tail' (cf. Tangut myiiy 1.39 'tail')

kë mU yphët 'to vomit' (cf. Tangut wya 1.19 'to vomit')

ylëV 'vapor' (cf. Tangut lwew 1.43 'smoke')

ytsi 'pillar' (cf. Tangut dzyị 2.60 'pillar')

ytshi 'give to drink' (cf. Tangut tyị 1.67 'nourish')

ytU 'collect' (cf. Tangut duu 1.5 'amass')

kë nu ymUt 'forget' (cf. Tangut myị̈ 2.61 'forget')

sU yno 'grass' (cf. Tangut nyu 1.3 'name of a vegetable')

tU ymngo 'dream' (with y-m-ng-!) (cf. Tangut myiiy 1.39 'dream')

ypGom 'ice' (with y-p-G-!) (cf. Tangut 'wọ 1.70 'ice')

kU yngu 'promise' (n.) (cf. Tangut ngwe 2.7 'promise' [n.])

kU ypum 'thick' (cf. Tangut wọ 1.70 'thick')

However, Guillaume told me that y-clusters may even occur in absolute initial position: e.g., yla 'yak-cow hybrid'. (You can hear it in "The Three Sisters". Be patient and wait for a QuickTime file to load. Then go to line 148 and press 'play'.)

So for a moment I wondered if 夷 'barbarian' in 夷X transcriptions of Tangut in the Pearl (listed here) could have represented syllables with y-initial clusters.

Then I remembered that all the Xs in 夷X-transcriptions represented Chinese velar-initial syllables. If 夷 were y-, then 夷X stood for Tangut yk-syllables. But the absence of transcriptions with 夷 plus non-velar Xs is suspicious. If Tangut had yk-, wouldn't it also have other y-clusters like yt-, yp-? This is a variation of the argument I used against interpreting 夷X as ik- or Gk-: positing one sequence implies the existence of others, and if those others can't be found, then perhaps that sequence never existed.

Obviously, 夷 must have stood for something. It either represented a feature to be added to the following velar stop (i.e., voicing), or it stood for an initial (i.e., it was half of a fanqie formula). The consistent use of 夷 before velars leads me to favor the former interpretation.

Next: What Greek letter does y resemble? *yw- is the only cluster that isn't exotic to me since it might have been in Middle Chinese: e.g., 唯 ywi 'only'. yw- also resembles a labiopalatal glide [ẅ] (IPA [ɥ]). None of the other clusters have any parallel in any language I've personally encountered. Some JG y-clusters come from Proto-rGyalrong (PG) l-clusters:

PG lp(h)- > JG yp(h)-

PG lm- > JG ym-

PG lk- > JG yk-

PG lq- > JG yR-

but others are retentions of Proto-rGyalrong clusters (yt-, yn-, yng-, yl-). See Jacques (2003: 333-334). LIP SOUNDS LIGHT (PART 12)

Nishida and Arakawa (according to Kotaka) reconstruct an f in Tangut. If Tangut had an f, tangraphs pronounced with f (i.e., Homophones Chapter II tangraphs) should be used to transcribe Chinese f.

Grinstead (1972: 188) has compiled a list of tangraphs representing Chinese syllables pronounced with f- in modern standard Mandarin (and in modern northwestern dialects, though I am not sure if Grinstead was aware of that).

Although there is no guarantee that these syllables were also pronounced with f- in Tangut period northwestern Chinese (TNC), I will make that assumption since as far as I know, syllables of their initial class were not used to transcribe Indic in the late pre-Tangut period, indicating that they had initials absent from Indic (e.g., f- or pf-).

(They do appear in early transcriptions of Indic because they originally had stop initials: e.g., 梵 Md fan, modern NW fã, fæ̃, fẼ < Old Chinese *bramh for brahma and vaaM [Coblin 1994: 293].)

(I am going to star all reconstructions from now on except for Tangut because by definition all Tangut forms not in tangraphy are reconstructions.)

This initial class was also inconsistently transcribed in Tibetan during the late pre-Tangut period, implying that they had an initial (*f or *pf) absent in Tibetan: e.g., 梵 was transcribed as bam, Hpwam, and wyam (and in Khotanese Brahmi as hvaM:mä; I need to study what the colon means).

It is simplest to assume that these f- remained (more or less) unchanged in TNC between the late pre-Tangut period and modern times:

Late pre-Tangut NW Chinese *(p)f- > TNC *(p)f-? > modern NW f-

That of course assumes that TNC dialects are ancestral to modern NW dialects. There is no guarantee that they are.

Now for Grinstead's list which I've expanded with other information (all columns but the first two):

Md pron. of Chn syl
TangraphHomophones chapterRhymeTangraphic TelecodeGrinstead's glossNishidaSofronovGong
fa, fenVIII1.183620FLOWERxwa (not fạ̣ which he reconstructed for TT2467 FLOWER)xwâxiwa (not two syls but xia with labial medial; I'd prefer xwia)
fan1.253126BRAHMAxwə̃xwẫxiwã (see note above)

1.565559PERFUME; prob. a loanword from Chn 香; see here; (Nishida [1966: 347] has 軒 'high'; thinks could be borrowing from Chn)xyãxyõxyow
fang, fen, feng2.434019transcription char for NTC equivs of Md feng, fan, xiong (: NTC x-) (Nishida [1966: 453] has 部姓 'clan surname')Gaxyõxyow
fangUnknown; resembles TT3537 but has TT2850 終 END to 2.48 on right side
feiI1.113450transcription char for NTC equivs of Md pei, pi (: NTC ph-) (Nishida [1966: 431] has 壁 'wall', thinks is borrowing from Chn)pipipyi
feng1.490633transcription char for NTC equivs of Md feng, bo (: NTC p-) (Nishida [1966: 378] has 封域 'feudal territory'?)poHpopo
fengII1.774844AUTHORITY (Nishida [1966: 496] has Jpn 豐たか yutaka 'abundant', equiv to Md feng!)mvIrwewer

fu VIII1.13552transcription char for NTC equiv to Md fu (Nishida [1966: 436] has 部姓 'clan surname')? (sic)xuxu
fuII1.310862SUPPRESS? (sic)wəə

This table is not complete: e.g., the TNC equivalent of 飛 Md fei was transcribed in the Pearl (line 165) as TT3016 (Homophones Chapter II; see part 10 for reconstructions).

Four assumptions:

1. If Tangut had an [f], it would most likely be in Chinese loans, though an indigenous [f] would also be possible.

2. Such an [f] would most likely be in the phonetic inventory of educated Tangut-Chinese bilinguals. (If [f] were indigenous, then all speakers of Tangut dialects with that sound would have it.)

3. The Tangut dialect recorded in Homophones was a educated variety. If any variety of Tangut would have an [f], it would be that variety.

4. Tangut syllables with labiodental [f] would be assigned to Homophones chapter II ("Light Lip Sounds").

With those four points in mind, I conclude that the Homophones dialect probably did not have [f]. If it did, all of those tangraphs used to transcribe Chinese f-syllables would be in Chapter II. But they are also in Chapters I ("Heavy Lip Sounds" = labials) and VIII ("Gullet Sounds" = 'gutterals' including velar fricatives). Moreover, they are also used to transcribe Chinese labial- and velar-initial syllables. It seems that Tangut p-, w- (phonetically [v]? see part 7), x(w)- were the closest available approximations of Chinese f.

Unless the dialects underlying the transcriptions were even more sinified than the Homophones dialect, I don't think they had [f] either. If speakers used 'authentic' Chinese pronunciations with [f] when reading Chinese names, then they were pronouncing (or at least trying to pronounce) Chinese (or, more precisely, TNC), and Chinese-only phonemes cannot be said to be part of their Tangut phonetic repetoire. We don't say that Anglophones who pronounce, say, French words in the French manner have nasalized vowels in their English phonetic repetoire.

It would be interesting to look at individual texts to see if they have consistent patterns for transcribing Chinese f: e.g., do some texts exclusively use Chapter II tangraphs whereas others exclusively use Chapter VIII tangraphs? The Pearl used both Chapter II and VIII tangraphs (see part 11). Such patterns would indicate different translators' transcriptional preferences.

Next: Where does Arakawa reconstruct f (at least according to Kotaka's tangraph list)? SYLLABIC YOD?

I have been looking into modern northwestern Chinese dialects to help me interpret the Chinese transcription evidence for Tangut. It is not clear what the relationship between those dialects and the Tangut-period Chinese dialects is: descent or superstratum-substratum?

Descent: Modern dialects are descended from the Tangut period ones.

Superstratum-substratum: Modern dialects are not descended from the Tangut period ones but were acquired by speakers of those dialects long ago, and traces of the substratal dialects may still remain.

Of course, there is no question that the Tangut period and modern dialects are both Chinese and therefore both related. The question is whether they are directly or indirectly related: 'mothers and daughters' or 'aunts and nieces'.

The Chinese transcriptions of Tangut in the Pearl contain occasional disinographic glosses (2SGs) in addition to monosinographic glosses (1SGs). Here are all ten examples of 2SGs containing 夷 'barbarian':

Pearl line2SGTibetan transcription from Nevsky (1926)TangraphHomophones chapterRhymeTangraphic TelecodeGrinstead's glossNishidaSofronovGong
064, 296夷格, 夷隔2.593772INJUREGkỊě'ệiGiẹ
176夷 皆dgye? (#61)1.342087TRUTHGkyE'êiGiey
196, 273夷隔, 夷客n/a1.443861LEARNGkewxêUGiew
211夷 隔dghi (#156)2.590121HANDBELLGkỊě'êi (sic for 'êi)Giẹe (sic for Giẹ?; no long tense vowels)
224夷 耿n/a2.531977TRANSPORTGkyẸ'ậiGiẹy
225, 356夷隔Hgi (#118)1.93409FORCE/

夷 'barbarian' also occurs thrice as a 1SG:

Pearl line1SGTibetan transcription from Nevsky (1926)TangraphHomophones chapterRhymeTangraphic TelecodeGrinstead's glossNishidaSofronovGong
313g-yi(H), g-yyi, yi(H) (#274)2.284979UNQUOTE'yï'I

The indisputable facts are:

1. All of the above graphs are listed in Homophones chapter VIII ("Gullet Sounds").

2. All the tangraphs belong to different rhymes.

3. There is no (obvious) correlation between Tangut tones and Chinese tones.

4. All the Tibetan transcriptions for the tangraphs transcribed as 夷X have complex initials: dgh-, dgh-, Hg-.

5. 夷 was used to transcribe Indic yin, yii in the 5th century. In modern Chinese languages, 夷 is homophonous with 以已異 if tone is ignored. In modern northwestern dialects, these three sinographs are pronounced as (ignoring tone):

y (no vowel!?; not IPA [y], but IPA [j]!): Xining

Hence the title: what is a syllabic yod? I thought a syllabic yod was the vowel [i]. How does syllabic [j] differ from [i]? Does a syllabic waw [w] distinct from the vowel [u] also exist?

i: other dialects: Dunhuang, Lanzhou, Xi'an

Presumably 夷 is also y or i in those modern dialects. William Wang's modern Chinese language database lists the Xi'an reading as i. No readings are available for other NW dialects but the sinograph is generally pronounced as (y)i throughout China (Wenzhou Hi with a voiced glottal fricative is an exception).

6. Pre-Tangut period Tibetan transcriptions for the second sinographs in the 2SGs have velar initials:


ke (Khotanese Brahmi: keyi, ka)

No transcriptions are available for the other second sinographs, but all other evidence points toward k- for 格 and 耿 and kh- for 客.

Thus 夷X sequences might have been pronounced as something like i-k(h)... in Tangut period NW Chinese. Since there are no 夷X sequences with non-velar X-s, it is unlikely that these represent Tangut disyllables. I know of no language in which initial i- can only precede velars. A language with ik- would also have i- before other types of consonants.

Nishida, Sofronov, and Gong interpreted 夷X very differently:

Nishida reconstructed a cluster Gk- (voiced velar fricative + voiceless velar stop).

Sofronov reconstructed a glottal stop '-.

Gong reconstructed a voiced velar fricative G-.

Nishida's Gk- is an isolated cluster in his reconstructed Tangut consonantal system (1964: 149) which contains no other voiced fricative + voiceless stop (or affricate) clusters: e.g., Vp-, ðt(s)-, zhch-. Usually phonological systems are, well, systems: the presence of one cluster may imply the presence of others. An isolate like Gk- is suspicious.

Sofronov's glottal stop '- is difficult to reconcile with the evidence. If the initial were a simple glottal stop, why write it with 夷 which probably did not begin with glottal stop? Initial glottal stop is correlated with the 'yin level' tone class in Chinese, and 夷 belongs to the 'yang level' tone class which is characterized by the absence of glottal stop. There were sinographs with pronounciations similar to the second elements of the 2SGs which probably had initial glottal stops: e.g., 厄. Why weren't they used? And if TT3861 were xeU with x-, why transcribe it as 夷隔 (ik...) or 夷客 (ikh...) rather than with an x-initial sinograph?

Gong's G- begs a similar question: why not transcribe the Tangut syllables with x-initial sinographs instead of 夷- (with y-/i-)? Example:

All other evidence indicates that 核 (not an obscure sinograph) was like 隔 or 客 except for its velar fricative initial.

Therefore if TT3861 LEARN were really xeU (Sofronov) or Giew (Gong), it could have been written as 核 instead of as 夷隔 or 夷客 or even as 夷核 ('read like 核 but with voicing of 夷'). Why wasn't it?

(Moreover, nothing leads me to believe that either 隔 or 客 end in anything like -U or -w, but that's another topic. In modern Xi'an, 隔 is kei and 客 is khei and my hypothetical 核 is xE.)

Was there some conscious decision to use 夷 as a phonetic symbol for G- since it was already being used to write TT4979 UNQUOTE which might have been (gə-)yi? (The function of final -H is uncertain.) If so, then 夷X meant 'read with G- plus rhyme of the following sinograph'.

Obviously, none of the three reconstructions account for the preinitials d- and H- in the Tibetan trasncriptions.

Nishida reconstructed a distinction between G- and Gk- whereas Sofronov reconstructed no G- at all and Gong reconstructed a G- corresponding to Nishida's G- and Gk-. I favor some sort of velar instead of Sofronov's glottal stop because of the Homophones chapter VIII word

TT3619 HEAD/TOP 1.4


in Tibetan as dguH, bgu (Nevsky 1926 #110)

in Chinese as o吳 with a subscript o indicating some modification of the reading

its homophone 吾 was transcribed in Tibetan as Hgo [nggo] in the pre-Tangut period but its modern Xi'an reading is u; not sure what the Tangut period reading was: u or something closer to nggo?

and reconstructed as

Nishida: GuH

Sofronov: 'u

Gong: Gu

(Also cf. Japhug rGyalrong tU ku 'head' [Jacques 2003: 23].)

If the word were simply 'u, why not transcribe it in Tibetan as u, or even as d-u or b-u [for də'u and bə'u]?

Even if o吳 were like u as in modern Xi'an, it is not hard to imagine how Tangut Gu 'head' could be transcribed as u. It is harder to imagine how a Tangut 'ệi 'injure' could be transcribed as 夷格 ik... or 夷隔 ik...

Gong's G also correlates with Japhug rGyalrong (JG) velars and uvulars and my lenition hypothesis (see "Beyond d-bris"). It is dangerous to reconstruct sounds on the basis of a related language, but at least we have the Tibetan transcriptions which are independent of JG.

Of course, there could have been two types of Tangut dialects: one with velars preceded by prefixes and another with unprefixed glottal stops (or zero initials, reduced even further from G-?). There is no need to assume that all data point to one and only one kind of Tangut.

Next: Another interpretation of the 夷 'barbarians'. BEYOND D-BRI(S): TANGUT LENITION THEORY (PART 1)

If Tangut (')w- < earlier p-, ph-, mb- after prefixes (as seen in part 2 of "d-bri(s)"), could other earlier Tangut stops and affricates have lenited in the same position?

C-t-, C-th-, C-nd- > (')y-?

C-ts-, C-tsh-, C-ndz- > z-?

C-ch-, C-chh-, C-ñj- > zh-?

C-k-, C-kh-, C-ngg- > G-?

If this hypothesis is correct, we should expect to see Tangut (')y-, z-, zh-, G- correspond to Japhug rGyalrong (JG) stops and affricates. Looking at Jacques 2003, I found the following correspondences:

Tangut initialJaphug rGyalrong of gDong-brgyadnumber of instancespages of Jacques 2003notes
y-Gurzh- (sic)17Tangut has retroflex vowel corresponding to JG r-
'y-kU rc-111
qa zh-121JG zh- < proto-rGyalrong y- (Jacques 2004: 144)
kU zh-121
tU zh-121
kU rR-123Tangut has retroflex vowel corresponding to JG r-
z-kU rts-213, 14
kU zr-117
G-kë sq-27, 8
ta q-112
të fk-112
të kh-118
të mk-219
tU k-123


1. There are no cases of rGyalrong dental stops corresponding to Tangut (')y-. Therefore Tangut dental stops did not lenite to (')y-.

2. Tangut (')y- may partly originate from lenited r- (assuming that JG r- is conservative).

JG rzh- < proto-rGyalrong rə-ly-, ry- as well as rzh-.

The proto-rGyalrong ancestor of JG Gurzha 'hundred' (: Tangut yir 2.27 'hundred') may have been rya; cf. Written Tibetan brgya 'hundred' < earlier b-rya? Cf. Starostin's Proto-Sino-Tibetan (p-)ryaa. (Although I don't endorse his PST, his data may be worth a look, assuming it is accurate.)

(But Matisoff 2003 analyzed brgya as b-r-gya. Proto-rGyalrong had no gy- [Jacques 2004: 332], and I'm not sure what pre-PGR gy- became in PGR. Tibetan gy- corresponds to the palatal stops c- and -j- in GJ.)


TT5770 HUNDRED yir 2.72 < C-rya? (b-rya like pre-WT?; also cf. Old Chinese 百 prak 'hundred' whose -k can't be explained)


TT2425 AUTHORIZE ryar 1.92 < rya (no pre-initial; vowel affected by autoretroflexion)

(It's not clear why some pre-Tangut a > yi and some didn't. Maybe there were two kinds of a-s: a back type that remained low and a front type prone to raising. Or raising was somehow triggered by the vowel of a presyllable:

C-high vowel + rya > yir


C-non-high vowel + rya > yar

Perhaps early Tangut had two or more presyllable vowels: cf. JG which has high, mid, and low presyllable vowels: -U-, -ë-, -a-. (Before I learned of Japhug, I proposed two presyllable vowels for Old Chinese in a paper I submitted in 2005.)

JG rc- < proto-rGyalrong rky-?

root of GJ kU rcat 'eight' < rkyat < borrowed from Tibetan since -k- reflects Tibetan 'intrusive -g-'?; cf. WT brgyad < b-ryat? (in turn, cf. Old Chinese 八 pret 'eight' < p-ryat?) Perhaps

TT1632 EIGHT 'yar 1.82 < C-rya? (b-ryat like pre-WT?)

3. The glottal stop in Tangut 'y- might sometimes reflect earlier stops still present in the rGyalrong prefixes qa-, kU-, tU-:

C-y- > 'y-

y could not lenite any further after a preinitial since it was already barely a consonant. (Glides like y border on being vowels.)

JG rR- is from proto-rGyalrong rq-* (Jacques 2004: 330). Assuming that PGR was conservative, perhaps

TT0445 ROUGH 'yor 2.81 < qrom < rqom (cf. JG kU rRom < PGR rqom?)

4. Tangut z- may either be original (when it corresponds to JG z-) or derived from a lenited alveolar affricate: e.g.,

TT2385, TT2384 LEOPARD zerw 2.78 < C-tsew? (cf. JG kU rtsëG)

But Grinstead (1972: 111, 214) defined TT2385 as EAST!

The Pearl only used TT2384. Is TT2385 a ritual language word for LEOPARD that also doubles as a tangraph for a homophonous word EAST?

5. Tangut sh- might have lenited to zh- after a preinitial, but there is no evidence for this guess. The apparent JG cognate shëG of TT0834 JUNIPER zhyiw 1.46 has no presyllable, but maybe the pre-Tangut word did: C-shyiw > zhyiw.

6. There is a clear correlation between JG presyllable (+ preinitial) + k-, kh-, q- and Tangut G-. Therefore it appears that Tangut velar and uvular stops lenited to G- following a prefix that was later lost (at least in some dialects).

Next: Can this be confirmed by the Tibetan transcription evidence?

*Another source of JG rR- is rb- before rhymes ending in -q (Jacques 2004: 328).

Copyright 2002-2007 Amritavision