I was surprised to see the pronunciation of Zhuang s listed as [θ] in the Japanese Wikipedia article on Zhuang because neither of the two descriptions of Zhuang I had seen up until then - Ramsey's and Mori's - mentioned it. But two varieties of Wuming Zhuang in 中嶋幹起 Nakajima Motoki's "A Report on the Basic Vocabularies of the Tai Dialectal Variation of Wuming" (1980; PDF) did (and still do?) have such a sound. That article lists 276 words in three varieties of Tai:

- one variety of Wuming Zhuang recorded by Nakajima

- another variety recorded by Li Fang-kuei (1956)

- a different northern Tai language in Longzhou recorded by Li (1940)

The title refers to the s : θ : ɬ correspondence in 'three' and 'four' below:

Gloss Zhuang (Mori 2004) Wuming (Nakajima 1980) Wuming (Li 1956) Longzhou (Li 1940) Siamese Li's Proto-Tai (1977) via Proto-Tai'o'Matic Cantonese Middle Chinese Late Old Chinese
one ɗeew; it (lit.) it ʔit ʔit nɯŋ; ʔet (in 'eleven'); also cf. diaw 'alone' *hnɯŋ; *ʔjet; *ʔdiaw 'alone' jɐt *ʔit ʔit
two sooŋ; ŋej (lit.) ŋoi ŋoi jii sɔɔŋ; jii (in 'twelve') *sɔŋ; *ɲwi ji; sœœŋ 'pair' *ɲih; *ʂɔŋ *ni(t)s; *sroŋ
three saam θɑɑm θam ɬaam saam *sam saam *sam *shləm
four sej θooi θoi ɬii sii *si sej *sih *shlits
five ha; ŋu (lit.) hɑɑ ha haa haa *ha ŋ < *ŋu *ŋoʔ *ŋaʔ
six ʁok; lok (lit.) ɣɔk luk luk hok *xrok luk *luk *ruk
seven ɕat ɕat šät šit tɕet *čet tshɐt *tshit *shnit
eight peet pɛt pat peet pɛɛt *pɛt paat *pɛt *pret
nine kow kɑu kău kau kaw *kjəw kɐw *kuʔ *ku
ten ɕip ɕip šip ɬip sip *sip sɐp *dʑip *gip


General: Nearly all the Tai numerals are borrowed from Chinese.

'One': PT *hnɯŋ 'one' and PT*ʔdiaw 'alone' are native.

The latter superficially resembles Tangut

lew 1.43 < *CV-tek 1.43 'one'

tjịj 1.61 < *C-tje 'alone'

cf. Matisoff's (2003: 660) Proto-Tibeto-Burman *t(j)ak, *t(j)ik 'one'

but the similarity is fortuituous.

PT *ʔjet looks as if it was borrowed from a transitional stage between MC *ʔit and Cantonese jɐt, but Chinese borrowings in Vietnamese show no sign of an *e-stage.

'Two': The *-w- in PT *ɲwi corresponds to nothing in the Chinese reconstructions. Perhaps PT borrowed the word from an early southern Chinese dialect with a medial glide from an earlier labial prefix:

?*P-ni(t)s > *ɲwih

'Three' and 'four': Although Longzhou's lateral fricative ɬ- seems to preserve OC *-hl-, it comes from an *s in Li's PT reconstruction. Note also that 'ten' has ɬ- in Longzhou but never had *hl in OC.

In 1962, Pulleyblank reconstructed OC which cthorresponds to *hl in other more recent reconstructions.

'Four': Wuming has a labial vowel implying that the PT form could have been *swi. My explanation for *-w- in 'two' may also apply to 'four':

?*P-shli(t)s > *swih

Perhaps Wuming simply borrowed from a *swih-dialect of Chinese, whereas the rest of Tai borrowed from *sih dialects. Does any modern Chinese language have a labial glide or vowel in 'four'?

Written Tibetan bzhi < *b-lji 'four' also has a labial prefix, though this could be a coincidence.

'Five': Most of the Tai forms are very strange because they combine an h- implying an MC or LOC *x- < *hŋ- combined with an -a that preserves the original OC vowel. Presumably Tai reflects an early southern Chinese *hŋaʔ, whose voiceless initial arose from the fusion of a voiceless prefix with root-initial *ŋ-.

Mori's literary Zhuang ŋu could either be from Cantonese syllabic ŋ with an epenthetic vowel or from early Cantonese *ŋu.

'Six': PT *xrok has an unexpected *x- and nonhigh vowel. It seems to reflect an early southern Chinese *hrouk whose OC ancestor *Cʌ-ruk had an emphatic prefix absent from mainstream Chinese.

Wuming (Nakajima) ɣ- is from *r- < *hr- < *xr-.

The northern Tai l-forms are later borrowings.

Mori's literary Zhuang lok seems to combine the newer Chinese initial l- with the -ok of ʁok. Maybe there was an early southern Chinese *lo(w)k < *hrouk.

(Mori used the official Zhuang romanization. I write r as ʁ since he compared it with French r. But perhaps he was trying to describe [ɣ].)

'Seven': I don't know why Siamese and PT do not have an aspirated affricate initial.

The comments on the vowel of PT*ʔjet 'one' also more or less apply to 'seven'.

'Eight': Wuming (Li) pat seems to be a later borrowing than the others whose vowels reflect MC or late OC *ɛ.

'Nine': The early MC form may have been *k(ɨ)əw in the south. Cf. Viet cậu ̣[kəw] 'mother's brother' < Chn 舅 which rhymed with 九 'nine' in MC.

'Ten': The Tai forms have a voiceless initial which seems to indicate that the word was borrowed after devoicing in Chinese. Another possibility is that an early southern Chinese dialect had a voiceless initial from the fusion of an earlier voiceless prefix-*dʑ- cluster. 中嶋幹起 A TRUNK RISING FROM THE CENTRAL ISLAND

It's a small world. Last night, I found a Zhuang word list collected by 中嶋幹起 Nakajima Motoki, coauthor of these books on Tangut. All roads lead back to the 'state of ten thousand secrets'!

Just now, I discovered a partial but still very long list of Nakajima's publications that doesn't include that word list. It turns out that he's a Sinologist who's worked on Qiang, Manchu, Mongolian, and Khitan* in addition to a wide variety of Chinese languages. The list only goes up to 2002. I wonder what he's done since then. As far as I can tell, his Tangut site hasn't been updated since 2003. I hope it never goes down, as it has been an invaluable reference for the past two years.

*A good choice for a Tangutologist. I'm surprised that I can't find anything he's written on Jurchen, which would be the perfect intersection of his interests in Manchu and Khitan. (The Jurchen language is related to Manchu, but its script is related to khitanography.) ƧЗЧƼƄ

may be my least intelligible post title so far. It's certainly the least pronounceable, since it has no consonants or vowels. Those five letters represent standard Zhuang tones 2 through 6 in the now defunct PRC romanization of Zhuang. Can you guess why?

The choice of ч is particularly neat since it could stand for Russian четыре 'four'.

ƅ can be awfully hard to distinguish from b, depending on the font, and both can occur at the ends of syllables.

In modern Zhuang romanization, ƨзчƽƅ have been replaced by zjxqh. (Q for the fifth tone reminds me of Latin quinque.) Letters for segments were also replaced:

ƃ [ɓ] > mb

resembles Cyrillic Б but is not [b] and has a distinct Unicode codepoint

ƌ [ɗ] > nd

ə [a] > ae

short counterpart of a [aa]; phonetic descriptions and data are mostly from Mori's Zhuang site and Wikipedia

ɯ [ɯ] > w

Most letters did not change. Some still have unexpected phonetic values:

c [ɕ]

r [ɣ]

s [θ]

according to the Japanese Wikipedia article, which implies standard Zhuang has no [s] - does any language have [θ] but not [s]?

2.21.3:52: Some Zhuang varieties do have [θ] but not [s]. I'll discuss them tomorrow.

The letter e following another vowel signals 'nondefault length':

long default length short default length except before glides u [w] and i [j]
Default length [aa] [oo] [i] [u] [ɯ]
Spelling a o i u w
Nondefault length [a] [o] [ii] [uu] [ɯɯ]
Spelling ae oe ie (but i before u) ue (but u before i) we (but w before i)

The length of e is automatically determined by the following consonant, so there is no need for a digraph ee:

Short [e] before Long [e] before
[j] [k ŋ t n p m w]

No syllable can end in a short vowel, yet Zhuang has syllables ending in -ae. How is this possible? Comparisons with Siamese point toward the answer:

'chicken': Zh gaeq : S ไก่ kaj

'can': Zh ndaej : S ได้ daj

note that Zh -j is a tone letter, not [j]

'go': Zh bae : S ไป paj

Syllable-final ae is actually [aj], not [a]. -ae is presumably an abbreviation of -aei. -i represents final [j] elsewhere: e.g.,

-ai [aaj], -oi [ooj], -ui [uuj], -wj [ɯɯj], -ei [ej]

But ae before a consonant other than [j] is [a], so ae in that position corresponds to Siamese [a]:

'evening': Zh haemh [ham]: S ค่ำ kham

'we': Zh raeuz [raw]: S เรา raw

'wine': Zh laeuj [law]: S เหล้า law

These correspondences also apply to loanwords from Cantonese:

'unit of weight': Zh gaen [kan]: Ct 斤 kan

'unit of capacity': Zh daeuj [taw]: Ct 斗 taw

'north': Zh baek [pak]: Ct 北 pak

'west': Zh sae [saj]: Ct 西 sae

The numbers underlying the older Zhuang tone letters match the numbers that I use for Middle Chinese tone categories. The table below lists these numbers, their letters in nonitalic old and italic new Zhuang romanization, and a numerical description of each tone.

Tone category *Voiceless initial *Voiced initial
V 1/zero/zero [24] 2/ƨ/z [31]
Q 3/з/j [55] 4/ч/x [42]
H 5/ƽ/q [35] 6/ƅ/h [33]
C long vowels: 3/p, t, k/p, t, k [55]
short vowels: 5/p, t, k/p, t, k [35]
6/b, d, g/b, d, g [33]

All syllables in the C category end in voiceless stops: [p t k]. However, these stops are written as b, d, g if they are preceded by mid level tones. Vowel length and graphic coda 'voicing' are sufficient to indicate the tones of C category syllables; there is no need to add tone letters 5 and 6.

Maybe I should rename the C category 'K' to avoid confusion with the C category of Tai linguistics (which corresponds to my Q category!). K represents the coda -k, and it is a back consonant like Q and H.

The three-way split of the K category in Zhuang is reminiscent of similar three-way splits in Siamese and Cantonese:

Category K syllables
Earlier initial *Voiceless (K1) *Voiced (K2)
Vowel length Short Long Short Long
Zhuang [55] [35] [33]
Siamese Low High Falling
Cantonese High Mid Low

Notice that Zhuang's tone system is more like its nonrelative and neighbor Cantonese than its relative and nonneighbor Thai:

- Zhuang and Cantonese split the voiceless K1 category, whereas Siamese split the voiced K2 category

- Zhuang and Cantonese are voiced-low languages in which *voiced initials are associated with lower tones, whereas Siamese is voiced-high

Zhuang differs from the other two in one respect: its long-vowel tone rises, whereas Cantonese and Thai have a long-vowel tone that has a lower endpoint than its short-vowel counterpart. 化 CHANGING MY MIND ABOUT 亻 PEOPLE

This morning I realized that I overlooked an obvious sinographic source of the tangraphic phonetic element 亻 khwa/xwa (examples here): the left side of 化 'change', *xiwa in Gong's reconstruction of 12th century northwestern Chinese.

That hadn't occurred to me because I was looking for exact homophones (ignoring tones). Hence I proposed 科 12NWC ?*khwa 'class' and 和 12NWC ?*xwa 'harmony'. I almost mentioned 禾 ?*xwa 'rice plant' last night, but I left it out since it's less common than the other two sinographs and therefore would be a less likely source. I was thinking that 亻 could be 禾 without a horizontal line and the diagonal strokes at the bottom.

Deriving 亻 from 化 would be far simpler from a graphic point of view, but the rhymes would have to be reconciled. -wa is a Grade I rhyme in both Gong's 12NWC and Tangut reconstructions, whereas *-iwa is a Grade II rhyme. Would the Tangut use part of a sinograph with a Grade II reading to create tangraphs with Grade I readings? Maybe.

In any case, I doubt 化 ended in *-iwa, which is distinct from *-jwa in Gong's reconstruction of 12NWC. I have never seen such a distinction in any pair of monosyllables. Perhaps *xiwa could be reinterpreted as *xwia which would sound almost like Vietnamese khuya [xwiə] 'late'. But I doubt that Grade II in either 12NWC or Tangut had a medial *-i-, for reasons that I'll explain in a future installment of "A *kra-zy Idea". 徐冰 SLOW ICE AND THE 谷文達 VALLEY OF LITERARY ACHIEVEMENT

Years ago, I was blown away by 徐冰 Xu Bing's 天書 Book from the Sky (lit. 'heaven book'):

A Book from The Sky is comprised of printed volumes and scrolls containing four thousand "false" Chinese characters invented by the artist and then painstakingly hand-cut onto wooden printing blocks.

His quasisinographs (quasinographs?) reminded me of parasinography: khitanography, jurchenography, and tangraphy.

Today, I discovered another quasinographic artist, 谷文達 Gu Wenda*, creator of the Temple of Heaven (no relation to the 'Heaven Book'):

an entirely human hair made temple of pseudo-chinese, english, hindi, arabic and synthesized english-chinese, chinese ming dynasty’s furniture tv monitors, a video film heaven.

Gu's 碑林 Forest of Stone Steles contains Tang Dynasty poetry into English written in sinography: e.g.,

月光 Late Middle Chinese *ŋwyet kwaŋ 'moonlight' > moonlight > 虻癩忒 Md mang lai te 'horsefly leprosy error'

Gu tried to rework the literal translations of the transcription sinographs into a new English poem. He explains the process here:

the phonetic translation back to chinese from the english version of tang poems is a very complicated and tiring process. the following are four steps describing the sound translation process:

#1, one word at a time i search for a chinese word which sounds as exact as possible to the english version of the tang poem; i call this chinese sounds mimic english. overall, there are a multitude of chinese characters with the same or similar pronunciations that could work.

#2, from the many choices, i try to select the one chinese word which allows itself to be a building block to the new story being created.

#3, i repeat the same process for the next word trying to make sense within the developing context, and so on ... for each word.

#4, in order to make all of the english sound mimic chinese work together as phrases and sentences in creating the post tang poem, i often have to revise previous words until all the english sound mimic chinese have successfully constructed a readable new post tang poem.

Here's a simpler series of works (signed "谷氏楷典 'Gu's Square-Style** [Calligraphic] Classics") using sinography as-is: e.g.,

Sotheby's > 素思碧寺 Md suo si bi si 'simple thoughts green temple'

(see the neon sign here - it's red, not green!)

*This is the standard Mandarin version of his name which in IPA is [ku wən ta]. His name in his native language (Shanghainese) is presumably something like koʔ vəŋ daʔ. I got those readings from Glossika's online dictionary.

In the Wikipedia article on Shanghainese, the rhyme that Glossika romanized as -əŋ appears as -əɲ with a palatal nasal coda. If that notation is not an error for -ŋ, Shanghainese would be the only modern Chinese language with that I know of. (It is debatable whether Middle Chinese dialects had palatal nasal and stop codas that are preserved in Sino-Vietnamese -nh [ɲ] and -ch [c].)

**The actual graph for 'square style' that Gu uses has the components of 楷 stacked top to bottom rather than left to right. This seems to be a graph that Gu invented. (The Taiwanese government's allographic dictionary does not include it among its six variants of 楷.)

Gu has been experimenting with his own variations of sinography. I would treat these graphs as idiosyncratic sinographs and reserve the term quasinographs for characters that superficially resemble sinographs but lack sinographic structure (i.e., do not use the existing inventory of semantic and phonetic elements***) and do not represent sounds in any real language****. Parasinography, on the other hand, does represent sounds in real languages (i.e., Khitan, Jurchen, and Tangut).

Sinography Quasinography Parasinography
Contains sinographic semantic and phonetic elements + - -
Represents sounds in real languages + - +

Quasinography is not writing because it only mimics the form of written language and lacks semantophonetic substance. It is aesthetic, but it is not literary.

***Quasinographs can share graphic elements with real sinography, but those graphic elements have no meanings or readings.

Parasinographs can also share graphic elements with sinography proper, though those elements may have different functions: e.g., 亻 is a left-hand semantic element for 'person' in sinography but may be a phonetic for khwa/xwa in tangraphy (that may have originated as an abbrevation of 科 ?*khwa and/or 和 ?*xwa in the NW Chn dialect known to the Tangut). 亻 appears on the left of only four tangraphs (glosses based on Nishida 1966: 359):


the only one of the four whose reading does not fit the khwa/xwa archetype

the right side is shared with TT0590 njɨɨ 1.32 TWO

TT3102 khwa 2.14 CHINESE-CLOTH

CHINESE is on the right

probably a variant spelling of the following word

TT3103 khwa 2.14 CLOTH

ONE-OF-TWO is in the center, followed by the mysterious right-hand element ヒ - why?

TT3104 xwa 1.17 MONK

has HAND on the right - why?

borrowed from NW Chn 和尚 ?*xwa ɕo 'monk'

TT3101 ɕiar 1.81 shares an initial with the second graph of 'monk', 尚 *ɕo, but the vowels don't match, so I doubt that its 亻 stands for 'monk'.

although 尚 *ɕo was once 尚 *ɕaŋ, this a-reading was probably obsolete by the time tangraphy was invented, so it could not have influenced the design of TT3101

****I wrote "any real language" rather than "any Sinitic language" because I view vietography (喃 nôm), koreography (國字/국자 kukca 'national characters'), and japanography (國字 kokuji 'national characters') as branches of sinography rather than as parasinography. I suppose I could name all three sinoxenography by analogy with the term Sinoxenic for localized Chinese readings of sinographs by non-Chinese speakers. I would also include Zhuang sawndip (zhuangraphy) as a kind of sinoxenography.

Here's a sample of sawndip. It contains a graph


which coincidentally looks like the Vietnamese graph for 'heaven' from "Vietographic Virtues", though it is glossed in Chinese as 上 'top'.

More samples are listed here under "文献样品" (document samples). THE VIETOGRAPHIC VIRTUES OF VISTA

This is the first post I've typed on my new Windows Vista laptop using KompoZer, the successor to Nvu, the program I've been using for the past three years.

On Sunday, I discovered that Vista has built-in support for Unicode CJK Unified Ideographs Extensions A and B. I'd rather call it 'CJKV' since many of the characters are vietographs for native Vietnamese words such as

𡗶 trời/giời 'heaven' = 天 thiên 'heaven' + 上 thượng 'top' (both semantic)

𣋀 sao 'star' = 牢 lao 'prison' (phonetic) + 星 tinh 'star' (semantic)

the vietograph was devised when 'star' was earlier Vietnamese *Craaw; I can't determine the original initial consonant without more evidence: e.g., other Vietic words for 'star' and/or another vietograph containing a phonetic element representing the lost initial

The second graph appears in my 1997 edition of Ngũ thiên tự with the elements reversed (星+牢) . The same is true of

𡨸 chữ 'character' = 宁 trữ 'space between gate and gate screen' (phonetic*) + 字 tự 'character' (semantic)

字+宁 in Ngũ thiên tự

which is a vietograph devised for an early Chinese loan that is not considered to be Sino-Vietnamese. Its semantic element represents a borrowing of that same Chinese word from a later dialect**.

There are some true sinographs in the CJKV Extensions as well: e.g.,

𡦼 = 宀 'roof' + 人 'person'

which I mentioned in part 1 of "A *kra-zy Idea" as a variant of 家 'house' (with 豕 'pig' as phonetic). (I've got to write part 3 soon.)

Vista also supports the Hong Kong Supplementary Character Set (香港增補字符集) which includes Cantonese-specific cantographs such as

𣲷 Cnt nap in 濕𣲷𣲷 sap nap nap 'very wet'

the root is pan-Chinese; the rest is not

𣲷 combines 氵 'water' (semantic) with the obscure phonetic 㘝 'take and collect things' (Cnt nap?), a drawing of a 又 hand in a 囗 box

For other examples of this kind of reduplicated intensive, see Bauer and Benedict (1997: 299-300).

*宁 'space between gate and gate screen' is an example of a graph whose main function is phonetic: e.g., in the common graph 貯 'to store' (Viet trữ, Md zhu). If 宁 were not obscure, it might not have been used as a simplified form of 寧 representing an unrelated morpheme 'peace' (Viet ninh, Md ning). In simplified Chinese, 宁 has a double phonetic function:

- as a zhu-phonetic in 貯佇竚紵羜

- as a ning-phonetic in 拧柠咛泞狞聍苧

simplified from 擰檸嚀濘獰聹薴

Nothing in the graphs themselves indicates which reading of 宁 is applicable. Such polyphonic phonetics also exist in tangraphy.

I can't display any tangraphs because I haven't installed Mojikyo or copied my tangraphic GIFs onto this new machine yet.

Early Vietnamese had no alveolar fricatives like *dz, so early Middle Chinese *dzɨh was borrowed as early Viet *dʑɨh (> modern chữ) with a palatal initial.

I assume this borrowing occurred before Vietnamese developed tones; at this point EMC (and early Viet?) *-h was probably breathy voice rather than an actual final [h].

Later Vietnamese seems to have developed a *z or *dz, so late Middle Chinese *dzɨ6 was borrowed as c. 10th c. Viet *(d)zɨ4 (> modern tự).

The Vietnamese nặng tone (which I write as 4 since it is the voiced-initial Q-category) must have been phonetically similar to tone 6 (yang departing) in the LMC dialect that was the source of Sino-Vietnamese. RISING LEVELS: THREE DOZEN SINO-TANGUT TONAL CORRESPONDENCES

11 years ago, I couldn't find any strong correlations between Tangut tones and the tones of the sinographs used to transcribe Tangut in the Pearl. I don't have that data on hand, but I do have a small sample from part 2 of "An Uncertain Rhyme".

Tangut level-tone tangraphs in this sample tended to be transcribed with level-tone sinographs:

Tangut tone Tone class of transcriptive sinograph Number of occurrences
level yin level: 1 1
yang level: 2 7
yin rising: 3 1
yang rising: 4' (all with sonorant initials) 1
yin departing: 5 0
yang departing: 6 1
rising yin level: 1 4
yang level: 2 0
yin rising: 3 0
yang rising: 4 (all with sonorant initials) 2
yin departing: 5 2
yang departing:6 0

The reverse is also true: level-tone sinographs in this sample tended to be transcribed by level-tone tangraphs:

Chinese tone class Number of level-tone transcriptive tangraphs Number of rising-tone transcriptive tangraphs
yin level: 1 3 2
yang level: 2 4 0
yin rising: 3 0 1
yang rising: 4' (all with sonorant initials) 0 1
yin departing: 5 0 3
yang departing: 6 1 2

These tables could be simplified by combining the Chinese tonal categories to match Tangut:

Tangut tone Number of level-tone transcriptive sinographs Number of non-level ('oblique') tone transcriptive sinographs
level 8 3
rising 4 4

Although the Tangut rising tone had a 50/50 chance of being transcribed with a level or oblique tone sinograph, the reverse is not true: oblique-tone sinographs were mostly transcribed with rising-tone tangraphs.

Chinese tone class Number of level-tone transcriptive tangraphs Number of rising-tone transcriptive tangraphs
level 7 2
non-level ('oblique') 1 7

The size of the sample (19 transcriptive sinographs and 17 transcriptive tangraphs) is too small to be representative. Too bad, because the general patterns fits what I would expect to see if my hypothesis about the glottal origin of the Tangut rising tone is correct:

OC Late OC Middle Chinese tone category Expected direction Tangut tone Pre-Tangut
*-V, *-G, *-N (final vowels, glides, nasals: i.e., sonorants) level <-transcribes> level *-V, *-G, *-N (final vowels, glides, nasals: i.e., sonorants)
*-ʔ rising <-transcribes-> rising *-ʔ
*-(C)s *-h departing *-h < *-(C)s

Many of the tonal systems in the Sinosphere can be described using what I call the VQHC scheme:

V: final vowels and other sonorants

Q: final glottal stop

H: final fricatives

the above three are 'live' in Tai terminology

C: final stops

'dead' in Tai terminology

this class has been split between level and rising in Tangut; the details of the split have yet to be determined (and the mystery of the handful of possibly stop-final 'entering' tone tangraphs in Precious Rhymes of the Tangraphic Sea has yet to be solved)

This four-way categorization corresponds to ACBD using Tai terminology. (The Q and H categories are reversed in Tai linguistics. D simply means 'fourth' but it conveniently is the initial letter of 'dead'.)

Even the Austronesian Tsat language of Hainan more or less conforms to the VQHC scheme if I am reading what remains of these charts correctly:

*voiceless initial *voiced initial
V category 33 11
C > Q category (all final voiceless stops merged to glottal stop) 24 42
H category 55

The Tsat V category tones are what J. Marvin Brown would call 'V-low' (V-L) since *voiced-initial syllables have lower tones (11) than *voiceless-initial syllables (33).

The Tsat C > Q category, on the other hand, is 'V-high' (V-H) because *voiced-initial syllables start with high tones that drop (42) whereas *voiceless-initial syllables start with low tones that rise (24).

I wonder if two earlier tones merged to form the Tsat H category, and if those tones were V-low or V-high.

Brown (1985: 21) seems to think that tone splits conditioned by initial voicing are initially 100% V-L or V-H but "eventually wear down to 50 percent". Perhaps Tsat has reached that 50% point.

Tangut, however, died before it ever had a chance to develop a phonemic tone split. It may have had a phonetic tone split in one direction or the other, but such allotones would not become distinctive until the voicing that originally conditioned them was lost: e.g., if the level tone were 33 after voiceless initials and 11 after voiced initials:

before devoicing of initial obstruents after devoicing of initial obstruents
different initial phonemes, same phonemic tone same initial phoneme, different phonemic tones
phonemic notation phonetic notation phonemic notation phonetic notation
/khalevel/ [ka33] /kha33/ [kha33]
/galevel/ [ga11] /kha11/ [kha11]

No evidence suggests that Tangut devoiced initial obstruents before its death, whereas the Chinese dialect known to the Tangut had shifted earlier voiced obstruents to voiceless aspirates: e.g., *g > *kh. The modern northwestern Chinese dialects only show this shift in level tone syllables, as in eastern Mandarin dialects. This is yet another reason that I suspect that the current NW dialects are not direct descendants of the dialect known to the Tangut, though traces of that earlier dialect may still remain.

*The exceptions were fricatives: e.g., *z > *s (not *sh: i.e., aspirated s as in Burmese). R43: AN UNCERTAIN RHYME (PART 2)

In "十二世紀末漢语西北方音韻母系统的構擬" (1995), Gong reconstructed the rhyme of 定 'certainly' as *-jij in late 12th century northwestern Chinese. This is reminiscent of the -jij that he reconstructs for Tangut rhymes 36 and 37. I am suspicious of all these -jij because

- I know of no language with -jij.

if a modern Chinese language had [jij], it might be transcribed by PRC linguists as iii, since they tend to write [j] as i.

- Gong's Tangut reconstruction distinguishes between -jii (R14) and -jij (R36/R37). This seems to be a very fine distinction. If I attempt to make the two sound different, the results are [jiː] and [jiɪ].

- R36/R37 -jij was generally transcribed in Tibetan with a mid vowel -e, not a high vowel -i. (There are exceptional transcriptions ending in -i, -ï and even -oH!)

- R36/R37 -jij was generally used to transcribe Sanskrit e, not Sanskrit i (though this is not strong evidence if the Tangut transcribed Chinese transcriptions of Sanskrit which had raised *e to i)

I have reinterpreted Gong's R36/R37 -jij as -je, and I suspect that his 12NWC *-jij should also be *-je, a rhyme that is not in his reconstruction.

12NWC *-jij sinographs and R36/37 -jij tangraphs transcribed each other in the Pearl, indicating that their rhymes were similar (i.e., they were both *-je?). Numbers after sinographic readings indicate tone categories*. None of the following tables are comprehensive.

Pearl Graph Transcribed as
20.2.4 TT1175 ɕjij R36 1.35 ɕjij2
(no examples of 12NWC -ji transcribed with R36 1.35 seem to exist in the Pearl, but in the Tangut translation of Leilin, TT0690 tɕhjij R36 1.35 transcribes 程 tɕhjij2,tɕhjij3, and 鄭 tɕhjij6)
7.2.10 TT1092 tɕjij R36 2.32 tɕjij1
16.5.2 TT5366 tɕjij R36 2.32
20.4.9 TT1974 wjij R36 2.32 wjij4'
17.6.3 tɕjij5 TT5366 tɕjij R36 2.32
24.5.9 TT3005 mjij R37 1.36 mjij2
26.4.6 TT3510 mjij R37 1.36
31.4.11 TT4276 tshjij R37 1.36 tshjij1
18.3.3 TT4729 tjij R37 1.36 tjij3
10.6.6 TT5816 djij R37 1.36 djij2
10.4.10 mjij2 TT0090 mjij R37 1.36
27.3.12 mjij2
23.5.9 phjij2 TT3533 phjij R37 1.36
4.6.9 sjij1 TT4436 sjij R37 1.36
20.6.8 TT3606 tsjij R37 2.33 tsjij1
16.4.7 TT4461 tjij R37 2.33 tjij1
34.6.4 thjij6 TT2897 thjij R37 2.33
14.6.2 tsjij1 TT3606 tsjij R37 2.33
13.1.8 tsjij3

12NWC *-jij (my *-je) was also transcribed with other Tangut rhymes which were the long and tense counterparts of R36/R37. I would reconstruct these rhymes as *-jee and *-jẹ:

Pearl Graph Transcribed as
18.6.4 mjij6 TT3724 mjiij R40 2.35
19.6.6 ɕjij2 TT0779 ɕjịj R64 1.61
12.2.2 ljij4' TT1381 ljịj R64 2.54

R79 -jirj, the retroflex version of R36/R37, was not used to transcribe any sinographs in the Pearl or the Tangut translation of Leilin since 12NWC had no retroflex vowels. Nishida (1964: 62) and Sofronov (1968 II: 46) list its Tibetan transcription as -e. (Nishida also lists -u [!?], but I cannot confirm this in Sofronov.) Hence I reconstruct R79 as -jer.

R43 must have been similar to R36/R37 since it too was used to transcribe 12NWC -jij (my *-je). Transcription pairs in bold indicate tangraphs for Chinese loanwords used to transcribe the sinographs for those loanwords and vice versa.

Pearl Graph Transcribed as
19.6.7 TT0177 ɕjɨj R43 1.42 ɕjij2
27.6.3 TT3922 ljɨj R43 1.42 ljij6
30.5.9, 36.6.9 ljij6 TT2339 ljɨj R43 1.42
9.6.2 tjij1 TT4159 tjɨj R43 1.42
11.2.10 tɕjij1 TT5538 tɕjɨj R43 1.42
27.6.8, 28.4.11 tɕjij5 TT1146 tɕjɨj R43 2.37
28.4.11 TT1146 tɕjɨj R43 2.37 tɕjij5
27.3.1 TT3301 ɕjɨj R43 2.37 ɕjij5
25.6.2 TT3973 rjɨj R43 2.37 ljij4'
4.4.4 tshjij1 TT1229 tshjɨj R43 2.37
27.3.3 ɕjij5 TT3301 ɕjɨj R43 2.37

I could reconstruct R43 as -jə by lowering Gong's ɨ and dropping the -j, just as I reconstructed R36/R37 as -je by lowering Gong's e and dropping the -j. However, dropping -j throughout the rhyme group that R43 belongs to would make R41 and R42, the j-less Grade I and II counterparts of Grade III R43, the same as R28-29 (solution 1 below). Two other possibilities keep R28-29 and R41-42 distinct:

Rhyme Grade Gong Solution 1 Solution 2 Solution 3
R28 I
R41 I -əj -əj -ej
R29 II -iə ʕ ʕ ʕ
R42 II -iəj ʕj -eʕj
R30/R31 III -jɨ -jɨ -jɨ -jɨ
R43 III -jɨj -jə -jəj -jej

(All three solutions of mine incorporate the 'emphatic' interpretation of Grade II.)

I prefer solution 3 since it places e-rhyme groups next to each other:

R28-33: group

R34-40: -e group

R41-43: -ej group

R44-49: -ew group

Both solution 2 and Gong's reconstruction alternate ə- and e-rhymes:

R28-33: group

R34-40: -e group

R41-43: -əj group

R44-49: -ew group

That would go against the general tendency for rhymes with similar vowels to be grouped together in Tangraphic Sea.

The -ej group of solution 3 is to -e what the -ow group is to -o:

Mid front palatal vowel Mid front palatal vowel + labial glide Mid back labial vowel
no glide R34-40: -e group R44-49: -ew group (mixing characteristics of R34-43 and R50-60) R50-55: -o group
glide R41-43: -ej group R56-60: -ow group

The strange placement of R50 -jwo between R49 -jiw (possibly -jew?) and R51 -o may indicate that it was meant to be transitional (palatal like R49 but an o-rhyme like R51).

Proposing a mid vowel for R36/R37 and R43 can also account for cases in which those rhymes correspond to 12NWC low (lower mid?) vowels:

Pearl Graph Transcribed as
20.6.4, 22.4.2 sja5 TT0054 ɕjij R36 1.35
(in the Tangut translation of Leilin,sja1 is transcribed as TT1175 ɕjij R36 1.35)
30.6.5, 30.6.6 TT1174 sjij R37 1.36 sja2
34.6.6 TT3668 sjij R37 1.36
(in the Tangut translation of Leilin, 冶野 ja4' and 夜 ja6 are transcribed as TT3708 ʔjɨj R43 1.42)
22.3.8 TT0675 ɕjɨj R43 2.37 捨 ɕja3
21.6.11 TT3606 ʔjɨj R43 2.37 ja4'

I suspect that 12NWC *-ja was more like *-jɛ (which is Pulleyblank's reconstruction of the 13th century eastern Chinese counterpart of that rhyme).

R37 was sometimes transcribed with 12NWC *-ji:

Pearl Graph Transcribed as
20.2.6, 20.5.1 TT1235 njij R37 1.36 nji4'
32.3.6 TT2875 njij R37 2.33
17.1.8 TT5673 njij R37 2.33

Gong's R37 -jij is not far from his 12NWC *-ji, but it does not match his 12NWC *-ja very well (unless the Tangut ignored the -a after the medial -j-). My R37 -je is like both 12NWC *-ja (or my *-jɛ) and *-ji.

Next: How have others reconstructed these rhymes? Can rhyme alternations and comparative evidence confirm a front vowel in R41-43?

*The number of tones in 12NWC is unknown. The numbers refer to MC tone categories:

1/2 'level'

3/4 'rising'

5/6 'departing'

(7/8 'entering' is not represented in the data.)

Odd numbers indicate tones for voiceless-initial syllables and even numbers indicate tones for voiced-initial syllables. The apostrophe after 4 indicates a syllable with a voiced sonorant initial. In modern Chinese languages, 4 may merge with 6, but 4' tends to merge with 3.

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