Last night I finished reading this biography of the mother of the professor who introduced me to Tangut twice - once on a plane to Japan in 1988 and again in a classroom in 1994. Many or even most of the posts on this blog would not exist if I had not met the son of Valentina Valerianovna Lyovina. I thank her for her indirect yet enormous impact on my life. The story of her life inspired me this week when I needed the strength to go on. And go on she did - from Russia to Yugoslavia, Canada, and ultimately America. In her honor I shall transcribe her name in the script that I studied thanks to her son:

5156 1va (transcription character)

5267 1lɨẽ (transcription character)

0648 2ti 'to remain'

3274 1na (surname character)

Three of those characters were used to transcribe the Sanskrit syllables va, ti, and na; the exception is the second which was used to transcribe Chinese l-syllables with front nasal vowels: 靈林凌陵菱䔖綾令伶零領連蓮廉鐮.

The second syllable could also be transcribed as

3421 2lɛ̃ (syllable of 2lɛ̃ 2lɛ̃ 'medium'?*) or 1661 1lɨĩ (transcription character).

I am not confident about the nasalization of the vowel of 3421.

1661 may be closest to the Russian and English pronunciation of unstressed -len-.

Tangut l- may have been velar [ɫ] as it usually could be followed by the Grade III medial -ɨ- but not the Grade IV -i-.

Tangut had no syllables ending in -n; foreign nasal-final syllables were transcribed with characters for Tangut syllables ending in nasal vowels.

*7.6.2:31: The Tangraphic Sea definition of 3421 is

'2lɛ̃ TOPIC 1140 1413 is' (what I gloss as 'is' is the Tangut equivalent of the Classical Chinese copula 也; I discussed both in parts 1 and 2 of "A Fami-*l-j-l Resemblance")

'5254 5254 is'

'5963 5963 is'

'not small big GENITIVE say': i.e., 'how one says neither small nor big'; the genitive indicates that 'not small big' was nominalized, so a very literal translation would be 'the saying of not small [or] big'

I have not glossed some of the words in the definition because their definitions are circular:

1140 is defined as 0669 1140 'short', 5254 5254, 3421, and 'a body that is small but not big'

1413 is defined as 5963 5963, 1140 1413, 5254 5254, and 3421 3421

5254 is defined as 5963 5963, 1140 1413, 3421 3421, and 'neither big nor small'

5963 is defined as 1140 1413, 5254, 4417 '?', and 'neither big nor small'

'Neither big nor small' implies that 3421 is 'medium', but 3421 and its definition 5254 5254 also define 1140 whose other glosses include 'short' and 'a body that is small but not big'. How can 3421 and 5254 5254 be 'small' and 'not small' at the same time?

Li Fanwen (2008: 554) translated 3421 as an adjective 'equal, even, moderate'. I can understand how he got 'moderate' from 'neither big nor small', but not 'equal' or 'even'.

Kychanov and Arakawa (2006: 499) translated 3421 as an adjective 'middle-sized' but 3421 3421 as an adverb 'equally, just right'. Is there any text with 3421 3421 as an adverb? All examples of 3421 and 3421 3421 that I have seen in Li Fanwen (2008) are only in dictionaries. DID THE TANGUT KNIT? (PART 2)

When I looked up 'knit' in the English index of Li Fanwen's 2008 Tangut dictionary last night, I only found

2958 2e 'knitted wool'

whose entry mentioned its near-synonym

5930 2kəu 'knitted wool, woollen blanket'

but later I found two characters that he defined as 'knit' in English on pages 247 and 734:


1481 1tʂhəõ 'to knit, weave' (defined in Chinese as 結 'to tie'; Kychanov and Arakawa 2006: 'tie, knot, clog, bind', путы 'fetters') =

top and bottom left of 1539 1tʂhwəẽ 'to tie, fasten' (semantic) +

bottom right of 2176 1tʂɨəʳ 'to tie' (only in dictionaries?; semantic)


4626 2dwa 'to knit' (defined in Chinese as 織 'to weave'; Kychanov and Arakawa 2006: 163: 'weave, knit') =

left of 4640 2dwa 'many, much' (only in dictionaries?; phonetic)

left of 0635 1niə (first syllable of 0635 1424 1niə 1thiu 'blood relations'; semantic: 'blood ties' > 'tie'?)

I doubt either verb meant 'to knit' for the reason I gave at the start of my last post. I assume 1481 is 'to tie' and 4626 is 'to weave'.

1481 1tʂhəõ 'to tie' and 1539 1tʂhwəẽ 'to tie' are obviously cognates and even combine to form a disyllabic verb

1481 1539 1tʂhəõ 1tʂhwəẽ '[to] tie, clog, shackles, fetters' (Kychanov and Arakawa 2006: 661)

The two words go back to *choN and *P-cheN; the labial prefix of the latter conditioned -w-. Did *P-cheN originate as the second half of a reduplicative compound *choN-*cheN which came to be an independent word *cheN? See Gong (2003: 612) for other examples of o-e reduplicative compounds.

4626 2dwa may go back to *P-da-H, and its *da may be the root of

0630 1la < *Cʌ-Ta 'to weave' and 2497 2la < *Cʌ-Ta-H 'to weave'

(*T may be *t, *th, or *d; voicing and aspiration are neutralized when obstruents lenited in intervocalic position.)

A d-word for 'weave' is in some Qiangic languages (e.g., Ersu dɛ, Lyuzu de, Xumi dyi; the last two were also glossed as 'knit'!), but not in rGyalrong (e.g., Casmi ka-tia, Tshobdun ka-ta, Japhug kɤ-taʁ, Somang ka-tak, Zbu kɐ-tɐχ < *-q; did *-q become zero in Tangut instead of conditioning vowel 'length'?).

7.5.2:03: Perhaps prenasalized forms are the 'missing link' between the d- and t-words for 'to weave': e.g.,


Namuyi ndæ (also glossed 'knit'!), Guiqiong nthɑ, Zhaba a-ntha 'weave a basket'


Daofu nthɑ (also glossed 'knit'!)

Could 4626 2dwa go back to *P-N-t(h)a-H? If so, then there is no need to reconstruct *d- in the Tangut root for 'to weave', and the la-words for 'to weave' may go back to *Cʌ-t(h)a(-H). Perhaps *Cʌ- was *Nʌ-, and the three words were originally very similar:

*Nʌ-t(h)a > *Nʌ-la > 0630 1la

*Nʌ-t(h)a-H > *Nʌ-la-H > 2497 2la

*P-Nʌ-t(h)a-H > *P-Nda-H > 2497 2dwa

*Nʌ- was completely lost after *t(h) lenited to *-l- following its vowel in the first two words, but it lost its vowel in the third word and fused with the root initial. I wonder how many other *t-roots have l- and d-derivatives in Tangut. DID THE TANGUT KNIT?

Although Li Fanwen's (2008: 484) dictionary defined Tangut

2958 2e and 5930 2kəu

as 'knitted wool', Kychanov and Arakawa (2006: 377) defined both as 'princess'**, and I doubt the Tangut knew what knitting was if knitting is only about a millennium old and initially spread westward from the Middle East. Perhaps the Tangut could have spoken about knitting by recycling a word for 'weave' such as


0630 1la 'to weave' = left of 0435 2kiʳw 'to weave' + center and right of 2374 2pʊ̣ 'to weave'

which has a rising tone variant


2497 2la 'to weave' = center and right of 2374 2pʊ̣ 'to weave' + right of 0630 1la 'to weave'

and which may be cognate to Old Chinese 織 *tək 'to weave' and tak-type words glossed as 'weave' and 'knit' (!) in other Sino-Tibetan languages.

The two verbs combine into

1la 2la 'to entwine, thread though; plait silk' (Kychanov and Arakawa 2006: 432).

The initial l- of 1la and 2la is a lenited *T-:

*Cʌ-Ta(-H) > 1la ~ 2la

The initial of the prefix *Cʌ- is unknown. The vowel of the prefix was low since it did not condition raising in the root before being lost.

The function of the suffix *-H is unknown.

The rhyme is problematic if I want to relate these la-words to OC *tək 'weave', Somang rGyalrong ka-tak 'to knit', etc.

- It would be simple if Old Chinese and pre-Tangut vowels usually matched, and OC corresponded to pre-Tangut *ə, but Tangut a is from pre-Tangut *a, not *ə.

- I expect pre-Tangut *-k to be reflected as vowel 'length' in Tangut: *-Vk > -VV.

Compare with Tangut

3752 3296 2miə 2niaa 'Tangut'

whose pre-Tangut ancestor was borrowed as Tibetan Mi-nyag (implying *mə-njak).

(I put 'length' in quotation marks because I no longer think Tangut had distinctive vowel length. I do not know what the real difference between 'short' rhyme 17 -a and the less frequent 'long' rhyme 22 -aa was.)

Sino-Tibetan t-k words in turn resemble Thai ถัก thak 'to knit' and Lao ຖັກ thak 'id.' which have no cognates outside southwestern Tai unless they are related to tak 'knot' in Po-ai, a northern Tai language (cf. how English knit and knot are cognates). Pittayaporn (2009: 89) did not reconstruct *th- in Proto-Tai (PT) and regarded th-words "as post-PT lexical innovations, either borrowings or forms derived after the establishment of the contrastive aspiration."

If thak 'to knit' is a borrowing from 織 OC *tək, its aspiration is unexpected. Is that aspiration a trace of a prefix in a southern dialect of Chinese? Moreover, its vowel is also unexpected. OC (later *ɨə) would have been borrowed into early Tai as or which would have become ɯ and o* in Thai and Lao, not a. Maybe thak is from a southern Late Old Chinese */tʰək/ [tʰʌq] with aspiration and a lowered vowel due to an earlier prefix:

*sʌ-tək > *sˁʌˁ-ˁtˁʌˁqˁ > *sˁtˁʌˁqˁ > */tʰək/ [tʰʌq]

7.4.0:28: If there ever was such a form, it does not have any modern reflexes in southern Chinese languages and Sino-Vietnamese. Taiwanese tsit, Cantonese zik, and Sino-Vietnamese chức all point to Middle Chinese *tɕɨk from Old Chinese *tək. I cannot find any aspirated forms of 織 in any Chinese variety at 小學堂.

*7.4.0:37: In Pittayaporn (2009), always rounded to o before *-k in Thai except in *lɤk 'deep' which became Thai ลึก lɯk instead of ลก *lok. I do not know why Thai ลึก was not marked with "-v" for vocalic irregularity.

**7.4.1:48: Kychanov and Arakawa (2006: 377-378) also defined the disyllabic word

2958 5930 2e 2kəu

as 'princess', whereas Li Fanwen (2008: 484) glossed that as 綫毯 'thread blanket' and 毛毯 'woolen blanket', presumably because each character had the note

1743 5760 2lɨu 1tshʊ  'thread rough***' = 'rough thread'

in the D version of Homophones.

Li glossed 2958 by itself as 毛綫 'wool' in Chinese and 5930 by itself as 毯 'blanket' in Chinese and as 'woollen blanket' in addition to 'knitting wool' in English. (His dictionary has bilingual glosses for individual characters but only Chinese glosses for polysyllabic words and phrases.)

As far as I know, neither 2958 nor 5930 occur by themselves or as a sequence outside dictionaries, so 2958 5930 may be a disyllabic 'ritual language' word, and 2958 and 5930 may not be monosyllabic words.

I do not know

- why Li defined 2958 5930 as something other than a redundant compound of 'rough thread', and how he was able to define 2958 and 5930 slightly differently as monosyllabic words.

- why Kychanov and Arakawa's definitions for 2958, 5930, and 2958 5930 are so different from Li Fanwen's.

***5760 1tshʊ  'rough' was borrowed from Chinese 粗 'id.' A FAMI-*L-J-L RESEMBLANCE? (PART 2)

I forgot to mention in my previous post that I doubt 隹/維/惟 ever had *l- in Old Chinese. 維/惟 (but not 隹) definitely had some sort of palatal segment in Middle Chinese dialects, judging from Sino-Korean yu (borrowed from northeastern MC) and Sino-Vietnamese duy < *jwi (borrowed from southeastern MC). Although MC *j- generally comes from OC *l-, it is not necessary to reconstruct *l- in this case, as a simple OC *wi would also have become MC *jwi or *ɥi, etc. Moreover, the cluster *lw- would be sui generis in my reconstruction and nonexistent in Baxter, Sagart, and Schuessler's latest reconstructions. I just wanted to see how far I could take my *l-j-l hypothesis and fit 隹/維/惟 into it. Schuessler (1987: 632) once reconstructed 隹/維/惟 as *ljuəj but has since opted for *wi (see p. 37 of his 2009 book). *wi matches Thurgood's (1982) Proto-Sino-Tibetan copula *wəj and Matisoff's (1985) Proto-Tibeto-Burman copula *waj. (I have replaced Thurgood and Matisoff's *y with *j for ease of comparison.

Lowes (2006) gathered copulas and similar verbs from 71 Sino-Tibetan languages and mapped them. Some have l- and w-; others don't. Although Gong's (2003) description of Tangut is in her bibliography, she didn't list the Tangut copula

0508 2ŋwʊ

which resembles Cogtse rGyalrong ŋos in her "Remaining" category. See #2156 'be, is' in Nagano and Prins' database for cognates in other rGyalrong varieties.

Lacking familiarity with the many languages in her paper, I will reserve further comments on it and focus instead on Chinese and Tangut. Could the apparent vowel alternations in the following forms be remnants of earlier verb paradigms? Chinese glosses are brief; see Schuessler (1987) for fuller glosses. Tangut glosses are based on Gong (2003) except for 1lɨə which is based on Sofronov (1968 I: 262).

Root Language Rhyme type -i < -(ə)j?
< -j(ə)l?
-e- -a
*w- Old Chinese 隹/維/惟
*wi < *wəj? 'to be' (impersonal)

*waj 'to act as, do'

*wets < *waj-t-s or *wej-t-s 'should'

*wəʔ 'there is, to have'
2vɨe < *Cɯ-wajH or *Cɯ-we(j)H 'there is'
*l-j-l? Old Chinese
*ʔlil < *ʔljəl? 'to be'

(objective; later copula)
*ləʔ 'should, indeed' (later perfective)
1lɨə < *lə  (intensive particle)
*m- Old Chinese
*məj < *-l? 'to not be'

*Cɯ-maj < *-l? 'there is no'

*met < *maj-t or *mej-t 'to not have'

*mə 'should not'

*Cɯ-ma 'there is no'
1mi 'not' (before nonauxiliary verbs)

1mie < *Cɯ-maj or *Cɯ-me(j) 'there is no', 2mie < *Cɯ-majH or *Cɯ-me(j)H 'not yet'

1mɨə < *mə 'not' (before auxiliary verbs)

2niaa < *mjaCH? 'no!', 'don't ...!'

There are even more Old Chinese negatives, but they either start with *p- (e.g., 不 *pə) or have rhymes that are variations of those in the table above (e.g., 莫 *mak). Notes on the five rhyme types that I did include:

1. -j(ə)j/l

隹/維/惟 OC *wi could be from *wəj or be the zero grade of *w-j. Although OC *-j may be partly from *-l, the *-j of *wəj could not be from *-l if it is descended from Thurgood's Proto-Sino-Tibetan *wəj.

It is interesting that Lowes (2006) did not list any probable modern reflexes of PST *wəj, though she devoted a section of her paper to the proto-form. STEDT only lists reflexes in Nungic and Loloish.

微 OC *məj may be the schwa grade of *m-j or *m-l.

Tangut 1mi may be the zero grade of *m-j or *m-l (if *l merged with *j in pre-Tangut) or from *Ci-ma after 'brightening'. See 5 for other ma-negatives. I would prefer to reconstruct a high-frequency function word as monosyllabic.

2. -aj/l

爲 OC *waj is the *a-grade of *w-j. See above on why I don't derive its *-j from *-l.

Tangut 2vɨe may also be from *w-j. See 4 below for a homophonous optative prefix absent from the table above.

也 OC *ljalʔ is an *a-grade form of *l-j-l.

靡 OC *maj and Tangut 1mie and 2mie may be *a-grade forms of *m-j or *m-l (if *l merged with *j in pre-Tangut).

3. -e

叀 OC *wets may be an *a-grade or *e-grade form of *w-j.

蔑 OC *met may be an *a-grade or *e-grade form of *m-j or *m-l.


These forms could have lost their root-final consonants because they were originally unstressed.

The perfective and optative prefixes

2vɨə- < *wəH and 2ve- < *weH or *wə-j-H?

are not cognate to 有 OC *wəʔ and 叀 OC *wets since they are originally directional prefixes meaning 'there, outside'.

5. -a

Earlier reconstructions of 無 OC *Cɯ-ma have a medial *-j- corresponding to the *-j- of my pre-Tangut *mjaCH: e.g., Gong Hwang-cherng's OC *mjag. However, recent reconstructions have abandoned OC *-j-, as it usually corresponded to nothing in other languages: e.g., Written Tibetan ma 'not' and Written Burmese ma 'not' (even though mya is possible in both Written Tibetan and Written Burmese).

I would normally derive Tangut 2niaa from *Cɯ-naH, but I wanted to see if I could make it part of the m-family. If Tangut ni- is partly from *mj-, perhaps all miV are from Cɯ-mV:

Pre-Tangut Tangut
*Cɯ-mV miV
*(Cɯ-)mjV niV

Are there any other Tangut niV-words with m-cognates? The closest thing that comes to mind is

1nɨaa < *Cɯ-naC or *mjaC 'black': cf. 黑 OC *sʌ-mək 'id.'


1. it has -ɨ- and -a-, not -i- and -ə-. (Medial -ɨ- after a dental is unusual and should be investigated.)

2. there is no Chinese-internal reason for reconstructing *-j- in OC *sʌ-mək (which is not to say that *-j- is impossible; I suspect it was lost after the emphatic consonants conditioned by *Cʌ-prefixes).

3. it is a better phonetic match for Written Tibetan nag 'black'.

This next word also has phonetic problems:

2nieʳ < *Cɯ-nerH or *rɯ-ne(n)H or *mjerH or *rɯ-mje(n)H 'face': cf. 面 OC *mens 'id.'

Although it at least has -i- unlike 'black', once again, there is no Chinese-internal reason for reconstructing *-j-, and it is not certain Tangut ever had a final nasal in this word since *r-...-e and *r-...-eN merged as -eʳ. A FAMI-*L-J-L RESEMBLANCE?

Last night when I mentioned the Chinese third person pronoun 伊 (now obsolete in standard Mandarin, though surviving in Taiwanese), I realized that a homophonous copula written with the same character* might share a root *l-j-l with the copula 也. Tonight I tried to see how far I could extend the *l-j-l family. Possible affixes of unknown function are in red (indicating that they cast doubt on this hypothesis - it is too easy to 'relate' unrelated words by dismissing all 'extra' segments and syllables as 'affixes').

Sinograph My Old Chinese Zhou Chinese gloss from Schuessler 1987
*ʔlil personal equational copula
*ljajʔ < *ljalʔ? marks statement as objective fact (later copula)
*ləʔ? should, indeed (later perfective)
隹/維/惟 *lwi < *lwil? < *P-lil? impersonal equational copula

The phonetic of 伊 Md yi < *ʔlil is 尹 Md yin < *lwirʔ (< *P-lirʔ) 'be straight, straighten, administrator (i.e., one who straightens)'. I reconstruct *-l and *-r for characters sharing a phonetic that are later read with -i/j and -n.

*ʔlil is a 'zero grade' form of the root *√l-j-l. The medial *-j- became *-i- since it was not followed by a vowel.

*ljalʔ is an *a-grade form of the root *√l-j-l. I am not certain that the consonant after *a was originally *l instead of *j. It is difficult to believe that a pause marker (another function of 也 according to Schuessler 1987) could have five segments instead of being a simple V or CV syllable. Maybe it had a shorter reading in that function - a reading like that of 矣.

The reconstruction of 矣 is elusive, as it belongs to a phonetic series with various initials without an obvious common denominator. The simple reconstruction *ləʔ could be an unstressed reduction of some pre-Chinese form of the verb *√l-j-l. But I see no reason why a deontic form would be more prone to reduction than a realis form.

The Tangut translation equivalent of both 也 and 矣 is

5285 1lɨə < *lə

and it is tempting to see a direct connection between pre-Tangut *lə and 矣 *ləʔ.

Maybe 矣 *ləʔ has nothing to do with *√l-j-l and is simply a variant spelling of 已 *ləʔ 'to finish' (and by extension, 'already'). 矣 and 已 might have also phonetically differed in some way that can no longer be reconstructed.

隹/維/惟 *P-lil is another 'zero grade' form of the root *√l-j-l. Although the shift of *PC- to *Cw- is phonetically plausible, there are not enough phonetic and word family alternations to fully justify reconstructing a prefix *P-. I am much more confident about reconstructing *P- in pre-Tangut to account for sets of words such as

0618 1tsia < *Cɯ-tsa (*Kɯ-tsa?) 'hot'

(The prefix *Kɯ- may have been lost after conditioning vowel breaking and before it could condition aspiration in the root initial *ts-. Tibetan tsha < *tsa preserves the bare root.)

1829 1tsha < *Kɯ-tsa 'hot'

1825 1tshwia < *P-Kɯ-tsa 'to heat'

See Gong (2002: 45-46) for more examples in Tangut. I interpret his Chinese example as involving vowel alternation rather than zero ~ *-w-alternation:

熱 Middle Chinese *ɲiet < Old Chinese *Cɯ-ŋet 'hot'

爇 Middle Chinese *ɲwiet < Old Chinese *Cɯ-ŋot 'to burn'

Then again, maybe there never was an *o in 'to burn':

*Pɯ-ŋet > *Pɯ-ŋiet > *Pŋiet > *ŋwiet > *ɲwiet

But I hesitate to reconstruct a Proto-Sino-Tibetan *P-causative prefix, much less relate it to the Proto-Austronesian causative prefix *pa- and/or Anderson's (2004: 162) Proto-Austroasiatic causative prefix *’B-. Lookalikes abound  within as well as among languages: e.g., Vietnamese also has a l-copula (< Proto-Vietic *la 'to work'?) that is unrelated to any of the Chinese forms above. In any case, if 隹/維/惟 ever had a *P-prefix, it could not be causative, since 隹/維/惟 is 'be', not 'cause to be' (unless 'cause to be' was downgraded to 'be').

*The semantics of 伊 'he/she/it ~ is' are reminiscent of those of modern standard Mandarin 是 'is' which originated as 'this'. I thought 伊 might have undergone a similar shift from pronoun to verb, but Schuessler (1987: 742) listed the verbal meaning before the pronominal meaning (implying the verbal meaning is older?), and my *l-j-j hypothesis requires 伊 to have been a verb first. However, I cannot think of a case in which a verb became a pronoun, though I can make up a scenario of reinterpretation:

1. 'X is Y'

2. 'is Y' (X is dropped as an assumed subject)

3. 'he's Y' ('is' comes to refer to the subject)

4. 'he' ('he's' is used before verbs as well as nouns, becoming a pronoun)

If that occurred, then Taiwanese 伊是 i si 'he is' is etymologically 'is' + 'this'!

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