I was uncomfortable about claiming that Tangut lew 1.43 'one' had a dental stop root because I couldn't find an internal cognate with initial *t(h)- or *d-. But it turns out that there are two words that might fit those criteria:

TT2507 ONE/ALONE tjịj 1.61 < pre-Tangut *C-tjij (?*k-tjej)

transcribed in Tibetan as gteH (Nevsky 1926: 78, 1960 II: 653)

the tense vowel presumably reflects an earlier cluster

TT2704 ONE tjɨ 1.30

cf. TT2422 ONE ljɨ 1.29 - Gong reconstructs 1.29 and 1.30 identically

Grinstead also glossed TT2704 as IF, but I don't think it can mean IF by itself, as it seems to form a disyllabic word 'if' with TT2071 IF tjij 1.36.  Could the first syllable of tjɨ tjij 'if' be partially reduplicated and written phonetically with a tangraph for an unrelated homophone?  Or can 'if' be semantically derived from 'one'?

Nevsky (1960 II: 653) compared TT2507 to 獨 Mandarin du 'alone' (< Old Chinese *dok)*.  However

TT2507 tjịj 1.61

TT2704 tjɨ 1.30

TT2422 ljɨ 1.29

lack the final -w that should correspond to Old Chinese *-k. This absence of a -w also complicates a cognate relationship between them and lew 1.43. One could argue that the -w-less Tangut words for 'one' preserved an originally open syllable root, whereas

TT0100 lew 1.43 < *CV-tek

Japhug ptɯɣ 'one' (in sqa ptɯɣ 'eleven') < ?*p-tik

Written Tibetan gcig

Written Burmese tac

隻 Old Chinese*tek 'single'

reflect a suffix *-k.  But I know of no non-Tangut evidence for reconstructing an open syllable word for 'one'.  Could pre-Tangut have lost the *-k in some forms of 'one'?  As a high-frequency word, it would be particularly susceptible to phonetic erosion.

*I've wondered if 獨 Old Chinese *dok  should be reconstructed as *N-tok, sharing a root *t-k with 隻 *tek 'single'.  Since I hypothesize that nonhigh OC vowels may have conditioned emphasis, perhaps 'single' was *Cɯ-tek with a high-vowelled presyllable that 'deemphasized' the root:

*Cɯ-tek*Cɯ-tek*tek ONE MORE TIME: THE 'ONE' AND 'ONLY'

Last night, I wrote,

So did lew 1.43 ['one'] come from an pre-Tangut *CV-tek?

I forgot about the Tibetan transcriptions of 'one'

gliH, gli, kli

which don't match the Tangut period northwestern Chinese transcription 婁 ?*ləw.

For 11 years, I've struggled to reconcile these transcriptions. I was under the assumption that they represented the same Tangut pronunciation. Now I think that they may represent different dialects:

Stage 1Stage 2Stage 3
Pre-Tangut *kV-tek (cf. Old Chinese 隻 *tek 'single', Written Tibetan gcig 'one')*kV-lek (root initial lenition after presyllabic prefix)Dialect A:*ɣliɣ (transcribed as gli[H]; cf. how Proto-rGyalrong *kl- became Japhung ɣl-)
Dialect B: *kliɣ and/or *kliʔ (transcribed as kli)
Dialect C: *lew (phonetically [ləɰ]?; transcribed as 婁 ?*ləw)

I now wonder if the lateral initial of 'one' was primary rather than secondary.

Nevsky (1926: 1) compares 'one' which he reconstructed as *li or *δi to forms in a language he identified as "Čuṅ-kia-tsï":

leao, dleo, leou

deou, leu-deou-la (trisyllabic!?)

I assume the name of the language would now be written in Pinyin as Zhongjiazi. Does anyone know what language this is? I presume that these words are unrelated lookalikes.

lew 1.43 'one' has two spellngs in Tangut:

TT0100 (the numeral) and TT3603 ('only')

Nevsky compared TT0100 to

TT2422 ljɨ 1.29

transcribed in Tibetan as gliH, gli, liH, li

the PERSON on the right is derived from the left of TT3317 ONLY

which he translated as 'one, alone, only' (Nevsky 1926: 61, 1960 II: 398). TT2422 is also an equational sentence final particle like Old Chinese 也 *lajʔ 'be'. For a semantic shift between 'be' and 'only', cf. Old Chinese 唯 *t-wi 'be' > 'only' (but not 'one'!).

If ljɨ 1.29 is cognate to Old Chinese 也 *lajʔ and TT0100/TT3317, then 'one' had a pre-Tangut initial *l- and is not cognate to Sino-Tibetan tik-words for 'one'. The -w of lew might reflect a pre-Tangut *-k suffix: *le-k. However, I am uncomfortable about proposing such a suffix unless I can find other examples of X ~ Xw (< *X-k) alternation. Moreover, the vowels in ljɨ 1.29 and lew 1.43 don't match.* So perhaps these are two unrelated words with similar semantics.

*But the vowel of ljɨ 1.29 does match the vowel of

TT4601 ONE gjɨ 2.28

whose initial may reflect the same velar prefix as the k/g-transcriptions of lew 1.43 'one'.

Perhaps TT4601 has ljɨ 1.29 as its root:

*N-k-ljɨ-H > *ŋgljɨ > gjɨ

Transcriptions indicate that Tangut g- may have been prenasalized [ŋg] or at least of partial prenasalized origin. A nasal prefix could have prevented *gl- from becoming *ɣl- > l-.

The final -H (an unknown glottal) accounts for the rising tone. THE LEW-TEK SOLUTION

Right before I finished writing "Use It or Lose It?", I realized that Tangut lew 1.43 'one' not only wasn't an innovation but also could help me fill a hole in my lenition theory.

I've theorized that Tangut had a set of lenited or 'softened' initials originating from earlier stops preceded by presyllables:

*CV-velars > ɣ-

*CV-labials > w-

I wasn't able to figure out whatever happened to acute initials preceded by presyllables. I did consider the possibilty that the lenited forms didn't last long and reverted to stops: e.g.,

*CV-t(h)- > *CV-θ- > *θ- > th-

*CV-d- > *CV-ð- > *ð- > d-

But this hypothesis would predict alternations between t- and th- (< *CV-t-) which Gong (1988) did not find.

Then it occurred to me that Tangut dental stops might have lenited into liquids: e.g.,

*CV-t(h)/d- > *CV-l- > l-

If that were the case, then Tangut lenition would look very much like Korean intervocalic lenition:

Type of stop





Earlier Korean

Middle Korean













w- ?[v-]



Tangut has a large number of l-syllables. About 11% of all tangraphs (689/~6,000) were pronounced with initial l(h)-. For comparison, only 7% (41/579) of Xinhua zidian (1971 ed.) is devoted to sinographs with initial l- in Mandarin. Could some of this 11% consist of secondary liquids from earlier dental stops? Could lh- be from *CV-th-?

My hypothesis predicts that there should be

- word families with dental stop and lateral-initial members

- external dental stop-initial cognates for lateral-initial Tangut words

Sino-Tibetan languages have two types of words for 'one': it-words and tik-words. Since

*CV-t- would become Tangut l-

final *-k became -w in Tangut (Gong 1995: 55).

a Tangut word for 'one' would look like l-w - which matches the consonants of lew 1.43 'one'!

So did lew 1.43 come from an pre-Tangut *CV-tek? I can't be sure until I find many other l(h)-words with external t-cognates. The only two I could quickly find were

Tangut ljij 2.33 < ?*CV-taH 'see' : Written Tibetan lta, Old Chinese 睹 *t

Tangut lu 2.1 < ?*CV-tuk-H (?*-k-s) 'burn' : Written Tibetan dugs-pa 'kindle', Written Burmese tauk 'to blaze', Old Chinese 燭 *tok 'torch'

(I use -H to symbolize a lost pre-Tangut laryngeal [< *-ʔ, *-s] that was the source of the Tangut rising tone.)

More examples are needed to establish this correspondence. USE IT OR LOSE IT?

Although this idea may have intuitive appeal, I'm not sold on it:

We propose that the frequency with which specific words are used in everyday language exerts a general and law-like influence on their rates of evolution. Our findings are consistent with social models of word change that emphasize the role of selection, and suggest that owing to the ways that humans use language, some words will evolve slowly and others rapidly across all languages.

I object to the use of the term 'evolution'. When I first read that abstract, I wondered if the authors meant that high-frequency words underwent sound changes more quickly than other words. But they seem to be using 'evolution' to refer to lexical replacement.

The authors only used Indo-European data. Can the same results can be replicated with data from other families around the world?

Furthermore, there are cases when words of roughly equal frequency are replaced, whereas (seemingly) infrequent words are stable. The same qha/kha word for 'bitter' is in Sinitic, Tibetan, Burmese, Qiang, Tangut, etc. Although 'bitter' is not exactly marginal, it's presumably less frequent than the numerals for 'two' to 'ten' which have been been replaced by Chinese loanwords in Thai.

The original Sino-Tibetan word for 'seven' has been replaced in Tibetan and possibly also in Tangut:


Old Chinese *s-hnit

Written Burmese khunac < *nit

gDong-brgyad rGyalrong kɯ ɕnɯs

but Written Tibetan bdun!

and why does Tangut have ɕjạ 1.64, with -a vocalism - low vowels raise in Tangut, but the reverse is unknown to me. (Gong [1995: 72] does not list ɕjạ 1.64 as cognate to the OC and WB words for 'seven'.) Is -a from a suffix? The initial does look like Qiang forms:

Ronghong ɕtɕə (LaPolla and Huang [2003: 383] derived this from Proto-Tibeto-Burman *s-nis)

Mawo ɕiŋ (the nasal must be from a suffix)

Taoping stə

'One' is surely more frequent than 'seven', yet Tangut and Qiang have their own word a instead of a tik or it-word. Moreover, Tangut has a word lew 1.43 which has no cognates that I know of.

Next: Maybe Tangut lew 1.43 isn't an isolate after all.

ADDENDUM: Pronouns are yet another high-frequency lexical category. Yet they are highly prone to borrowing (i.e., replacement) in East Asia

due in particular to "the intensive pursuit of politeness" which characterizes speaker interaction in East Asian societies (Schmidt 1906) ... Chinese personal 1sg. and 2sg. pronouns have partly or completely replaced the native pronouns in varieties of central Tai (Sagart 1999: 146)

Indo-European *Heg 'I' has been much more stable than 余 Old Chinese *la 'I' (Sagart 1999: 142-144) which is extinct in modern Sinitic languages (though its Middle Chinese pronounciation has been borrowed into Japanese as yo, which is a marginal pronoun for 'I').

Although Korean na 'I', uri 'we', nO 'thou', and nOhUy 'you' have remained stable since Middle Korean, all the Old Japanese pronouns are gone in standard Japanese except for wa(re) 'I' which is archaic. (Granted, Old Japanese is older than Middle Korean, so the comparison is unfair.) *R-L / *L-R

In "More *kl-ues", I discovered that lateral initials only occurred after a limited number of rhymes. It occurred to me that I should look at the vowels in those rhymes:

InitialRhymeRhyme #Overall rhyme #RhymeRhyme #Overall rhyme #RhymeRhyme #Overall rhyme #RhymeRhyme #Overall rhyme #
lh--ejr2.6677-jɨr1.8692none; only three lhw- syllables in Tangut-jwɨɨr1.92100

Overall rhyme numbers refer to the numbers of rhymes if tones are disregarded: e.g., overall rhyme 100 includes both -jɨɨr 1.92 and -jɨɨr 2.85.

Although Tangut allowed l(h)- to combine with nonretroflex -a and -u -

l-la (23)laa (4)lạ (5)lã (7)lu (24)luu (1)lụ (11)
lh-lha (5)̣(no lhaa)lhạ (3)lhã (3)lhu (2)lhuu (1)(no lhụ)

(11.28.00:17: lateral-initial syllables with glides before a and u also exist: e.g., ljwa, lhjụ, etc.)

(numbers in parentheses indicate the number of tangraphs with a given reading)

there were no syllables combining l(h)(j)(w)- with -a(a)r or -u(u)r. How likely is it that pre-Tangut had no syllables like

*rl(j)(w)a(a), *lr(j)(w)a(a), *l(j)(w)a(a)r

*rl(j)(w)u(u), *lr(j)(w)u(u), *l(j)(w)u(u)r

I would expect any random consonantal skeleton to occur more frequently with a than with any other vowel, unless a came from an earlier non-a vowel. And it seems strange that Tangut has five instances of lwər but zero instances of l(h)(j)(w)ur.

Tangut has only one instance of a dental followed by -ur: TT0431 POINT-OUT nur 1.75. There are no syllables like tur, thur, dur. In September, I proposed that

-or > -wər after dentals and alveolar l

Now I wonder if that change should be rewritten as

-ur, -or > -wər after dentals and alveolar l-

(but why do rur and ror exist?*)

nur 1.75 could be an archaism that did not become the expected -wər. Or it could be a reduction of an earlier *n-r-Cu, *n-Cru, or *n-Cur with a non-dental root initial. I can't explain it away as a loanword since I don't know of any similar-sounding Chinese or Tibetan word.

Tangut -(w)ər corresponds to gDong-brgyad rGyalrong -ɯr < Proto-rGyalrong *-ir, *-ɯr, *-ur (i.e., any high vowel before *-r; Jacques 2003: 20). Unfortunately, Jacques lists no cognates for Tangut syllables of the type Twər / lwər, and even if he had, other comparative evidence would be needed to determine that those syllables originally had a rounded vowel.

The missing lar-type syllables might have been 'brightened' to ljiir, etc. Unfortunately, I can't find any correspondences between the retroflex rhymes after Tangut laterals and gDong-brgyad rGyalrong -a(r) which could confirm this hypothesis.

*Perhaps rur and ror exist because they come from pre-Tangut open syllables ru and ro.

The breaking of *u after dentals and l might only have occured before pre-Tangut rhotic codas:

*tur > twər


*ru, *ro > rur, ror

(all vowels of *r-initial syllables developed retroflexion)

This hypothesis assumes pre-Tangut had no syllables like

*tru, *thru, *dru, *lru or

*rtu, *rthu, *rdu, *rlu

which would have become tur, etc. since they had no rhotic codas.

nur 1.75 could then be explained as originating from a rhotic coda-less*nru or *rnu which would not have been subject to the *u-breaking rule. nur 1.75 would not be an exception, but a regular product of a sound change. MORE *KL-UES

In my first *kl-ues post, I proposed the following correspondences:

Proto-rGyalrong *l(j)- : Tangut lh- < *C-l- (without the tenseness trigger)

Proto-rGyalrong *C(V-)l- : Tangut lh- < *C-l- (with the tenseness trigger)

Incorporating my last post, I would now rewrite this as:

Proto-rGyalrong *l(j)- : Tangut lhV < *CV-l-

Proto-rGyalrong *C(V-)l- : Tangut lhṾ < *C-l-

Can a similar pattern be found for Tangut plain l? No, for there was no correlation between Proto-rGyalrong *C(V-)l- and Tangut lṾ̣ in Jacques' (2003) comparisons.

(11.27.00:16: In fact, tense vowels after Tangut l[j]- have the opposite pattern of correlation: i.e., with Proto-rGyalrong *lj-.)

FrequencygDong-brgyad rGyalrongProto-rGyalrongTangut
Plain vowelTense vowelRetroflex vowel
4l-*l-l-, lj-l-, lj-None, not even when PR has *r-
1li- (in qa liaʁ 'eagle')?*rlj-l-
1ɬ- ("aspirated lateral"; rare)unknownl-
8j-*lj-l-, lj-l-
1jl-*jl-l-None, unlike lh-!

All of the gDong-brgyad rGyalrong words in the above comparisons have presyllables, so there is no correlation between GBR presyllables and Tangut vowel tension.

I can only conclude that Tangut vowel tension and retroflexion reflect prefixes present in pre-Tangut that were absent in Proto-rGyalrong.

The very nice correlation between Proto-rGyalrong *C(ə-)l- and Tangut lhṾ with a tense vowel may be an artifact of a very limited data set consisting of only three comparisons:

GlossgDong-brgyad rGyalrongProto-rGyalrongTangut
manuretɯ ɣli*kl-lhjị 2.60
bowtɯ di*tl-lhjɨ ̣1.69
moonsla*sə-l-lhjị 2.60

11.27.1:41: There is only a very weak correlation between PR *lj- and Tangut lj- in Gong's reconstruction. There are several possibilities:

- PR retained an original *-j- that was lost in Tangut

- Tangut retained an original *-j- that was lost in PR

- PR and Tangut retain *-j- in different words

- PR and Tangut developed *-j- in different words

- Tangut -j- has been incorrectly reconstructed in at least some cases.

I suspect the fifth possibility. Chinese transcriptions imply that Tangut did have lj-, even though there are no instances of ly- in Nevsky's (1926) collection of Tibetan transcriptions of Tangut. This absence may simply reflect the absence of the letter combination ly in Tibetan. Then again, the transcriptions contain un-Tibetan letter combinations, so if the transcribers heard a [lj] or [ʎ], they could have written ly. Perhaps the transcriptions reflect dialects of Tangut with and without lj-.

I wonder if the distribution of retroflexion after laterals is significant. In Gong's reconstruction, retroflex vowels never follow a simple initial l-. They are found only in certain retroflex rhymes after l- followed by glides and lh- with or without glides:

689 l-syllables-Ø--j--w--jw-
l-0/2028/251 (all in rhymes 1.92, 1.93, 1.94)5/30 (all in rhyme 2.76)0/33
lh-5/65 (all in rhyme 2.66)7/88 (all in rhymes 1.86, 1.92, 1.94)0/31/17 (rhyme 1.92)

It is curious that

lj- and lhj- are 25% and 35% more common than l- and lj-

lhjw- is almost six times as common as lhw- PREINITIALS VS. PRESYLLABLES: TENSE VS. LAX

In my last post I asked,

I don't think a simple prefix was enough to trigger tenseness. There may have been something more. What was it?

My guess is that the tense vs. lax distinction in Tangut reflects an earlier distinction between nonrhotic preinitials and (optional) presyllables:

Syllable with preinitial other than *r-*C-CVCṾ (tense)
Syllable with (pre)initial *r-, medial *-r-, or final *-r*r(C)V, *CrV, *CVrCVr (retroflex)
Syllable with presyllable*CV-CVCV (lax)
Syllable without preinitial or presyllable*CV

Vowels also tensed after earlier clusters in Korean, though this tension is nonphonemic because it always precedes tense consonants. Tangut may have developed tense consonants from earlier clusters. The tense/lax contrast in consonants was lost, though the contrast remained in vowels:

Stage 1Stage 2Stage 3Stage 4
consonant cluster + lax voweltense consonant + lax voweltense consonant + tense vowelconsonant + tense vowel
Middle Korean and pre-Tangut(hypothetical stage before tenseness spreads into vowel)Modern KoreanTangut

Pre-Tangut presyllables may have

- dropped without a trace

- or dropped after the syllable initial lenited intervocalically:

*CV-kV > *CV-ɣV > ɣV

Labial stops may have lenited to w. Dental stops did not seem to lenite.

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2007 Amritavision