The first character in the fragment I discussed in my last post


3543 1dzwy1 'chapter' = 3020 1ja'3 'reality' + 5334 1dzwy1 'to catch' (phonetic)

contains a component

found in only three other characters:


3020 1ja'3 'reality' = left of 3543 1dzwy1 'chapter' (circular!) + 'speech' (left of 1817 2daq1 'to know')


3523 2ja'3 'skinny, wan and sallow' = left of 3020 1ja'3 'reality' (phonetic) + all of 4675 2rer4 'toil' (why?)


3021 1ja'3 'chapter' = possibly left of 3543 1dzwy1 'chapter' + ?

I suppose their shared component is a phonogram for ja'3 that is semantic in 3543 if it is an abbreviation for 3021 1ja'3 'chapter':


3543 1dzwy1 'chapter' = 3021 'chapter'? + 5334 1dzwy1 (phonetic)

If 3021 predated 3543, then 3543 cannot be the true source of the left side of 3021 (though such a circular analysis might have appeared in the Tangraphic Sea if it had an entry for 3021). 3020 1ja'3 or 3523 2ja'3 may be an abbreviated phonetic in 3021 1ja'3. HOW LONG DOES IT TAKE TO LEARN THE GOLDEN GUIDE?

Andrew West identified a fragment of Tangut writing practice (British Library Or. 12380/2625) that puzzled me back in 2011 as being from the preface to the Golden Guide:

Location 01B508 01B509 01B510 01B511 01B601
Li Fanwen number 3543 0113 1771 4735 2814
My transcription 1dzwy1 1shen3 2seq4 2ja3 2lhiq4
Gloss chapter to accomplish wisdom sharp month

Here is the full context of this passage from Kotaka (2005) with his translation:

1ngwy1 2di4 2gwi4 2pho'4 1lyr'3 1ny'4 1dzwy1 1shen3

lit. 'five character phrase collect four two chapter accomplish

'Collecting phrases of five characters, [we] made four-two chapters.'

1nwy1 2ja3 2lhiq4 1oq2 2zeq4

lit. 'know sharp month round finish

'The sharp of mind may finish [them] in one month.'

Kotaka thought

1lyr'3 1ny'4 'four-two'

meant 'eight', though he broke the Golden Guide into three sections with ten subsections in this 2003 article:

I. The Tangut world and its government

A. The creation of heaven and earth, the seasons, the Heavenly Stems and Earthly Branches, the changing of the days and years

B. The life and work of the court, the imperial family, and its officials

II. Names

A. Tangut surnames

B. The Tangut people and surrounding peoples; the Jurchen are excluded, leading Nishida (1997) to guess that the Golden Guide was composed before the rise of the Jurchen Empire in 1115

C. Chinese surnames

III. Tangut daily life

A. Marriage and kinship terms

B. Luxury items and clothes

C. Utensils, food, monks

D. Beasts, birds, domestic animals

E. Daily life of the masses

Although Kotaka (2005) used IOM #741 as his basis, his typeset version of the Golden Guide has

2699 instead of 1771

for 01B510. Both IOM #741 and British Library Or. 12380/2625 clearly have 1771. I translate

1seq4 2ja3

as '[those with] sharp wisdom' or 'the wise and sharp'.

Kotaka translated

2lhiq4 1oq2

as 'one month'. The first half is 'month', but what is the second half 'round' doing? Maybe this phrase means 'in about a month'. Are there any other instances of quantity words followed by 'round'?

Lastly, Kotaka translated


as a verb 'finish', though Kychanov and Arakawa (2006: 283) and Li (2008: 518) translated it as an adjective 'honest'. Kotaka's translation is closer to Nishida's (1966: 432) 'to do'.

The phrase

1seq4 2ja3 2lhiq4 1oq2 2zeq4

'wisdom sharp month round ?'
is parallel in structure to the following phrase:

1viq1 1lwen1 1kew4 1mi4 1chen3

'foolish obtuse year not ?'

which Kotaka translated as 'Even the stupid will take less than a year [to finish it].'

1chen3 normally means 'correct' but can also mean 'to pass'. 'Correct' is similar in meaning to 'honest' for 2zeq4. Could 1kew4 1mi4 1chen3 'year not correct' mean 'not precisely a year'?

I would then expect 'not precisely a month' in the previous line. But if 2lhiq4 2zeq4 were 'precisely a month' (lit. 'month honest'), wouldn't a modifier follow 'honest' instead of precede it: i.e., *2lhiq4 2zeq4 1oq2 'about (lit. round) an honest month' instead of 2lhiq4 1oq2 2zeq4 'honest approximate (lit. round) month'?

Maybe 2lhiq4 1oq2 2zeq4 means 'precisely a full month' if 1oq2 is 'full' as in

1oq2 1sy1 'full'

which may be a redundant compound since 1sy1 means 'full' by itself. CLEAR FOUNTAIN TEMPLE?

I often play the game of 'guess the Chinese characters' when seeing Korean in hangul or romanization or Vietnamese in Quốc ngữ (國語 'national language'; i.e., the Vietnamese alphabet).

A Buddhist temple named Chùa Thanh Nguyên was on the front page of Sunday's Honolulu Star-Advertiser. (Yes, this post is four days late.)

I think the characters for Thanh Nguyên are 清源 'clear fountain'. There are 清源寺 'Clear Fountain Temples' in China, Korea, and Japan.

The Sino-Vietnamese reading for 寺 'temple' is tự; its native (?) equivalent is chùa, written as 廚 chù 'kitchen' (phonetic) with or without 寺 'temple' (semantic) beneath it in the nom script.

For years I have been puzzled by chùa which should go back to an earlier *ɟuə.

I would expect a word for a Buddhist temple to be a loanword. (English temple is from Latin templum.) But there is no Chinese word for 'temple' like *ɟuə. tự is from a southern Late Middle Chinese form like *sɨ̀ which in turn goes back to Late Old Chinese *zɨəʰ. *zɨəʰ is close to *ɟuə, but not close enough. If *zɨəʰ were borrowed into early Vietnamese, it might have become *ɟ(ɨ)ə(h) which would have developed into modern *chờ, *chỡ, chừa, or *chữa without a labial vowel, not chùa with a labial vowel.

Moreover, I can't find any Vietic cognates of chùa, so I am not sure it is a native word. Could its nom spelling 廚 be etymological: i.e., is it a borrowing from Early Middle Chinese 廚 *ɖuə 'kitchen'? Although the initials are not a problem (Early Middle Chinese 重 *ɖuoŋʰ 'heavy, important' corresponds to Vietnamese chuộng 'to esteem' [i.e., regard as important]), there is a vast semantic gap between 'temple' and 'kitchen'. Is a shift from 'kitchen' to 'temple' attested in another language?

While I'm on this topic, I should mention another Vietnamese word that I cannot explain. A Buddhist monk is a thầy chùa. Thầy 'master' has no Chinese source that I know of. Moreover, its huyền tone would normally indicate a *voiced initial, even though I doubt Proto-Vietic had voiced aspirated *dʱ- or *ʑ-. (Sino-Vietnamese th- followed by a huyền tone is from southern Late Middle Chinese *ʑ-.) Nom spellings of thầy contain the voiced-initial phonetic 柴 sài < *draːj with optional semantic additions 亻 'person' or 師 'master'.

Another puzzling word of this type that is certainly native is thịt 'meat' whose Vietic cognates have voiceless s- even though a nặng tone also normally indicates a *voiced initial. (Proto-Vietic *s- normally became Vietnamese t-, not th-.) Nom spellings of thịt contain the voiced-initial phonetic 舌 thiệt < *ʑiet with optional semantic additions 月 'meat' or 肉 'meat'.

Could the tones of 'master' and 'meat' reflect voiced prefixes? Did those prefixes condition aspiration? Or did Proto-Vietic have *dʱ- or *ʑ-? MOCANG, MOZANG, MIZANG, MYDZON?

While looking through for references to the Uyghur in the Tangut Empire in Dunnell (1996), I found a passage about Lady Mocang (Chinese Wikipedia entry), mother of Emperor Yizong. Mocang is how Dunnell read 没藏; the second character can also be read as zang in Mandarin. Is there any premodern note specifying the pronunciation of 藏?

Dunnell (1996: 56) also mentions another Chinese transcription pronounced Mizang in modern Mandarin in 李燾 Li Tao's 續資治通鑑長編 Xu Zizhi tongjian changbian, but I only see 没藏 Mocang/Mozang in this online edition. Alas, she does not provide characters for Mizang in the back of her book.

Although the surname Sinified as Mocang/Mozang (and Mizang?) is important in Tangut history, I don't know what its Tangut characters were. The closest matches in the surnames in Mixed Categories are

3682 5164 1my1 1dzon1 and 2888 5164 2my1 1dzon1

3682 is a word for 'merit' that may not be attested outside dictionaries. Its meaning is not apparent from its Tangraphic Sea analysis:


3682 = 2216 1teq4 'swift' (semantic?) + 3513 1my1 'heaven' (phonetic and semantic?)

2888 means 'surname' and is obviously from 'person' + 'surname'; it also appears in the surname

2888 1085 2my1 1zi4

with 1085 'man'. I would not expect 'surname' in a surname.

5164 is not attested alone; it is a second syllable in six other surnames in Mixed Categories:

2214 5164 2ly3 1dzon1 and 4698 5164 2ren4 1dzon1
3219 5164 1pa1 1dzon1 and 1989 5164 2vy3 1dzon1
3334 5164 1ma4 1dzon1 and 3889 5164 2be'4 1dzon1

2214, 3219, 1989, and 3889 are surname characters.

4698 is a surname and toponym character.

3334 means 'female'.

Does the Tangraphic Sea analysis of 5164 tell us something about the eight -dzon families?


5164 = 5031 1by1 'second half of Lyby, ancestor of the Black-Headed Tangut' + 1332 1de1 'to pass on' + 2132 2ew4 'achievement'

Were those families considered to be direct descendants of Lyby?

3682 1my1 and 2888 2my1 are good phonetic matches for 没 which would have been pronounced something like 4my1 in Tangut period northwestern Chinese, but 5164 1dzon1 has a voiced initial unlike 1tshon1 and 3tshon1, the Tangut period northwestern Chinese readings of 藏.

2.18.0:45: Conversely, 藏 did have a voiced initial in the eastern dialect(s) underlying the later Phags-pa Chinese reading ꡐꡃ <tsaŋ> [dzaŋ] (with both tones 1 and 3), but the final does not match Tangut -on (which may be from pre-Tangut *-am, *-em, *-om, or *-um; see Jacques 2014: 206). SOUTHERN MOUNTAINS IN XINJIANG MANDARIN

Yesterday I rediscovered a note from last October for a future post idea. I had found this passage in Hahn (1998: 383) which happened to contain the Uyghur equivalent of Turkish adamlar 'men' from my previous post (emphasis mine):

Loanwords [in Uyghur] may contain both front and back elements, but any suffix attached to such a loanword stem takes its harmonic information only from the last syllable of the stem. This may be illustrated by means of the suffixes -lAr and -dA attached to adem 'human being' < Arabic ʕādam, polek 'Pole' < Russian poljak, šenduŋ 'Shandong' < Mandarin Shāndōng and Xunen 'Hunan' < Mandarin Húnán: ademler 'human beings', polekler 'Poles', Šenduŋda 'in Shandong', Xunende 'in Hunan'.

Why does Uyghur front e correspond to standard Mandarin nonfront a? I think the key word is "standard". I don't have any data on the Mandarin spoken in Xinjiang, so I can only guess that the source dialect of those loanwords might be like other Mandarin dialects which have fronted the vowel of *-an: e.g., 南 næ̃ 'south' and 山 sæ̃ 'mountain' in Xi'an which is 2,500 km from Ürümqi.

The nasal front vowels in the Xi'an forms for 'south' and 'mountain' might reflect the pre-Mandarin substratum dialect underlying Tangut

3382 1nin1 (transcription of Chinese 南 'south' in the Tangut translation of The Art of War)

3763 1shan2 'mountain' (borrowing from Chinese 山 'id.')

The -n of my Tangut transcriptions is not a consonant [n]; it indicates vowel nasalization.

Gong reconstructed the rhyme of 3382 as -ẽ. I think -in1 might have been something like [ə̃j], [ẽj], or [ɪ̃]. In any case, it was not [ĩ] like -in4. VOWEL HARMONY IN OLD CHINESE, PRE-TANGUT, QIANG, TIBETAN, AND XIONGNU

Back in 2002, I proposed that presyllabic vowels conditioned vowel 'warping' in Old Chinese:

low-vowel presyllable + high-vowel syllable > falling diphthong

e.g., *Cʌ-Ci > Cei

high-vowel presyllable + low-vowel syllable > rising diphthong

e.g., *Cɯ-Ca > Cɨa

Ten years later, I proposed a similar origin for Tangut grades:

low-vowel presyllable + high-vowel syllable > Grades I and II (depending on medial)

e.g., *Cʌ-Ci > Ci1 and *Cʌ-Cri > Ci2

high-vowel presyllable + low-vowel syllable > Grades III and IV (depending on initial)

e.g., *Cɯ-sha > sha3, *Cɯ-sa > sa4

In all of the above cases, harmony is from presyllable to syllable (= prefix to root in many cases) and only partial. The second vowel does not completely change to match the first.

However, in Yadu Qiang, a relative of both Tangut and Chinese, harmony is from root to affix and total according to Nate Sims:

a + = ɑ-hɑ 'one bunch' (backing of a)

a + = ɛ-pɛ 'one bowl' (raising of a)

(a 'one' is cognate to Tangut

5981 0a0 'one'.)

+ pʰu = tu-pʰu 'up-flee' = 'to flee upward'

ʁwɑkʰu + = ʁwɑkʰu-pu 'sarcasm-do' = 'to be sarcastic (raising and rounding of ə)

Did vowel harmony develop in opposite directions in different branches of Sino-Tibetan? Lhasa Tibetan even has harmony "from affixes to stems and vice versa" (DeLancey 2003: 271): e.g.,

Hgro gyi yinukijiː̃] 'go (conjunct future)' (stem vowel raising)

zhim-po [ɕimpu] 'delicious' (suffix vowel raising)

That harmony is relatively recent, as it is not reflected in the orthography based on Classical Tibetan.

I don't know how old Yadu Qiang harmony is.

My proposed Pre-Tangut harmony must predate the 11th century when Tangut was first written (i.e., when Tangut had already lost the presyllables that conditioned harmony).

My Old Chinese harmony is even older.

I used to think that vowel harmony might have spread from Old Chinese to core 'Altaic' (Turkic, Mongolic, Tungusic), but I somehow overlooked the obvious fact that my proposed Old Chinese harmony is nothing like 'Altaic' harmony which is also root-to-affix and total: e.g.,

Turkish adam-lar 'men', Türk-ler 'Turks'

Written Mongolian aqa-nar 'older brothers', degüü-ner 'younger brothers'

Manchu sagda-sa 'old men', gege-se 'older sisters'

Korean pad-a 'receive and ...', ŏb-ŏ 'carry and ...'

I have not seen any evidence for vowel harmony in Xiongnu. The Xiongnu title 'crown prince' was transcribed in Old Chinese as 護于 with a mixture of syllable types (A and B in Chinese terminology). (But could 'crown prince' have been a compound noun without vowel harmony? Its second half may have been shared with the Tangut title for 'supreme ruler', transcribed in Old Chinese as 單于 and interpreted by Vovin as 'north-ruler'.)

If I am right about Chinese and Tangut, then

- Chinese and Altaic vowel harmony are unrelated

- Altaic vowel harmony had nothing to do with Xiongnu

Perhaps Mongolic* got it from Turkic or vice versa when speakers of both were under Xiongnu rule

- Tangut vowel harmony had nothing to do with Tangut's neighbor Uyghur

*I use 'Mongolic' here in a broad sense to refer to the ancestor of all languages related to the Mongolic languages: e.g., Khitan, which is, strictly speaking, a Para-Mongolic language that is a sister to Mongolic proper.

Tangut fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2014 Amritavision