(I'll cover over-Gg-rowth in part 6.)

In part 4, the Tangraphic Sea fanqie appeared to support Nishida's reconstruction of Gk- in Tangut:

Gk-syllable < G(k)-syllable + k(h)-syllable

However, the overall fanqie pattern is

(C2)C1-syllable < C1-syllable + C3+syllable

C1 = initial of what David Boxenhorn calls the 'kernel' syllable

C2 = preinitial

C3 = initial which may have no relevance to the syllable being spelled

rather than

C2C1-syllable < C2-syllable + C1+syllable

(and if this pattern were correct, how could it be applied to syllables without preinitials? Would ka < kV + ka, whereas Gka < G(k)a + k(h)a?)

Suppose that the Gk-pattern for fanqie were universal. Given the Tibetan transcriptions

d-wi, dbi, biH (Nevsky 1926 #240)


TT2413 DO 1.10

its fanqie should have the following initials:

dw- or db- < d- + b- or w-

with d- = C2 = preinitial

from Homophones chapter III (dentals)

and -b-/-w- = C1 = initial

from Homophones chapters I (labials) or II (labiodentals; w may have been [v])

But its fanqie is

TT4242 from Homophones chapter II (labiodentals) +

instead of the expected Homophones chapter III (dentals)

TT3450 from Homophones chapter I (labials)

The final speller matches the prediction* but the initial speller certainly does not. And TT2413 DO 1.10 is listed in the labiodental chapter of Homophones, not the dental chapter (despite its dental-initial transcriptions d-wi and dbi). In Gong's reconstruction, the fanqie is:

wyi 1.10 < w 2.27 + pyi 1.10

Given a correlation between Tangraphic Sea initial spellers and Homophones initial classification, we would expect a further correlation between them and Tibetan preinitials. But matches between Homophones classes and Tibetan initials are rare:

Homophones chapterExpected preinitials in Tibetan transcriptionsActual attested preinitials in Tibetan transcriptions from Nishida (1964)
I (labials)labials, H- (if representing a homorganic nasal [m])d-, H-, r-
II (labiodentals)labials, H- (if representing a homorganic nasal [m]); the Tibetan alphabet has no letters for labiodentalsd-, b-, l-
III (dentals)dentals, H- (if representing a homorganic nasal [n])g-, gr-, b-, H-, r-, Hr-, l-, s-
IV ('retroflexes'?)dentals or palatals, H- (if representing a homorganic nasal [n] or [ñ]); maybe even the Tibetan letters for Sanskrit retroflexesh- (sic)
V (velars)velars, H- (if representing a homorganic nasal [ng])d-, b-, br-, m-, H-, r-, h- (sic)
VI (alveolar fricatives and affricates)alveolar fricatives and affricatesg-, b-, H-, r-
VII (palatals) palatals, H- (if representing a homorganic nasal [ñ])g-, b-, H-, r-, h- (sic)
VIII (velar fricatives and glottals)velars, h-, H-, '-g-, d-, b-, r-
IX (liquids, rhotics - and the palatal zh!)l-, lh-, r-, zh-k-**, g-, gr-, b-

Next: Instead of part 6 ... the most shocking Tangut article ever. This is not hyperbole.

* I suspect that the root initial was originally p-, though the choice of a p-initial final speller may have been coincidental.

Guillaume Jacques (2003: 6) proposed that TT2413 DO wyi 1.10 is cognate to gDong-brgyad rGyalrong kë pa 'to close' (but 'to do' in other rGyalrongic languages?). If that is correct, then I suspect that Tangut w lenited after a preinitial that was lost in some dialects:

Proto-Tangut *t-pyi 'to do'

(I am presuming that all Tangut dialects reflect the *-a > -yi shift in this word.)

Conservative dialect *t-pyi, *t-wyi [tvi] (transcribed in Tibetan as dbi, d-wi)

Innovative dialect wyi [vi] (transcribed in Tibetan as biH; function of -H still unclear)

(I am leaving Gong's reconstructions unstarred.)

I am guessing that the Tangut preinitial was t- but written as Tibetan d- since

- I am assuming that Tangut preinitials were like rGyalrong preinitials (i.e., voiceless stops were permitted but voiced stops weren't)

- Tibetan spelling allows voiced stops but not voiceless stops as preinitials

Other interpretations of Tibetan d- are possible: e.g., Tangut [G] (suggested by Wolfenden 1934) or [ð]. The precise phonetic value of the preinitial is less important than presuming its very existence.

I wrote [tvi] and [vi] and I would prefer to write *t-pi, etc. without *-y- because I think Gong's -y- is often redundant. I will explain why in a future post.

The use of Tibetan b-, ww-, wh- for Tangut w- suggests that Tangut w- was not a simple [w]. It may have been a labiodental [v], explaining why w-syllables were listed in the labiodental chapter of Homophones.

Guillaume also proposed a similar cognate pair for 'year':

gDong-brgyad rGyalrong U xpa : TT1605 wyi 1.10

No Tibetan transcriptions for TT1605 are available in Nevsky (1926).

gDong-brgyad rGyalrong xp- is from Proto-rGyalrong *kə-p- (Jacques 2004: 275). Perhaps the proto-Tangut word for 'year' was *k-pyi. (But without Tibetan evidence, other Tangut preinitials cannot be ruled out: e.g., *t-pyi might also be possible.)

The initials of the fanqie for TT1605 YEAR (w-, sh-) do not match the proposed *k-p- sequence:


TT1605 wyi 1.10 < TT0577 wyu 1.2 + TT3699 shyi 1.10

rather than

k-pyi < k- + pyi

It is clear that this Tangraphic Sea fanqie are non-etymological. There is no other evidence (e.g., a Tibetan transcription) for a root initial sh- in YEAR.

Even if one rejects Gong's reconstructed initials, one can see that the Homophones chapter numbers for the fanqie tangraphs do not support the etymological fanqie hypothesis:

II (w-) < II (w-) + VII (sh-)

rather than

? (k-p-) < V (k-) + I (p-)

The use of different initial spellers for DO wyi 1.10 and YEAR wyi 1.10 may or may not reflect different preinitials: e.g.,

syllables with *t-p- in dialect 1, w- in dialect 2: spell with TT4242 wyï 2.27 (< *t-p-?)

syllables with *k-p- in dialect 1, w- in dialect 2: spell with TT0577 wyu 1.2 (< *k-p-?)

I would like to investigate this possibility later. If the above hypothesis is correct, there should be no overlap between the fanqie of t-p- and k-p-syllables: e.g., YEAR (< *k-p-?) should not be used as an initial speller for syllables used as initial spellers for *t-p-syllables. Unfortunately, without more Tibetan material, it is difficult to determine what preinitials, if any, most Tangut syllables had (at least in some dialects). Nevsky (1926) only lists transcriptions for 334 syllables which are only about 6% of the total of c. 5,800. (Note that this total refers to tangraph readings and not to unique, nonhomophonous syllables; I know of no estimate for the latter.)

** I regard k- as a preinitial in the Tibetan transcription kli for

TT0100 ONE 1.43

Other Tibetan transcriptions are gli and gliH (Nevsky 1926 #1).

The Chinese transcription of ONE in the Pearl is 婁 with initial l-. (Its rhyme in the Chinese dialect of that time and place is unknown.)

The northwestern pronunciation of 樓, presumably a homophone of its phonetic element 婁, was transcribed in Tibetan as leHu and liHu in the pre-Tangut period.

(Tibetan spelling does not allow syllables to end in -w, so VHu = foreign -Vw. There is no reason to assume that the vowel of 樓 had a breathy quality or that 樓 had a disyllabic Chinese reading.)

In modern northwestern dialects, 樓 is pronounced with nonfront vowels (Coblin 1994: 266):

Xining lU

Dunhuang lou

Lanzhou nəu

(Early 20th c. Lanzhou ləU, as heard by Karlgren)

Early 20th c. Pingliang lu (as heard by Karlgren)

Xi'an lou (Karlgren heard the same thing c. a century ago)

I cannot find any modern northwestern readings of 婁 (a low-frequency sinograph).

I conclude that the Tangut word for 'one' may have been something like *k-lïw or *k-ləw with a nonfront vowel. (Did k- lenite to G- before l- in some dialects?) ï and ə are similar to the back unrounded vowel U (IPA [ɯ]).

Although Gong (1997) reconstructed lew with a mid front vowel, I cannot explain why lew would have been transcribed with Tibetan i rather than e. I would expect the Tibetan transcription of lew to be ལེའུུ le(Hu), not kli, gli, gliH.

(Can anyone see Tibetan letters? I'm trying out Tibetan in Unicode for the first time on this blog.) CZECH OUT THESE CLUSTERS (PART 4): THE G-KOP-OUT

Fanqie spellings in Chinese - and presumably also in Tangut - are based on the following premise:

Given a fanqie formula XY

in which X is the initial

and Y is the final

all fanqie spellings starting with X (XA, XB ...) should represent syllables with initial X-

all fanqie spellings ending with Y (CY, DY ...) should represent syllables with final -Y

Therefore given the first fanqie formula in the Tangraphic Sea

TT5504 UNCLEAN pe 1.8 +

TT1960 HIDE lu 1.1

in which p- is the initial

and -u 1.1 is the final

all fanqie spellings starting with TT5504 UNCLEAN pe 1.8 should represent syllables with initial p-

all fanqie spellings ending with TT1960 HIDE lu 1.1 should represent syllables with final -u 1.1

Let's test these predictions using Gong's reconstruction:

All fanqie spellings starting with TT5504 UNCLEAN pe 1.8 represent syllables with initial p-: pu 1.1 and po 1.49.

The above fanqie formula (pe + lu 1.1) was used to indicate the pronunciation (pu 1.1) of the four homophones

TT2876 (a surname) pu 1.1

TT2877 (a kind of bird) pu 1.1

TT4868 CHOKE pu 1.1

TT3249 WEED pu 1.1

All fanqie spellings starting with TT2876 (a surname) pu 1.1 represent syllables with initial p-:

pe 1.8 (TT5504 UNCLEAN, its own initial speller), pẽ 1.15, pa 1.17, pã 1.24, 1.27, pey 1.33, pow 1.54, r 1.84.

TT2877, 4868, 3249 were not used as fanqie initial spellers.

All fanqie spellings ending with TT1960 HIDE lu 1.1 represent syllables with final -u 1.1:

pu 1.1, phu 1.1, tu 1.1, thu 1.1, tshu 1.1, su 1.1

Both fanqie spellings ending with TT2876 (a surname) pu 1.1 represent syllables with final -u 1.1:

tsu 1.1, zu 1.1

TT2877, 4868, 3249 were not used as fanqie final spellers.

Both fanqie spellings ending with TT3392 tu 1.1 (whose final speller is TT1960 HIDE lu 1.1) represent syllables ending in -u 1.1:

nu 1.1, ku 1.1

The fanqie spelling ending in TT3805 (a surname) su 1.1 (whose final speller is TT1960 HIDE lu 1.1) represents the syllable gu 1.1.

In reality, it is actually Gong's reconstruction that should be tested against the spellings, not the other way around. The fanqie are hard facts; Gong's reconstruction isn't.

A reconstruction should reflect equations such as these:

Initial X: Represented by equivalent fanqie spellers TT5504 = TT2876

Final Y: Represented by equivalent fanqie spellers TT1960 = TT2876 = TT3392 = TT3805

Gong has reconstructed X as p- and Y as -u. (The rhyme number 1.1 really isn't a reconstruction, as the above tangraphs are listed in the level [first] tone rhyme 1 section of the Tangraphic Sea.)

In Tangraphic Sea, TT3409 and TT4384 have the initial speller TT5688, which in turn has the initial speller TT1070. In Gong's reconstruction, all four all have the same initial, but Nishida reconstructed three different initials!

Tangut TelecodeGong (1997)Nishida (1964-1966)
TT3409Gie 1.9GkIě 1.9
TT4384n/a; presumably would be GkIě 1.9 since these are in the same homophone group in Homophones as well as in Tangraphic Sea
TT5688Gyu 1.3'yyuH 1.3
TT1070Gyị̈ 1.69Gị̈ 1.69

The initial speller for TT1070 is TT5712 which still has a G in Gong's reconstruction (Gyu 1.3) but not in Nishida's ('yyuH 1.3).

Using Gong's reconstruction, the following statement sounds reasonable:

Gie 1.9 has the same initial as Gyu 1.3 which in turn has the same initial as Gyị̈ 1.69 whose initial is the same as Gyu 1.3. In short, G- = G- = G- = G-.

Using Nishida's reconstruction, the following statement sounds unreasonable:

GkIě 1.9 has the same initial as 'yyuH 1.3 which in turn has the same initial as Gị̈ 1.69 whose initial is the same as 'yyuH 1.3. In short, Gk- = 'y- = G- = 'y-!?

Changing 'y- to G- might help a little. Let's look at the fanqie for 'yyuH 1.3 again, but with Gy- instead of 'y-:

TT5688 GyuH 1.3 < TT1070 Gị̈ 1.69 + TT1548 (no Nishida reconstruction) 1.3

Gong reconstructed TT5688 as Gyu 1.3 with G-.

TT1548 is tyu 1.3 in Gong's reconstruction which is equivalent to tyuH 1.3 in Nishida's reconstruction.

Now the formula can be rewritten as Gk- = G- = G- = G-. Still not perfect, but better. How could Nishida explain the -k-? Notice that the fanqie for the level tone tangraphs transcribed in the Pearl with 夷 'barbarian' followed by a velar-initial sinograph all have velar-initial final spellers:

TT3409 GkIě 1.9 < TT5688 GyuH 1.3 + TT0560 unknown velar (Gong reconstructed k-) + - 1.9

TT2087 GkyE 1.34 < TT0875 (no Nishida reconstruction; Gong reconstructed Gyï) 1.29 + TT1328 khe 1.34 (sic! -e is Nishida's value for 1.37!)

TT3861Gkew < TT3409 Gk 1.9 + TT5271 kew 1.44

TT5476 GkIěr 1.78 < TT5576 Gur 1.75 + TT3928 unknown velar (Gong reconstructed k-) + -r 1.78

Stripped of detail, the fanqie spellings are:

TT3409: Gk- < G- + k-

TT2087: Gk- < G- + kh-

TT3861: Gk- < Gk- + k-

TT5476: Gk- < G- + k-

Is it a coincidence that 夷 + velar transcriptions correlate with G(k)- + k(h)- fanqie spellings?

In Chinese-style fanqie, the initial of the second speller has nothing to do with the initial of the syllable being spelled: e.g., Go + ke = Ge, not Gke. But there is no guarantee that Tangut fanqie had to work the same way. Were there attempts in Tangut fanqie to represent complex initials by using the first speller to represent the preinitial and the second speller to represent the rest of the syllable: e.g., Go + ke = Gke? If that were the case, we would expect

preinitials in Tibetan transcriptions to match initials of Tangut fanqie initial spellers

initials in Tibetan transcriptions to match initials of Tangut fanqie final spellers

But the preinitials in Tibetan transcriptions generally match nothing in the Tangut fanqie, which is perhaps why they have been ignored for so long. (The possibility that the preinitials reflect a Tangut dialect more conservative than the Sinified dialect preserved in the dictionaries has been overlooked.)

Unfortunately, there is very little Tibetan transcription data for Nishida's Gk-syllables:

RhymeTibetan transcriptionNishida's reconstructionGong's reconstructionFanqie (initials only)
1.9Hgi, dghiGkIěGieG- + k-
1.34dgheGkyEGieyG- + kh-
2.59Hge, dghiGkỊěGiẹ(not available)

H- - assuming it was a voiced fricative [H] and not a nasal [ng] that assimilated to a following g(h)- - sort of matches G-, but d- doesn't, unless it had lenited to [G] in preinitial position. g(h)- doesn't match the k(h)- of the fanqie, unless one supposes that the transcribed dialect had voiced (aspirate) velars corresponding to voiceless velars in the dialect spelled in fanqie: e.g.,

Hg-, dgh- = [Gg] (dialect 1) : [Gk] (dialect 2 in fanqie)

(Strangely, the fanqie G- + kh- corresponds to a Chinese transcription 夷 皆 without a kh-, whereas the fanqie Gk- + k- corresponds to the only Chinese transcription 夷客 with a kh- as well as a kh-less 夷隔. Did the author of the Pearl intend to write 夷格 with k- [which appears twice in the Pearl] instead of 夷客 with kh- [which only appears once]?)

I'll admit there might be some (accidental) correlation between the Tibetan transcription (pre)initials and the fanqie initials, but there is a far larger number of counterexamples (added

Homophones chapterTangut TelecodeGrinstead's glossRhymeTibetan transcription with Nevsky (1926) numberGong's reconstructionFanqie (Gong's reconstructed initials only)
I (labials)1257HOUSE1.14(d)mi, (d)me (#5)myiim- + w- (not d- + m-)
II (labiodentals)2000FRIEND1.67d-wi (#51)wyị̈w- + p- (not d- + w-)
III (dentals)1427AND1.57gnoH, no, ño (#16)nioow*n- + ch- (not g- + n-)
IV ('retroflexes'?)4672IMPURE1.41hñe (sic; Nishida 1964: 98)niəydz- (sic!) + sh- (not x-** + ñ-)
V (velars)1267STRONG1.79rkhi(H) (#9)kyirk- + 'y- (not r- + k-)
VI (alveolar fricatives and affricates)1479PURE1.33gseh, se (#48)seys- + l- (not g- + s-)
VII (palatals)0008TO-LEAD1.50gshoH, zha (#4)shiosh- + kh- (not g- + sh-)
VIII (velar fricatives and glottals)2074RECEIVE1.34bge (#60)Giwey (one syllable; I'd write this as Gwiey)G- + k- (not b- + G-)
IX (liquids, rhotics - and the palatal zh!)0100ONE1.43gli(H), kli (#1)lewl- + s- (not g/k- + l-)

Next: Over-gG-rowth?

*The resemblance between TT1427 AND nioow 1.57 and Old Chinese 而 *nə 'and' is probably coincidental.

Old Chinese 而 *nə 'and' may be from an unstressed *no. Although -graphs are normally not phonetic in *o-graphs, 而 *nə is phonetic in graphs for *no-type syllables: e.g., 儒 *no 'scholar'. 而 also meant 'whiskers on an animal' and is a drawing of hair. Perhaps 'whiskers' was *səno with a contracted variant 須 *sno 'beard'.

**Gong's reconstruction has an initial x- but no initial h-. CZECH OUT THESE CLUSTERS (PART 3)

These Czech speakers say that j- in initial j-clusters is not pronounced, though others claim they do pronounce them! I suspect the latter are influenced by spelling. I've heard of people claiming to pronounce words as they are spelled (when in fact they don't), but not the reverse - i.e., people claiming they don't pronounce words as they are spelled (when in fact they do). People want to say that they sound 'correct' and assume that spelling represents the 'correct' form. If someone had asked me twenty years ago if I pronounced the final letter in thumb, I would have said 'yes'. Taking a phonetics class opened my eyes - no, ears - to what my English really sounded like.

One may wonder if the Tangut were able to accurately write phonetic dictionaries without phonetics classes. Actually, the Tangraphic Sea's structure reveals that some Tangut were versed in the Chinese style of phonetic science. (But whether sinocentric phonetics was sufficient training or not is another issue.) The Tangraphic Sea is arranged by tones and rhymes like a Chinese rhyme dictionary and each group of homophones has a fanqie spelling.

If Tangut had initials like Gg- (Huang 1983) or Gk- (Nishida 1964), I would expect the tangraphs representing those initials to only appear in fanqie for syllables with those initials. If Gg- or Gk- could also appear in, say, the fanqie for g- or k-syllables, then either the fanqie are not very useful or the reconstruction is wrong.

According to Nishida (1964: 132-135), the initial Gk- only appears in syllables belonging to six rhymes: 1.9, 1.34, 1.44, 1.78, 2.53, and 2.59. Therefore the tangraph(s) representing Gk- should ideally never appear in fanqie for syllables in other rhymes. But in fact they do. Here are the four fanqie for the level tone tangraphs transcribed in the Pearl with 夷 'barbarian' followed by a velar-initial sinograph. (Fanqie for the three rising tone tangraphs are presumably in the lost rising tone volume of Tangraphic Sea.) Reconstructions are Nishida's, not Gong's.)

TT3409 GkIě 1.9 < TT5688 'yyuH* 1.3 + TT0560 unknown velar + - 1.9

TT2087 GkyE 1.34 < TT0875 (no Nishida reconstruction) 1.29 + TT1328 khe 1.34 (sic! -e is Nishida's value for 1.37!)

TT3861Gkew < TT3409 Gk 1.9 + TT5271 kew 1.44

TT5476 GkIěr 1.78 < TT5576 Gur 1.75 + TT3928 (no Nishida reconstruction) 1.78

The only tangraph of the four that has a Gk- initial speller is TT3861. The other three initial spellers belong to rhymes other than the six Gk-rhymes. TT3409 has a 'y-initial speller, TT5476 has a G-initial speller, and TT2087 has an initial speller reconstructed by Sofronov as 'yə and by Gong as Gyï. These initial spellers appear in fanqie for syllables in other non-Gk-rhymes:

TT3615 'uH 1.4 < TT0875 (no Nishida reconstruction) 1.29 +TT0918 khu 1.4

TT1739 unknown laryngeal + -U 1.28 < TT3409 Gk 1.9 + TT0135 unknown velar + -U 1.28

TT4425 (no Nishida reconstruction) 1.55 < TT3409 Gk 1.9 + TT4043 kyõ 2.48 (with the rising tone counterpart of 1.55! sic!)

TT4957 (no Nishida reconstruction) 1.85 < TT3409 Gk 1.9 + TT2565 unknown laryngeal + -ər 1.85

TT0508 Gar 1.80 < TT5576 Gur 1.75 + TT2799 (no Nishida reconstruction) 1.80

TT2486 (no Nishida reconstruction) 1.84 < TT5576 Gur 1.75 + TT0455 kUr 1.84

TT3413 r 1.86 < TT5576 Gur 1.75 + TT1499 (no Nishida reconstruction) 1.86

TT0769 GAH 1.89 (sic! -AH is Nishida's value for 1.17) < TT5576 Gur 1.75 + TT3367 vor 1.89

The fanqie for the initial spellers in Gk-fanqie have initial spellers in other non-Gk-rhymes:

TT5688 'yyuH 1.3 < TT1070 Gị̈ 1.69 + TT1548 (no Nishida reconstruction) 1.3

TT0875 (no Nishida reconstruction) 1.29 < TT5712 'yyuH 1.3 (homophone of TT5688 above) + TT2422 lïH 1.29

Two questions to ponder for next time:

- Given the above fanqie, how could Nishida defend his reconstruction of Gk-?

Hint: TT3928 was listed in Homophones chapter V (velars).

- Why do you think Huang reconstructed Gg- after so many rhymes (unlike Nishida, who reconstructed Gk- after only six rhymes)?

*'yy- represents the initial 'y- followed by the glide -y- (-ǐ- in Nishida's notation). CZECH OUT THESE CLUSTERS (PART 2)

Given that Czech orthographic j-clusters seem to reflect earlier front vowel-consonant sequences, I'm surprised that I could only find five of them in this online dictionary. Although I would not expect the first element of a cluster to occur before every other consonant in a language, I would expect it to occur before most consonants in one or more phonetic classes:

- Given English sp-, one would predict - and find - st- and sk- (i.e., s- before other voiceless stops)

- Given Czech jm-, one would predict - and not find - jn- (i.e., a j- before another nasal).

(Did *jn- > [ny]? - e.g., *ene > *jne- > ně [nye]?)

Why wouldn't front vowels reduce to j- before n- and other consonants (e.g., b-)? Maybe reducible (unstressed?) front vowels were randomly distributed. If vowel reduction only occurred in high-frequency words like 'I am' and 'goes' (but what about 'acicular'?), then j- would only precede whatever consonants those words had.

Tangut had an initial (cluster?) transcribed in Chinese as 夷 'barbarian' + velar-initial graph which might imply yk-. Nishida interpreted 夷 + K-graph as Gk-. I presume that the Gg- reconstructed (by Huang Zhenhua, according to Li (1986: 126); Wenhai yanjiu [1983: 128]) more or less corresponds to Nishida's Gk-.

Given Tangut Gg-, I would predict - and not find - in Huang's reconstruction:

G- before other voiced stops: Gd-, Gb-

or other voiced fricative-voiced stop combinations: e.g., zd-, Vb-

Gg- is the only cluster of its kind. Moreover, I don't know of a single language with Gg- or Vb-. (Surprisingly, ancient Greek zeta is said to have been [zd] [Joseph in Comrie 1987: 418].)

The distributional oddities don't stop there. In Huang's chart of possible initial-rhyme combinations in the level tone (1983: 128-133), Gg- can occur before different 39 rhymes, whereas Nishida's Gk- can only occur before six rhymes:

Rhyme numberHuang's reconstruction (given if his Gg- occurred before it)Nishida's reconstruction (given if his Gk- occurred before it; rising tone syllables are listed in the row for their level tone counterparts)
1.9-ï, -ui-Iě
1.15-ïæn, -ïon
1.34-əĩ, -eĩ-yE
1.60-ɐi(2.53 -yẸ)
1.66(2.59 -Iẹ̌)
1.69-ïer, -ïEr

(I may comment on the complexity of Tangut rhyme reconstructions in a future post.)

On the one hand, Gg- is a widely distributed initial in Huang's reconstruction even though it is hardly a widely distributed initial in languages in general. It is possible for languages to have exotic sounds, but I would expect them to develop in specific environments with appropriate phonetic characteristics. What would cause Gg- to develop before all sorts of vowels? And if Gg- were so common, why are there so few Chinese transcriptions (and seemingly no Chinese transcriptions) implying it?

(One might be tempted to change Gg- to a simple g- which is absent from Huang's reconstruction, but then why would g-syllables be placed in the laryngeal chapter of Homophones? Gg- is not really laryngeal either, but its first half does sound like Huang's glottal H.)

Conversely, why would Nishida's Gk- only occur before ye-type rhymes (with the exception of -ew which still has the vowel e)? There is nothing about -(y)e that would encourage a Gk- to develop.

Next: Do the fanqie fit? CZECH OUT THESE CLUSTERS

(Lame title, I know. I'm too tired to write a real post.)

Today I ran across some Czech word written with a j-cluster (was it jde 'goes'?) which reminded me of the mysterious 夷 'barbarians' in the Chinese transcriptions of Tangut.

Judging from this online dictionary, Czech has the following five orthographic j-clusters:

(no jb-)jd-(g only in foreign words; native g became h [H])
jm-(no jn-)(no phoneme ng)
(f only in foreign words)js-jh-
(no jv-)jr-

Comparison with Russian reveals that the j-s in these clusters came from earlier front vowels:

jde 'goes': cf. Rus идет

jho 'yoke': cf. Rus иго

jméno 'name': cf. Rus имя

jsem 'I am': cf. Rus есмь (archaic)

I could only find one jr-word (jrhlicovity 'acicular') and I couldn't find any Russian cognate for it.

I have a Hippocrene Czech dictionary (not the best, but what do you expect for a freebie?) and it treats the j- as silent in its transcriptions for English speakers (hence /ee/ = [i]):

jměni /'mnye-nyee/ 'fortune, wealth'

jméno /'mé-no/

jmenovat /'me-no-vat/ '[to] name, call, appoint, nominate'

jsem /'sem/

Those are the only four jC-words in the dictionary. Neither /'/ nor /ny/ have anything to do with j-. The symbol /'/ represents stress, not palatalization. The háček on ě signals that the preceding consonant is 'soft' (historically palatalized): m + ě = [mñe] (< *mye, presumably).

Does the absence of a transcription for initial j- truly reflect a zero sound value, or is that merely a convenience for Anglophones? Is that j- only pronounced after a vowel? For example, is a jsem 'and I am' [aysem] rather than [a sem]?

Although I don't think Tangut had yC-initial syllables, its relative Japhug rGyalrong does has them (e.g., yla 'yak-cow hybrid'), and I wonder what other languages also have them. THROAT-Y? (PART 1)

There is no consensus on how to reconstruct the y-like initial consonant of Tangut that I've been examining over the past few days:

Nevsky, Li Xinkui, Li Fanwen (1986), me: y-

Nishida, Gong: 'y-

Sofronov: ' + palatal (semi)vowel

Huang Zhenhua: G-? (Wenhai yanjiu, p. 128-134)

Arakawa Shintarou: Kotaka's chart and description only has 'y-, but the Cyberpearl has both 'y- and y- (and has 'e [no -y- at all!] for TT5283 GENITIVE/DATIVE 'yiy 1.36!)

(The above is largely based on Li [1986: 127] with modifications.)

However, I think everyone would agree that the mystery initial (or initials?) was classified as a

TT0651 GULLET nywiy 1.36

TT0099 SOUND Giẹ 1.59

in the eighth chapter of Homophones, though they might reject Grinstead's translation of the first tangraph! (I prefer LARYNX.*) TT0651 0099 is a calque of the traditional Chinese phonological term 喉音 'throat / larynx sounds'.

One might expect 'throat sounds' to mean laryngeals (i.e., pharyngeals and glottlals). However, in modern Chinese languages, some syllables which had 'throat sounds' in Middle Chinese now have palatal y- or labial w-. Similarly, some 'laryngeal sound' Tangut syllables were transcribed in Tibetan with y- or w-.

This does not necessarily mean that y- and w-syllables were placed in the laryngeal chapter by mistake. They could have had initial glottal stops ('y- and 'w-) which could not be easily transcribed in Tibetan**.

If Tangut-period northwestern Chinese (TNC) had a 'y-/y- and 'w-/w- distinction like the 13th century Hphags-pa Chinese koine in the east, in theory 'y- and 'w- sinographs should be favored whenever possible*** to transcribe Tangut 'y- and 'w- syllables in the Pearl. But it's not clear that TNC had such a distinction.

Suppose that TNC had y- and w- (phonetically [v] as in modern NW Chn dialects?) but not 'y- and 'w-. And suppose that Tangut had y-, w- (phonetically [v]?) and 'w-.

(Why y-? Unlike Nishida, Gong, and possibly Arakawa, I doubt that a language could have 'y- without y-. Is there any language with 'y- and without y-? Having y- as the sole y-like initial is far more likely, though I don't know of any language with the trio of y-, w-, 'w-.)

Tangut w [v]-syllables would naturally be assigned to Homophones chapter II (light lip sounds = labiodentals).

Tangut 'w [w]-syllables would be assigned to Homophones chapter VIII (laryngeal sounds) along with Tangut '-syllables.

Both of those syllable types are where they would be expected in Homophones. But why weren't Tangut y-syllables assigned to Homophones chapter VII (proper tooth sounds = palatals) along with Tangut ch-, chh-, j-, sh- (but not zh-!) syllables?

If they were 'y-syllables, then that would explain why they are in chapter VIII along with other glottal stop-initial syllables. However, if they had no glottal stops, there has to be another explanation.

Next: The long shot explanation.

*TT0651 appears in Pearl 0205:

TT5108 THROAT (PHARYNX?) kor 1.59

TT0651 GULLET (LARYNX?) nywiy 1.36

Nishida (1964: 205) translated this as 'throat'. Its Chinese gloss is 咽喉 'throat' (< 'pharynx' + 'larynx').

kor 1.59 (< *rko? *kro? *kor?) could be cognate to Old Chinese 喉 *go 'throat'.

(There is no Chinese-internal evidence for an r in the OC word corresponding to the retroflexion of the Tangut vowel, but a *rgo or *gro would also be theoretically possible. Although Matisoff [2003: 672] reconstructed a Proto-Tibeto-Burman 'or 'throat', there is nothing indicating that Tangut k comes from an earlier glottal stop [and I don't know of a single language in which '- > k.])

(1:36: The OC initial of 喉 *go 'throat'. could have been a cluster like *ng-k- or *m-k- that later fused into g-, but there is no Chinese-internal evidence for this.)

**Although the transcribers could have combined the letter 'a with subscript y and w to write 'y- and 'w-, they did not 'do so. Perhaps that was simply because such letter combinations are alien to Tibetan. Or because the Tangut initials really weren't 'y- and 'w-.

For comparison, the Tibetan-based Hphags-pa alphabet had a special letter for Chinese 'y- and also had combinations of '- (a letter based on Tibetan Ha, not Tibetan 'a!) with -y- and -w- for 'y- and 'w-. (I am using Coblin 2006's interpretation of Hphags-pa. See p. 64 for examples of 'y- and ' + subscript -y- both transcribing what Coblin reconstructed as 'y-.)

(Hphags-pa Chinese actually had two kinds of subscript y, but only one [Coblin's ÿ] is relevant to the above discussion, so I have transliterated it as y without Coblin's diaeresis.)

***There may not have been enough 'y- and 'w- type syllables in earlier Chinese to transcribe Tangut syllables containing a wide range of vowels after 'y- and 'w-. DATIVE DETOUR

"Genitive Guidance" got me on the wrong track. Two wrong tracks, in fact. Fortunately, Guillaume Jacques got me off both of them in time..

First, he pointed out that

TT5283 GENITIVE/DATIVE 'yiy 1.36

is to be compared with a rGyalrong locative suffix -y that is still in use in eastern rGyalrong but not in Japhug (though it survives in, for example, qhuy 'this evening'*). I knew that comparing to the Tibetan genitive** was a long shot, but this blog is meant for experiments like that. And some are bound to fail. Science is not all about success.

So much for my plan to revise my lenition theory a bit by proposing that final consonants were lost in Tangut before g in the genitive/dative lenited to zero:

Stage one: CV(C)-gyiy

with CV(C) being the final syllable of a noun which may have ended in a consonant

Stage two: CV-gyiy

The final consonant of CVC has been lost (unless it was -y or -w)

Stage three: CV-Gyiy

-g- lenited to -G- between V and -yiy. Some dialects may have stopped at this point - was the Tibetan transcription g-y- representing a -G-?

Stage four: CV-yiy

G lenited to zero. Reflected in the Tibetan transcriptions ye and yi.

Second, he pointed out that there are errors in his indexes of Gong's reconstructions from Li (1997) and that Li's dictionary also contains errors. As Li himself wrote (1997: 13),

I am not saying that every word is correct or every problem is solved in my dictionary, which is impossible.

In the introduction to his dictionary, Li (1997: 15) did not list y- as a possible initial in Tangut. I did notice that but wondered if he had simply left out y- by accident. I really wish I could see Gong's phonological sketch of Tangut in the 2003 Thurgood and LaPolla volume on Sino-Tibetan (a mere $340 at Amazon!). I really wish I had Li's dictionary, period. I would appreciate it if anyone can help me get a copy. (I've been quoting from photocopies of the introductory matter.) I assume that y- is an error for 'y-, but I am not sure. In any case, the fanqie indicate a single y-like initial (either 'y- or y- depending on reconstruction).

Next: Throat-y?

*I doubt that any Japhug speakers actually think of qhuy as qhu 'late' + LOCATIVE anymore. A once productive case ending has been reduced to a phonetic fossil.

**I was surprised that LaPolla and Huang (2003: 223) compared the Qiang and Tibetan genitives since I was under the impression that the Tibetan genitive was a Tibetan innovation. LaPolla himself wrote,

... although all Tibeto-Burman languages have developed some sort of relation marking, none of the markers can be reconstructed to the oldest stage of the family [i.e., Proto-Tibeto-Burman]

On p. 3 of that paper, he reconstructed Tibetan-Western Himalayan *gi for the genitive (which still looks like the Tibetan transcription g-ye for TT5283 GENITIVE/DATIVE 'yiy 1.36, though we can't be sure that g-y- actually stood for a velar initial).

LaPolla also wrote (p. 3),

It seems that in Proto-Tibeto-Burman, as in many modern Tibeto-Burman languages, a genitive relationship was marked by no more than position immediately before the modified noun.

This possessor noun-possessed noun construction was still present in Tangut: e.g. (Nishida 1966: 568),

TT1143 DHARMA tsyiir 1.93

TT2267 DRUM bar 1.80

= 'drum [of the] Dharma'

from Pearl 212.

One might expect modifier-modified order to extend to adjectives and nouns, but this is not the case: e.g.,

TT2267 DRUM bar 1.80

TT5660 BIG khwey 2.30

= 'big drum'

(Cf. languages like Korean and Japanese, which are 'object-verb' and have possessor-possessed order like Tibeto-Burman but have adjectives before rather than after nouns. I used to think of the Kor/Jpn type order as 'normal' until I read Dryer [], who stated that noun-adjective order is "more common" in non-Asian OV languages:

The OV TB languages are in many respects atypical among OV languages in Asia, but normal for OV languages in the world as a whole, in that in most OV TB languages, some of these modifiers normally follow the modified word.)

LaPolla (p. 10) thought head-attribute (e.g., noun-adjective) order went all the way back to Proto-Tibeto-Burman:

We can therefore assume MODIFIER-MODIFIED order in N-N structures, and GENITIVE-HEAD, HEAD-ATTRIBUTE, NEGATIVE-VERB, and RELATIVE-NOUNword order patterns for PST [Proto-Sino-Tibetan].

Tangut also has negative-verb and relative-noun order: e.g. (Nishida 1966: 576, 582),

TT1491 NOT myi 1.11

TT2413 DO wyi 1.10

'not do'

(cf. Japanese which has shi-nai 'do-not', but Korean has both NEG + V and V + NEG constructions: 안 한다 an handa 'not' + 'do' and 하지 않다 haji anh-ta 'do' + 'not be'; 않 anh- 'not be' is from안 an 'not' + 하- ha- 'do')

(Why do rhymes 1.10 and 1.11 [and their rising tone counterparts 2.9 and 2.10] all end in -yi? Surely different rhymes should not be homophonous! I'll explain this some other time.)

TT4876 ALL ryur 1.76

TT3464 BUDDHA tha 1.17

TT5283 GENITIVE/DATIVE 'yiy 1.36

TT4351 ENTER 'o 2.42

TT3346 MUST lew (definition from Nishida 1966: 437)

TT5660 BIG lyịy 2.54 (preceding the noun!***)

TT4544 SAD wyuu 1.7

'the great compassion which all the Buddhas must enter' (avataMsakasuutra 41; tr. from Nishida 1966: 582)

***Nishida (1966: 274) treated TT5660 BIG lyịy 2.54 preceding a noun as part of a disyllabic noun rather than an adjective preceding a noun. Cf. black in blackbird (part of a noun; a specific kind of bird and not just any black bird) and black bird (adjective plus generic noun).

BIG SAD looks like a calque of Chinese 大慈 'great compassion' and Nishida's other examples

'great yin' = 'moon' (Pearl 051)

'great yang' = 'sun' (Pearl 051)

'great earth' (Pearl 115)

also look like probable calques of Chinese 太/大X 'great ...' words:

太 陰, 太陽, 大地 (with the same meanings; these are also the Chinese glosses in the Pearl)

Copyright 2002-2007 Amritavision