In "Two Strata of Nasals in Tangut?" I proposed that

- original root-final nasals were lost in Tangut

*-VN > -V

- Tangut nasal vowels are of foreign origin

non-T -VN ~ -Ṽ > T -Ṽ

or reflect nasal initials of second syllables of compounds:

*-V NV > *-VN > -Ṽ

This hypothesis predicts that no native root words should have nasal vowels. However, a counterexample quickly came to mind last night:

2diẹ̃ 'cloud' (which has also been reconstructed with an oral vowel*)

This word should go back to a compound *Sɯ-de-NVH** and its cognates in other languages should be open syllables like

Taoping Qiang χde³³

Xinlong Queyu ɕtie⁵⁵ 

Zhaba ʂtei

Daofu zdo

(All Qiangic except for Daofu which is rGyalrongic.)

However, other cognates also have nasal vowels or -m:


Jiulong Pumi d³⁵

Lanping Pumi zdĩ⁵⁵

Qinghua Pumi s⁵⁵

Taoba Pumi zə⁵⁵ r⁵⁵  (< *zəd?)

Mawo Qiang zdɑm ~ zdɤm

Shixing tĩ⁵⁵ rõ³³ ~ tɕi⁵⁵ rɑ̃³³  (< *tĩdṼ or *tĩd?; unsure if the first or second half is cognate)

rGyalrongic (the last three forms are from Jacques 2004: 245 who reconstructed Proto-rGyalrong *-im for the rhyme of 'cloud' on p. 251)

Maerkang rGyalrong zdᴇm

Caodeng rGyalrong zndimʔ ~ zndəmʔ

Benzhen rGyalrong zdiɐmʔ 

Japhug rGyalrong zdɯm

Somang rGyalrong zdém

Zbu rGyalrong zdím

Written Burmese တိမ် <tim>

Middle Chinese 曇 *dəm 'cloudy' (no Early Old Chinese attestations; could this be a borrowing from a substratum language?)

If Tangut is Qiangic, it is plausible that a Proto-Qiangic compound is the source of the Qiangic forms with nasal vowels and nasals. Perhaps

Longxi Qiang dá mù̥  'cloud, fog'

Mianchi Qiang dá mò 

preserve the second half of that compound (which STEDT traces back to Proto-Tibeto-Burman *r-muːk 'fog(gy), dark, dull').

However, it is not plausible that non-Qiangic languages (Burmese and Chinese) would also have cognates ending in -m unless one posits a compound at the Proto-Sino-Tibetan (!) level as a source of all those words. Would a couple of Qiang varietieshappen to preserve a compound that was lost elsewhere?

My guess is that this is an unusual case of nasal restoration. The original root for 'cloud' ended in an *-m that was lost in early Qiangic but preserved in rGyalrongic and Burmese. Later, Qiangic combined an open-syllable word for 'cloud' with an *m-initial adjective 'dark' that was

- preserved in Longxi and Mianchi Qiang

- reduced to -m in Mawo Qiang

- reflected in nasalization in Tangut, Pumi, and Shixing

The fact that Tangut had a nasal vowel in 2diẹ̃ 'cloud' almost a thousand years ago implies that a compound word for 'cloud' is very old.

Taoping Qiang and other Qiangic languages preserve the bare root of 'cloud' as a monosyllable ending in an oral vowel.

Another possible source of Tangut/Qiangic nasals and nasality in 'cloud' is an *m-word for 'sky, clouds, fog, rain' that is the ancestor of words for 'sky' like


1mə 'sky'


Jiulong Pumi my

Lanping Pumi mɤ⁵⁵ 

Xinlong Queyu mu⁵⁵

Yajiang Queyu mɯ⁵⁵


rGyalrongic Ergong zdo 'cloud' is a compound of 'cloud' and 'rain'. Its m truly corresponds to the -m of Mawo Qiang and the nasalization in Tangut, etc. whereas rGyalrong and Burmese -m actually corresponds to zero in Qiangic and Tangut.

The Chinese word could either preserve the original -m or be a loan from a Qiangic language with a compound for 'cloud'. That compound could have become a monosyllable or was reduced to a monosyllable in Chinese:

Q *dəmV > Chn *dəm

But are there any (other) Qiangic loans in Chinese? I have never heard of any. Has anyone looked for them? As far as I know, there are no ancient Chinese loans into Qiangic with the exception of Tangut (if it is Qiangic). I would expect loans in the other direction to be even rarer.

*Here is a list of all reconstructions for 'cloud' (or its rhyme if its initial is not available)

Kychanov and Sofronov 1963: 2-ɪ

Nishida 1964 and 1966: 2ne (though the rhyme is reconstructed on p. 59 of vol. 1 as 2-ɛ̣̃!)

Hashimoto 1965: 2-jɛ

Sofronov 1968: 2nɪ̣

Huang 1983: 2-ïɛ̃ (I chose the reconstruction closest to the others)

Li Fanwen 1986: 2nǐɛ̣̃ ̣(using the rhyme from the table on p. 189; the entry for the word itself on p. 290 has two other rhymes: 2nǐɒ and 2nǐɛ!)

Gong 1997: 2djɨ̣j

Arakawa 1999: 2-yenq (the same rhyme as mine despite a different notation)

Sofronov 2012: 2-ɛ (assuming he reconstructed it as Grade IV)

This site: 2diẹ̃

The only non-Tangut transcription of 'cloud' that I know of is late 12th century northwestern Chinese 寧 which Gong would read as *ndjij and I would read as *ndiẽ.

Tibetan transcriptions of its rhyme (2.55) are -e (twice), -eH (five times), and -eng (twice) (Tai 2008: 222). I opt for a nasal vowel because an oral vowel cannot account for Tibetan -ng.

**The prefix *Sɯ- conditioned the tense vowel (indicated with a subscript dot) and the breaking of *e to *ie:

*Sɯ-de- > *Sɯ-die- > *S-die- > *zdie- > *ddie- > *ddiẹ- (tension spreading into vowel) > diẹ-

The final *-H conditioned the second (i.e., 'rising') tone. It is also possible that the *-H was at the end of the root syllable:


My reconstruction of Tangut has almost no final consonants. All Tangut syllables end in vowels or -w (and one could treat -w as the final element of diphthongs ending in -u). Nonetheless, it is certain that Pre-Tangut had many final consonants like its relatives Written Tibetan, Written Burmese, and Chinese.

I used to think that Pre-Tangut final nasals generally conditioned Tangut nasal vowels:

Pre-Tangut *-VN > Tangut -Ṽ

There are no native Tangut words ending in nasalized -u-type vowels, so Pre-Tangut *-uN probably merged with Tangut oral -u vowels or nasal -õ-type vowels.

In some cases, nasalized tense and retroflex vowels lost their nasalization: e.g.,

Pre-Tangut *k-som > *s-som > *sọ̃ > 1sọ 'three'

Pre-Tangut *Cɯ-raŋ > *riẽʳ > 1rieʳ̣ 'horse'

One might predict that the Tangut cognate of Old Chinese 羊 *jaŋ 'sheep' should be ʔiẽ with a nasal vowel since Pre-Tangut *jaŋ lacks

- an *u that might have denasalized

- a preinitial *s- that conditioned tension which generally cannot coexist with nasalization (two exceptions are rhyme 65 -iẹ̃ and rhyme 76 -ɛ̣̃)

- a preinitial, initial, or final *r that conditioned retroflexion which generally cannot coexist with nasalization (two exceptions are rhyme 97-õʳ and rhyme 98 -iõʳ)

However, the actual word for 'sheep' is

2ʔie 'sheep'

with an oral vowel (and a second tone implying a final glottal *-H corresponding to nothing in Chinese). How can I account for its lack of nasalization?

Perhaps there are two strata of nasals in Tangut:

Primary nasals were lost without a trace: e.g., in 'three', 'horse', and 'sheep'.

Secondary nasals arose from

relatively recent loans from Chinese: e.g.,

2ʂɨẽ < Tangut period northwestern Chinese 聖 *ʂɨẽ 'sage'

compounds: *CV NV > *CVN (cf. the origin of secondary final consonants in Qiang*)

These new nasals also disappeared, but unlike primary nasals, they conditioned the nasalization of vowels. This nasalization may have been lost if the vowel was (a) *u, (b) tense, or (c) retroflex.

If this hypothesis is correct,

- all native root words should have oral vowels

- early borrowings of Chinese nasal-final words (if any) should have oral vowels

- all native words with nasalized vowels should be from compounds

I just came up with this idea and have yet to test it.

*LaPolla and Huang (1996: 23):

After the loss of the original [Qiang] finals, and the destressing of second syllables in two syllable compounds, the two syllables merged, with the initial of the original second syllable, or a reduced form of it, becoming the final of the original initial syllable.

11.24.10:30: Ronghong Qiang is similar to Tangut:

- Native words that once ended in nasals are now open syllables: e.g.,

*k-sum > χsə 'three' (cf. Tangut 1sọ 'id.')

*C-raŋ > 'horse' (cf. Tangut 1rieʳ̣ 'id.')

- Early** borrowings of Chinese nasal-final words have open syllables: e.g.,

*ŋ > RQ 'wolf'

*phen 'basin, pot, tub' > RQ pe 'tray' (why p- instead of ph-?)

扁豆 *piantəu > RQ petəu 'hyacinth bean'

(later borrowed again with a nasal as piantəu 'string bean')

- Verbs plus mi 'person' (cognate to Tangut 1mi 'Tangut') fused into words ending in -m: e.g.,

*ʁua-mi 'help-person' > ʁuam 'servant'

Here's an interesting case of a disyllabic Chinese word that was compressed into a Ronghong Qiang monosyllable:

蘿蔔 ?*luopɑu > RQ lup 'turnip'

Oddly a noncompressed variant of this word has a medial nasal in χolumpɑu < 胡蘿蔔 ?*χʊluopɑu 'carrot' (lit. 'foreign turnip').

The standard Mandarin word is huluobo [xuluopuo]. I presume that the local Mandarin pronunciation is closer to χolumpɑu, albeit without the nasal. Ronghong Qiang has both velar x and uvular χ, so speakers are not compelled to borrow Chinese velar [x] as uvular χ. The first vowel might be ʊ rather than o in local Mandarin since 胡 generally has an u-vowel in Mandarin dialects and Ronghong Qiang has no ʊ.

**11.24.10:36: These borrowings are not very old since they reflect sound changes from the last two millennia:

Early Old Chinese *r- > Late Old Chinese *l-

Middle Chinese *b- > Early Mandarin ph-

Early Mandarin *-un > local Mandarin -en

Early Old Chinese *-o > Late Old Chinese *-əu

I have not yet seen any archaic Chinese features. I think it's possible that the old and new strata of Chinese loanwords may only be a century or even just decades apart. DOUB-LƗỊ BLESSED

Today is Thanksgiving in the US. Last night I tried to come up with a Tangut translation of 'Thanksgiving'. That was tough since I don't know of a Tangut word for 'thank'. (If I only had a reverse index to Kychanov and Arakawa's dictionary! And/or Nishida's!) Here's the best I could do:

2lɨi ̣1lɨəə 1tseʳw 'kindness-repay(ing) season'

The first character is analyzed as a compound of two homophones in the Combined Homophones and Tangraphic Sea:


2lɨi ̣'kindness' = left of 2lɨi ̣'wide, broad' + right of 2lɨi ̣'lamb'

How many other Tangut characters have this structure? Why write 'kindness' with parts of two homophones instead of parts of semantically relevant characters?

Perhaps this analysis is backwards. 'Kindness' might be phonetic in 'wide' and 'lamb'. The Combined Homophones and Tangraphic Sea analyzed 'wide' as a phonosemantic compound:


2lɨi ̣'wide, broad' = left of 2lɨi ̣'kindness' (phonetic) + 1zie 'wide, vast, extensive' (semantic)

However, 'kindness' does not appear in the Combined Homophones and Tangraphic Sea analysis of 'lamb' as a semantic (?) compound:


2lɨi ̣'lamb' = 1zi 'man, male' (why?) + 2ʔie 'sheep' (semantic)

Does 2lɨi ̣'lamb' really refer to young male sheep?

Or could 1zi be phonetic if it was a lateral fricative [ɮ]? The Tangut initial that Gong reconstructed as z- is grouped with liquids in Homophones, implying that it was l- or r-like. Are there any other examples of z-phonetics in l(h)-graphs and vice versa? Or of ʐ-phonetics in r-graphs and vice versa?

Although I thought that the true analysis of 'lamb' was


2lɨi ̣'lamb' = 2lɨi ̣'kindness' (phonetic) + 2ʔie 'sheep' (semantic)

perhaps it was from 'man' + 'sheep' and it forms the beginning of a phonetic chain:


2lɨi ̣'lamb' > 2lɨi ̣'kindness' > 2lɨi ̣'wide, broad'

In this scenario, the graphs for 'lamb' and 'wide' share a common phonetic source ('kindness') even though they don't share any components!

Tangut fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2012 Amritavision