11.9.3.23:59: REVISITING THE TANGUT RITUAL LANGUAGE 4: A PEARL-Y PROBLEM

In part 3, I proposed that most known Tangut texts (e.g., translations of Buddhist and - I should have added - secular Chinese texts) are in the Mi language because it was more prestigious than the Lhwe language presumably spoken by commoners. A few partly or wholly Lhwe-based early cultural words (e.g., 'Tripitaka') were deeply embedded even into Mi, but in general, Mi was the language of high culture.

If the majority spoke Lhwe, why was the Tangut-Chinese glossary The Tangut-Chinese Timely Pearl in the Palm in Mi (Tangut in the title was 2mi) with a few exceptions: e.g.,

3057 5855 2ʒɨu 2lọ 'Twin Fish' (name of a constellation; a Mi-Lhwe mix)

5855 5549 2lọ 2tʃhɨw 'Twin Girls' (name of a constellation; a Lhwe-Mi mix)

I noted how these phrases had opposite word orders in part 1 of "Two Many Words". Could one order - the noun-numeral order atypical for Mi - be Lhwe word order? The coexistence of two word orders in Tangut reminds me of Vietnamese which contains both indigenous and Sinitic-style word / morpheme orders: e.g.,

Tiếng Việt 'Vietnamese language' (modified-modifier; indigenous order; loans from Chinese 聲 and 越 [the latter in turn from a name like *Wat in some ancient non-Chinese language)

Pháp Ngữ 'French language' (modifier-modified; Chinese order; loan from Chinese 法語)

Chinese and Vietnamese coexisted in Vietnam for nearly a millennium before Vietnam became independent. Similarly, Mi and Lhwe coexisted in the Tangut Empire for centuries. The Lhwe never got their own country because the Mongols conquered the Tangut Empire and made them part of an even larger, more diverse state.

The Pearl teaches Mi rather than Lhwe to Chinese because

- Mi was the official language of the Tangut Empire

- foreigners were not expected to learn the local languages, just as students going to Hainan to study 'Chinese' (Mandarin) do not study any of the local languages of Hainan which belong to four different families: Chinese (but not the Mandarin branch!), Kra-Dai (Hlai/Li, Be, and Zhuang), Hmong-Mien, and Austronesian.

- Lhwe were bilingual in Mi and could use a Mi-based glossary to learn Chinese, just as the peoples of Hainan are bilingual in Mandarin and can use materials in Mandarin to learn English

'Though wise men appear among the Chinese, the Hsi-hsia [Lhwe] will not revere them.'

(Translation by Nishida 1964: 187; also see Li Fanwen 2008: 150; the text reproduced in Kwanten 1982: 227 ends differently*)

One could interpret this as saying that the common Tangut - unlike the Sinophilic Mi - did not respect great Chinese men. The preceding passage was

'Though there are worthy men among the Hsi-Hsia [Mi], the Chinese will not respect them.'

(again in Nishida's translation)

Note that this does not state that there are worthy men among the Mi and the Lhwe. The Chinese could not even see the worth of the elite Mi, let alone the lesser Lhwe. In Chinese eyes, the Mi were 'barbarians'.

The goals of the Pearl were to facilitate Chinese recognition of the Mi and inculcate reverence for Chinese culture among the Lhwe. This sounds like a Mi-oriented worldview. I conclude that Kwyli Rirphu (a.k.a. 骨勒茂才 Gule Maocai in Chinese), the author of the Pearl, was either a Mi or a Lhwe who had fully absorbed the Mi POV.

*9.4.1:39: The Tangut text in Kwanten reads,

'Lhwe person how trust' = 'How can the Lhwe people trust [the wise men among the Chinese without knowing Chinese]?'

番人不崇

'Barbarian people do not revere [the wise men among the Chinese]'

11.9.2.23:59: REVISITING THE TANGUT RITUAL LANGUAGE 3: ALMOST ALL ABOUT MI

In part 2, I asked,

Conversely, why is the elite language so ... common?

In other words, why is almost all Tangut writing in the Mi language of the minority? Why is the Lhwe language of the majority so rare?

Andrew West compared what I consider to be the Lhwe and Mi languages to Old English and Latin. That led me to  compare Lhwe and Mi to Korean and Classical Chinese. Prior to the invention of hangul in the 15th century, there is very little material in Korean which was not a prestigious language in Korea. There are only 25 surviving poems in Old Korean (鄉歌 hyangga), and "[w]e quite honestly do not know what some hyangga mean, much less what they sounded like" (Lee and Ramsey 2011: 57). If Korean became extinct before the invention of hangul and one found a library of writings by Koreans, one would see nothing but book after book in prestigious Classical Chinese with occasional hints of nonprestigious Korean. Similarly, the majority of Tangut texts are in prestigious Mi with occasional hints of nonprestigious Lhwe. Writing is not necessarily a mirror of popular usage. The prestigious is preserved, but the rest is largely unwritten.

Andrew noted that Lhwe numerals occur in combination with Mi words in Mi texts: e.g.,

4344 2lheʳ (Lhwe) + 4730 2ʔu (Mi)̣ 'three storehouse' = 'the Tripitaka' (which may also be etymologically mixed: tri 'three' is Indo-Aryan but piṭaka 'basket' may be from a substratal language)

I suspect this word could be from an early wave of Lhwe-based Buddhist terminology. 2ʔụ 'storehouse' has no obvious Sino-Tibetan cognates and could be a Lhwe loan into Mi. So the entire compound could be of Lhwe origin. Or the Mi could have introduced storehouses and their word for them to the Lhwe.

Buddhist terms like 2lheʳ 2ʔụ 'Tripitaka' are comparable to

Korean 부처 Puchhŏ < 부텨 *Puthyə < *?Putke 'Buddha'

Japanese Hotoke < Potəkəy < *Potəkai 'Buddha'

which probably belong to the earliest strata of Buddhist terminology in those languages* and contain Korean peninsular morphemes with *-k-**.

Sino-Korean 불 Pul and Sino-Japanese Butsu 'Buddha' were borrowed from NE Late Middle Chinese 佛 *bur and Early Middle Chinese 佛 *but. They are part of a later wave of Buddhist terminology corresponding to a later wave of Mi-based Buddhist terminology:

Mi-based terms may have replaced many but not all of the earlier Lhwe-based terms. Or a handful of Lhwe terms were coined as the Lhwe were just getting into Buddhism whereas the Mi got deeply into Buddhism and coined a large Mi-based vocabulary for it. With rare, deeply embedded exceptions like 2lheʳ 2ʔu,̣ Tangut Buddhist terminology was based on Mi, just as nearly all Korean and Japanese Buddhist terminology is based on Chinese.

Next: A Pearl-y Problem

*9.3.1:06: Other examples of early Korean and Japanese Buddhist terminology:

K 절 chŏl < 뎔 tyər < ?*ter, J tera 'temple' (unrelated to Middle Chinese 寺 *zɨəh or Skt vihaara)

K 중 chung < 듕 tyuŋ 'monk' (unrelated to Middle Chinese 僧 *səŋ < Skt saṅgha)

**9.3.9:07: Korean *Putke probably contains a Shilla morpheme *ke corresponding to a Paekche morpheme *ka(i) in Japanese *Potəkai. (The final *-i could be a Japanese suffix added to a Paekche *-ka.) The meaning of this *k-morpheme is unknown. I don't know of any other proto-forms with it.

I also don't know why the Japanese form has *o instead of *u. I would not expect an *o for two reasons:

First, the word is probably from Paekche and judging from Sino-Japanese readings probably borrowed through Paekche, Paekche had a /u/ : /o/ distinction, so there would be no reason for a Paekche speaker to borrow Early Middle Chinese 佛 *but or Skt Bud(dha) as *pot(ə) with *-o- instead of *-u-.

Second, a non-final pre-Old Japanese *o normally raises to Old Japanese u, merging with original pre-OJ *u, but pre-OJ *Potəkai did not become OJ *Putəkəy. Was *o retained in a sacred name?

11.9.1.23:56: REVISITING THE TANGUT RITUAL LANGUAGE 2: INVERTING THE HIERARCHY

In part 1, I asked,

So what do I think the [Tangut] 'ritual language' was?

My current answer flips Nishida's 1986 answer on its head. Andrew West summarized Nishida's view:

Nishida hypothesised that Vocabulary I [Kepping's 'ritual language'] represents the language of the "black-headed people" who he considers were a nomadic people that formed the ruling class of Tangut society; and that Vocabulary II [Kepping's 'common language'] represents the language of the "red-faced people" who he considers were a sedentary, agricultural people that formed the bulk of Tangut society. In other words, Vocabulary I may represent a linguistic substratum that was only preserved in the odes of the ruling class.

Here's what I think happened:

The ancestors of the majority of the Tangut spoke some non-Sino-Tibetan language which was an isolate. Hence its words (e.g., 1ka 1ʔo 'moon') had no cognates in neighboring languages. Let's call these people the

0883 Lhwe (1lhwiẹ) 'Tangut'

after the 'ritual' equivalent of the 'common' word

2344 Mi (2mi) 'Tangut'

A Tibeto-Burman speaking people conquered the Lhwe. The two languages came into contact so intense that their phonologies converged* and even basic vocabulary was borrowed. Just as the Vietnamese borrowed đầu 'head' (cf. Middle Chinese 頭 *dəw 'head') from their Chinese rulers, the Lhwe borrowed 1ɣʊ 'head' (cf. Written Tibetan mgo 'head' and Old Chinese *goʔ 'ruler') from their Mi rulers. This is why the Mi word 1ɣʊ 'head' appears in the Lhwe glosses of Mi poetry; it has become a Lhwe word.

Line 1A of the Ode on Monthly Pleasures contains only one word that is unique to Lhwe: 1ka 1ʔo 'moon'. The other words are shared with Mi. If those words have Sino-Tibetan cognates (e.g., 1ɣʊ 'head'), then they are borrowings from Mi. Lhwe could be full of Mi borrowings. The cover of this Vietnamese orthographic dictionary has only one native Vietnamese word (mới 'new'); the rest - including the components of the author's name - are borrowings from Chinese. Shared words without ST cognates could be Lhwe words that were borrowed into Mi or the other way around.

If Andrew is correct and the 'ritual language' (= my Lhwe) lines are glosses rather than full lines, then they may not tell us anything about Lhwe grammar. Given an intense contact situation, I would expect Lhwe and Mi to have similar grammar, but I cannot rule out the possibility of great dissimilarity. There were and are many languages in contact with Chinese that have grammars varying from Chinese to different extents: e.g., Khitan and Jurchen (with 'Altaic'-style grammar) and Vietnamese (with 'Southeast Asian'-style grammar). Lhwe words in glosses may be isolated citation forms: e.g., 1ka 1ʔo 'month' could have had other forms depending on case and number. So it's possible that no true Lhwe sentences are preserved in extant writing. Conversely, why is the elite language so ... common?

Next: Why's It Almost All About Mi?

*The name of the Lhwe (1lhwiẹ) and the Lhwe numeral

4344 2lheʳ 'three'

contain unusual features shared with Mi:

1. the initial lh- (also in Written Tibetan)

2. phonemic vowel tension

3. phonemic vowel retroflexion

4. a two-way 'tonal' distinction (possibly a phonation distinction: e.g., plain vs. breathy?)

Whether these features resulted from Mi influence or vice versa is unknown, but I doubt that both languages happened to have this complex of characteristics prior to contact.

11.8.31.23:03: REVISITING THE TANGUT RITUAL LANGUAGE 1: ARTIFICIAL HAWAIIAN LANGUAGES

Andrew West's "The Myth of the Tangut Ritual Language" is worthy of an entire series of posts in response. For now I'm only going to address one point:

In 1996 the Russian Tangutologist Ksenia Kepping formulated the terms "common language" and "ritual language" to refer to these two forms of vocabulary, suggesting that the ritual language was an artificial language (lacking grammatical morphemes) created by Tangut shamans in ancient times, before the adoption of Buddhism, for ritual purposes.

Kepping's paper is online, though unfortunately it lacks the 'ritual language' text that was supposed to be in section E at the end. This text is in Nishida (1986).

On Monday night, I rediscovered Albert J. Schütz' The Voices of Eden: a History of Hawaiian Language Studies (1994) when looking for examples of Hawaiian dialectal variation. Along the way I found a section on two artificial Hawaiian languages that reminded me of Kepping's hypothesis.

The first was a priestly language incomprehensible to lay people. (The description sounds like glossolalia to me, so this may not have been a language.) The second was a language including grammatical morphemes devised by a Kamehameha (it is not clear which one) upon the birth of a son in 1800. This language died along with the son.

Two observers in the 1860s "wrote that the chiefs invented new words to keep themselves separate from the commoners" (Schütz 1994: 47).

I have never thought that the Tangut 'ritual language' was an arbitrary invention like these Hawaiian languages. I doubt that priests decided out of the blue to call the moon 1ka 1ʔo instead of 2lhị. So what do I think the 'ritual language' was?

Next: Inverting the Hierarchy
̣

11.8.30.23:36: PSEUDO-SINO-TANGUT

Although Tangut is undoubtedly full of Chinese loanwords - 'Sino-Tangut'? - like

5747 2ʃɔɔ 'double' (< Tangut period NW Chinese 雙/双 *ʃɔ̃)

3488 2twəəi 'pair, couple' (< Middle Chinese 對 *twəjh)

from my last entry, not everything that resembles Chinese may be a loanword.

Tangut has a number of negative morphemes

1918 1mi 'not' (the most generic of the set)

5643 1miə 'not' (for auxiliary verbs),

2194 1mie 'not' (for existential verbs), 'to not have'

1064 2mie 'to not yet be'

with an initial m- that is reminiscent of the *m- in Old Chinese, Middle Chinese and Tangut period northwestern Chinese negative morphemes.

莫 OC *mak > Late OC, MC *mak > TPNWC *mo

亡 OC *maŋ > Late OC *mɨaŋ > MC *muaŋ > TPNWC *waŋ

無 OC *ma > Late OC *mɨa > MC *muo > TPNWC *wu

毋 OC *mə or *mo > Late OC *mɨə or *muo > MC *muo > TPNWC *wu

勿 OC *mət > Late OC *mɨət > MC *mut > TPNWC *wu

But the Tangut vowels don't match the Chinese vowels (unless 毋 was OC *mə). This is not necessarily a problem because pre-Tangut *a is known to raise to various degrees: -ɛ, -e, -ie, -əi, -iə, -i (with -ɨ- instead of -i- depending on initial; Matisoff 2004; Tangut reconstructions are mine). I suspect different presyllables conditioned different degrees of raising. However, I am reluctant to reconstruct high-frequency morphemes with presyllables or complex initials: e.g., *C(i)-ma for 1mi. Perhaps pre-Tangut *-a always raised unless it were blocked by some conditioning factor:

*ma (cf. Written Tibetan ma 'not') > 1mi

But then that factor would have to be present in the pre-Tangut ancestors of all native Tangut -a syllables.

In any case, these Tangut m-negatives were probably inherited from Proto-Sino-Tibetan, the ancestor of Tangut and Chinese, rather than borrowed from Chinese long after it had split from the rest of Sino-Tibetan. The Tangut vowels don't match the Chinese vowels because both changed independently over millennia.

(8.31.00:18: Could pre-Tangut have had two negative morphemes, *mi and *ma, like Classical Tibetan mi and ma? Then 1mi would simply be a retention of *mi while the other morphemes might have been derivatives of *ma with vowel raising.

The distribution of Tangut 1mi and non-1mi m-negatives which may be from *ma does not parallel that of Classical Tibetan mi (for present and future forms of verbs) and ma (for past and imperative forms of verbs). Tangut does not have a Classical Tibetan-style four-way distinction between present, past, future, and imperative.)

Tangut also has non m-negatives:

1734 1ti 'don't ...!' (imperative)

1943 2niaa 'no' (as an answer to a question), 'don't ...!'

One might want to derive 2niaa from *miaa, but there are no other cases of *m- becoming n- in Tangut.

The analysis of Tangut negatives in this post is based on Gong (2003: 614-617).

Similar but 'separate'

Li Fanwen (2008: 874) identified

5550 1lɨə̣ 'to leave, to depart from'

as a borrowing from Chinese 離 'to separate'. However, 1lɨə̣ goes back to pre-Tangut *Slə which is not a good match for any reading of 離 through time:

OC *raj > Late OC *lɨaj > MC *lie > TPNWC *li

1lɨə̣ cannot be a cognate rather than a loanword. One could claim that pre-Tangut *S- (the source of vowel tension) was an affix that didn't have to correspond to anything in Chinese, but the pre-Tangut root segments *l- and *-ə do not correspond to OC *r- and *-aj.

I regard words like 1lɨə̣ as pseudo-Sino-Tangut.

Excluding pseudo-Sino-Tangut which can be regarded as a 'zero' (i.e., meaningless) category, Chinese-like words in Tangut belong to at least three strata:

2. Borrowings from Middle Chinese: e.g., 3488 2twəəi 'pair, couple' < Middle Chinese 對 *twəjh.

8.31.1:10: Notice that there is no known stratum of borrowings from Late Old Chinese. This suggests that there was little or no contact between Tangut and Chinese until the Middle Chinese stratum dating from a few centuries before the earliest Tangut texts.

The only Late Old Chinese-like Tangut word I can think of is

4106 1lɨə̣ < *S-lə 'to come; to call (to come?)' (not attested outside dictionaries?; 'ritual' Tangut?)

which resembles LOC 來 *lə < OC *mʌ-rə 'to come' but has somewhat different semantics and seems to be cognate to

3456 1lɨa 'to come' < *Cɯ-la

3456 is phonetic/semantic in 4106. The function of

'wood' (alphacode: box)

instead of the similar-looking

'motion' (alphacode: fis; in 4481 'to go')

at the top of 4106 is unknown. 4106 and 3456 have a circular analysis:

=+

4106 = 4307 1riaʳ 'to come, to invite' + 3456

=+

4307 = 4106 + 5479 1riaʳ 'illness, disease' (phonetic)

The shared parts of 4307 and 5479 (alphacode: juscia) do not comprise an independent tangraph.

8.31.1:41: juscia is obviously a reversal of the elements of 3456 (alphacode: ciajuu).

5479 is in turn derived from 4307:

=+

5479 = 4307 + 2857 2ŋo 'illness, disease' (semantic)

=+

3456 = 4106 + 4307

Next: Tangut 'Ritual' Language: A Hawaiian Parallel?

11.8.29.23:59: TWO MANY WORDS (PART 3)

The last of the four Tangut equivalents of Chinese 雙/双 'double' in the Chinese-Tangut index of Li Fanwen 2008 is

=+

5747 2ʃɔɔ <  Tangut period NW Chn 雙/双 *ʃɔ̃ =

bottom of 4326 2gəu (classifier used to count bows and chariots - perhaps also pages and items and/or corpses)

Does the 'wood' on top signify the substance out of which bows and chariots were made?

Why 'double' beneath 'wood'? Is it a reference to pairs of wheels? To the 兩 'both' in 輛 (a classifier for vehicles < 'two(-wheeled thing)'?)?

2gəu sounds like Middle Chinese 具 *guoh, a classifier for counting tools and corpses that is one of Li Fanwen's translations of 5747. 具 happens to be phonetic in 俱 *kuo, which Karlgren (1957: 50) translated as 'both (!), all, together'. 'Both' is just a very limited kind of 'all': 'two out of two'.

+ right of 3488 2twəəi 'pair, couple' < Middle Chinese 對 *twəjh

(Tangut period NW Chn 對 *twe would have been borrowed as 2twe(e). What Gong and I have reconstructed as vowel length may have been something else. It is not clear why Chinese words were borrowed both with and without long vowels since no stage of Chinese known to the Tangut is known to have had phonemic vowel length. Did the Tangut sometimes hear Chinese vowels as short and sometimes hear them as long?)

8.30.2:09: This tangraph is also read 3489 giaa 'stupid' with a fitting analysis in the Combined Homophones-Tangraphic Sea:

=+

left and top center of 2702 1lwɨẹ̃ 'ignorant, muddle-headed' analyzed as

=+

left of 1918 1mi 'not'

+ center and right of 1dʒwɨo 'bright, clever'

right of 3979 1vəị 'foolish, stupid':

=++

left of 3916 2si (an auxiliary word)

+ left of 4554 2lə 'foolish' (the resemblance to Tangut period NW Chinese 魯 *ləu 'stupid' is probably coincidental, as that would have been borrowed into Tangut as ləu)

+ left of 2442 1noo 'weak' (cognate to Old Chinese 弱 *nekʷ 'weak'? - long vowels in Tangut words may be reflexes of lost final stops)

Note how the right side of 3489 is split into top and bottom halves.

The doubled

4855 2dzie 'one' (for paired body parts: eyes, hands, feet; alphacode: hae)

of

5747

is reminscent of the doubling of 隹/又 in 雙/双. I finally realized that 4855 'one' may be derived from a cursive form of Chinese 隻 'single, individual'. Hence

5747 'double' = 4855 'single' x 2

3488 'pair' = 'not' (< Chn 不?; alphacode ) + 'single'

8.30.00:40: hae appears in 90 tangraphs, but it probably does not mean 'single' in many or even most of them.

11.8.28.21:49: TWO MANY WORDS (PART 2)

In Part 1, I looked at 5914 and 5855 which are probably derivatives of 4027, the most basic tangraph for 'two':

<>

5914 2lọ < 4027 1niəə > 5855 2lọ

Grinstead (1972: 75) and Nishida (1986: 30) identified 5914 as 'second son', but Li Fanwen (2008: 5914) defined it as 'two; second' and did not list any examples of it as 'second son' in isolation as opposed to the phrase

5914 1448 2lọ 2jɨw 'second son'

with 1448 2jɨw 'son'.

5855 was one of four Tangut equivalents of Chinese 雙 'double' in the Chinese-Tangut index of Li Fanwen 2008. The Tangraphic Sea derived a second equivalent (0468) from 5855:

=+

0468 1phɔ̃ 'double' =

left of 1572 1phɔ̃ 'white' (phonetic) +

right of 5855 2lọ 'double' (semantic)

Li Fanwen 2008 lists no examples of 0468 by itself outside a dictionary. Even in dictionary defintions, 0468 appears after 0517 in

0517 0468 2bəi 1phɔ̃

which was listed alongside 5914 and 5855 as an equivalent of 4027 in Tangraphic Sea. I suspect that 0517 0468 is a disyllabic ritual Tangut (RT) word for 'two' or something like it ('double'? 'pair'?), even though it doesn't appear in the ritual Tangut name of the second month:

0795 5855 1846 0863 2riə 2lọ 1ka 1ʔo

0795 is normally a prefix for an unknown direction (Gong 2003: 608) but is presumably representing an unrelated homophonous syllable (prefix?) before 5855.

5855 can stand alone. See Part 1 for examples.

1846 0863 is the RT word for 'month' corresponding to common Tangut (CT) 2lhi.̣

The analysis of 0517 is unknown, but the left side must be phonetic:

=+?

0517 = 1121 2bəi (or 1116, 2771, 3366, 4374 - the other four tangraphs containing 1121) +

0497 2ŋeʳw 'number' or one of the other 1,186 (!) tangraphs containing 'person' (alphacode: dex)

1121 is a surname tangraph. It may be based on Tangut period NW Chinese 馬 *mba 'horse'.

0517 strongly resembles 0518, yet another Tangut equivalent of Chinese 雙 'double':

0517 2bəi vs. 0518 2baʳ

Unlike 0517, 0518 can stand alone.

0518 has a rare element (alphacode hoi) that I covered in part 13 of my Tangraphic radicals series.

The Combined Homophones-Tangraphic Sea analysis of 0518 derived it from 0765, the only other tangraph with hoi, and its lookalike 0517:

=+

0518 = left of 0765 2khiəə 'steep' + right of 0517

Last year, I asked,

[I]s the shared initial [of 0518 and 0517] indicative of a common root b-?

0518 2baʳ < *r-baH or *baH-r

0517 2bəi < *CI-baH

*H is the pre-Tangut glottal source (*ʔ and/or *h) of the rising tone.

*CI- is a presyllable that conditions fronting of the vowel in the following syllable. For other cases of *a becoming Tangut -əi (Gong's -e), see section 2.2 of Matisoff (2004).

Next: Twins under Wood