I am interested in complex initial consonant clusters because I believe that they existed in languages like Old Chinese and pre-Tangut as transitional stages between polysyllables and single consonants:

early *CVC- > transitional *CC- > late C-

early Old Chinese *rɯ-n > Old Chinese *rnɨaʔ > Mandarin nü 'woman'

pre-pre-Tangut ?*sɯ-me > pre-Tangut ?*smie > Tangut miẹ R64 1.61 'woman'

By looking at the development of consonant clusters in attested languages, I can propose more realistic reconstructions. Although it's possible that Chinese and Tangut underwent sui generis paths of evolution, it is more probable that they didn't.

Slavic languages are famous for their initial consonant clusters: e.g., zdr- in Russian здравствуйте 'hello'. (What is its etymology?) And Czech is sometimes cited as an example of a language with a nearly perfect phonemic orthography. So when I learned a while back that Czech had the written clusters jd-, jm-, and js-, I wondered whether they should be taken at face value.

Yesterday a coworker mentioned Harkins' A Modern Czech Grammar (1953 - not exactly modern now!) and I went nuts looking for it. It contained the following passage about jC-clusters on p. 14:

Initial j may be dropped if a consonant follows:

jdu is pronounced "du"

jsou is pronounced "sou"

And I presume jmeno 'name' is [meno] (as indicated by my Czech-English dictionary, Trnka 1991).

The j- in these clusters may be silent (I've never actually heard Czech, so I don't really know), but they are not entirely arbitrary, as they correspond to i- or je- in Polish and и- or е- in Russian:

Cz jdu 'I go' : Pol idu and Rus иду 'id.'

Cz jsem 'I am': Pol jestem and Rus есмь (obsolete) 'id.' (was Pol -t- by analogy added by analogy with jest 'he/she/it is'?)

(Cz jsou 'they are' corresponds to Pol and Rus суть; cf. Skt santi and Lat sunt - was Cz j- added by analogy with other forms in the paradigm: jsi 'thou art', je[st] 'he/she/it is', etc.?)

Cz jmeno 'name' : Pol imię and Rus имя 'id.'

I assume that unaccented i- and je- in pre-Czech were reduced to j- which was then lost.

Is it possible that Old Chinese and/or pre-Tangut once had unaccented *i- and *jV- presyllables that were similarly reduced to j- and then lost?

jC-clusters exist in a modern East Asian language, Japhug rGyalrong, though some of them derive from proto-rGyalrong *lC- clusters, according to Guillaume Jacques (2004: 333-334):

*lC-type *jC-type
Proto-rGyalrong *lp- *lph- *lm- *lk- *lq- *jt- *jn- *jl- *jŋ-
Japhug rGyalrong jp- jph- jm- jk- jʁ- jt- jn- jl- jŋ-

It is odd that *jŋ- is the only example of *j- before a noncoronal. All other *jC-clusters contain coronals.

It is also strange that there is only a single aspirated *lC-cluster. Why are there no *lkh-, *jth-, etc.?

Since *l- and *j- are in complementary distribution, I wonder if one could simply reconstruct *l- in all instances, though this would entail the cluster *ll-. (I reluctantly reconstruct a tense *ll- as a pre-Tangut transitional initial between early pT *sl- and Tangut l- when followed by a tense vowel.) Maybe *jl- and some other *jC-clusters could be from earlier *iC- or *jVC-.

Guillaume Jacques (2004: 314, 316, 330) also reconstructed Proto-rGyalrong *wC-clusters. Could these be from earlier *uC- or *wVC-? BLUE + BIRD = ?

... bluebird? Not according to the Tangraphic Sea:


TT0647 pia R20 1.20 (miswritten as TT0646 in Sofronov 1968 II: 293) =

frame (= left and right sans center) of TT0646 pia R20 1.20 (second half of pia pie 'dark blue') +

left of 'bird' < left of TT5717 dʒwɨõ R58 1.56 'bird'

TT0647 represents the first half of pia piu 'butterfly':

I mentioned the second half on Tuesday when I was discussing tangraphs containing 几 with a line across it.

The Tangut word appears to be reduplicative since the second syllable is almost identical to the first. Are there other disyllabic words with the pattern Cia Ciu? Tangut linguistics has paid very little attention to disyllabic words, perhaps because Tangut dictionaries were organized by tangraphs which could represent parts of words as well as whole words.

The structure of the tangraphs for pia piu is even more opaque than the etymology of butterfly. Although fly is obvious, the reasoning behind butter is not.

I am a little surprised that pia contains


since its Chinese equivalent 蝴蝶 contains 虫 'insect'. I would expect both tangraphs to contain


but only one tangraph contains 'bird' and the second has no known reference to insects or birds.

pia 'blue' is obviously phonetic in pia 'butterfly'. I doubt that pia piu is a reduplication of a root pia 'blue': i.e., 'blue-blah'.

The structure of pia in 'dark blue' is obscure. Its Tangraphic Sea analysis is not very helpful:


TT0646 pia R20 1.20 (first half of 'dark blue') =

frame of TT0647 pia R20 1.20 (first half of 'butterfly') +

'person' < center of TT1429 pie R37 1.36 (second half of 'dark blue')

The actual analysis of pia may involve a phonetic element pia also found in

TT3075 pia R20 1.20 (first half of pia pie 'revere')

with 'fear'on the left; cf. right side of Chn 怖 'fear'

The functions of the remaining components


'surround'+ 'person''

of pia in 'dark blue' are unknown.

The analysis of pie in 'dark blue' is unknown. It looks like


'not' + 'person' + a vertical line

which hardly adds up to 'dark blue'.

I am surprised that there are homophonous reduplicated words pia pie 'dark blue' and pia pie 'revere'. (Was dark blue a reverential color?) I cannot think of any similar cases in Chinese.

The second halves of the two pia pie-s are written completely differently:

The pie of 'revere' is written as 'fear' plus an unknown element. CATTLE, KETTLE, KOTEL

You might have noticed that I had mistranslated

TT2591 riəʳ R92 2.77 '(three-legged) kettle, cauldron (< 'three-legged cauldron'?)'

on Monday as 'cattle' instead of 'kettle'. This is because I was more or less using the English definition in Kychanov (2006: 354):

cattle (sic!); three-legged cattle (sic!); cauldron

I did see the Russian definitions

котел 'kettle'

треножник 'tripod' (lit. 'three-leg-nik')

сковорода 'frying pan'

but I mistakenly assumed that котел was cognate to cattle! It may actually ultimately be from Latin catillus 'a small dish or plate' which is also the source of English kettle.

The actual Russian word for 'cattle' is скот which is obviously unrelated to its English translation (even though both happen to contain C-shaped first letters and the consonant t).

(2.13.0:29: I wonder if Russians notice the accidental resemblance between скот and Scot[land] and Scott.)

I thought it was strange that a graph for 'cauldron' could also mean 'cattle', but I couldn't rule such an oddity either, since many graphs do not represent the apparent sum of their parts: e.g., on Tuesday I mentioned


'sweet' = 'move' + 'tilt'


'adobe building; deprive' = 'tilt' + 'move'


'tongue' = 'person' + 'move'

I would expect the first two to mean 'lean toward' and the third to refer to human motion.


'cattle' (really 'kettle') = 'metal' + 'sweet'

made about as much sense as any of the above three combinations.

I considered the possibility that 'cattle' was simply coincidentally homophonous with 'cauldron', but 'three-legged cattle' (really 'three-legged kettle'!) seemed like a possible transitional meaning between the two.

If a word for 'three-legged kettle' sounds odd or overly specific, Grinstead's (1972) English-Tangut dictionary has entries for

'necklace of human heads' (yikes!)

'rack under rafter'

'thoughts in dreams' (could be useful)

'troubles of rebirth'

'vertebrae of neck'

'where the moon comes up!' (sic - with exclamation point!) SWEET METAL

The Tangut character

TT2591 riəʳ R92 2.77 '(three-legged) kettle, cauldron (< 'three-legged cauldron'?)'

has no known analysis but looks like 'metal' atop TT1079 lɨẹ̃ R65 2.55 'sweet':


This sequence of elements happens to match Tangut noun-adjective order: Tangut 'metal sweet' is equivalent to English 'sweet metal'.

I have no idea what 'sweet' has to do with kettles or cauldrons.

The analysis of 'sweet' is also unknown. It looks like 'move' + 'tilt' which makes no sense:


Its mirror image is TT4775 thiẽ R43 1.42 'adobe building; deprive; a place name and surname'


which sounds like the Tangut period NW Chinese word for 'sweet', 甜 *thiẽ. These two words are ultimately cognate since TPNWC *thiẽ goes back to Old Chinese *-hlim (?*-lim) or *hlem.

These two tangraphs share part of TT3909 lhwia R20 1.20 'tongue' which looks like 'person' + 'move' (why?):

lhwia may be from an earlier *Pɯ-lha which might be cognate to the Tangut and Old Chinese l-words for 'sweet'. WHAT LIES UNDER TANGUT MOUNTAINS?

The Tangut character for 'mountain'

seems to consist of

Nishida radical 30 'mountain'

atop Kychanov radical B300 resembling 几 with a line across it. At first glance, B300 seems to only occur in bottom position as in 'mountain' or

TT5553 piu R3 2.3, second half of pia piu 'butterfly' (beneath 'temperament - why?; no other piu-tangraphs contain it, so it cannot be phonetic)

or in bottom right-hand position as in

Tangraph Tangut Telecode Reconstruction Rhyme Tone.rhyme Gloss Notes
1056 dʒɨe R36 1.35 cold (n.); frost; to freeze 'ice' on bottom left
0983 lhooʳ R102 1.94 elbow; measure of length why is 'wood' on top?
0997 a surname 'person' (semantic) + 'elbow' (phonetic)
3479 R29 1.28 willow; weak; feeble; thin 'wood' on top; bottom right from 'elbow' (why?)

None of these four tangraphs have any common semantic or phonetic characteristics (with the exception of TT0997 which clearly contains TT0983).

According to Tangraphic Sea, B300 in those four tangraphs has the following sources:





The most interesting is the first derivation


which implies

- that the left-hand version of B300 is ㄇ with a line through it

- that the left-hand side of TT5190 jaaʳ R89 2.75 'day'

is a compound of a variant of

with a long final stroke over left-B300. This variant with or without left-B300 does not appear in any other tangraphs.

The right side of 'day' is the high-frequency mystery element ヒ. Perhaps the left side of 'day' is its core (though why those elements represent 'day' is a mystery) and ヒ is a filler (though I still don't know why some radicals need fillers and others don't).

I don't understand why 'day' is the source of B300 in 'cold' since the two are not homophones and have no semantic relationship. HERMITS + FRYING PANS =

... mountains in the Tangut script. The Tangraphic Sea analyzes TT2856 ŋəʳ R90 1.86 'mountain' as


top of TT2862 ʃɨi R10 1.10 'hermit' +

boittom of TT2576 ŋəʳ R90 1.86 'tongs; griddle; frying pan'

(Is such a scope of meaning really possible? Nothing cooks on tongs.)

This cannot be the real analysis of 'mountain', since ŋəʳ 'mountain' is obviously phonetic in ŋəʳ 'frying pan'. The Tangraphic Sea analyses 'frying pan' as


'metal' = top of TT2591 riəʳ R92 2.77 '(three-legged) kettle, cauldron (< 'three-legged cauldron'?)' +

all of TT2856 ŋəʳ R90 1.86 'mountain'

'Frying pan' is probably a simple semantophonetic compound: the metal thing that sounds like ŋəʳ

'mountain'. The tangraph for 'mountain' must predate the tangraph for 'frying pan'.

'Mountain' may be a compound of

Li Fanwen radical 230 ?'mountain'

with Kychanov radical B300 resembling 几 with a line across it. Crossed-out 几 is probably not a phonetic since I know of no tangraphs with it pronounced ŋə(ʳ).

The Tangraphic Sea analysis of 'hermit' is probably reliable:


?'mountain' < top of TT2856 ŋəʳ R90 1.86 'mountain' +

all of TT3344 dzwio R53 2.44 'person'

It must be a calque of Chinese

仙 'hermit' = 亻 'person' (semantic) + 山 'mountain' (semantic/phonetic)

and it too must postdate the tangraph for 'mountain'.

The association between mountains and hermits may be indigenous to Tangut culture since there is a native Tangut word for 'hermit'

ŋəʳ dʒɨee dị

lit. 'mountain dwell true/kind/calm'

It is tempting to regard ʃɨi 'hermit' as a loanword from Tangut period NW Chinese 仙 *siẽ or an even earlier *sien, but neither the initials nor rhymes match. Hence the vague similarity between the two words is coincidental. DIE FERSE DES KURZEN BERG (translation*)

Two people figured out the solution to Friday's mystery.

ŋəʳ viẹ khiə swiẽ

lit. 'mountain short heel'

is Jacob Kurtzberg, better known as Jack Kirby, king of comics! Congratulations!

ŋəʳ 'mountain'

corresponds to -berg 'mountain'.

I don't know of any cognate. The most similar word I can find is Old Chinese 嶽 *rŋok 'mountain; peak' whose rhyme doesn't match. I would expect the corresponding Tangut word to be ŋoʳ.

viẹ 'short'

corresponds to Kurtz- 'short' (the standard German spelling is kurz).

This word could be from an earlier *sɯ-Pe (P = unknown labial stop) or *sɯ-we with a high-vowelled presyllable conditioning upward vowel bending, vowel tensing, and possibly lenition:

*sɯ-Pe > *sɯ-Pie > *sɯ-bie > *sɯ-βie > *svie > *vvie > *vviẹ > viẹ


*sɯ-we > *sɯ-wie > *swie > *wwie > *wwiẹ > viẹ

Old Chinese 庳 *N-pe-ʔ 'short (in stature)' looks similar but is derived from OC 卑 *pe 'low'. The Tangut word viẹ refers to length, not height: e.g., bəị viẹ 'short spear' (Nevsky 1960 II: 568).

khiə swiẽ 'heel' (mistranslated by Nishida 1964: 206 as 'groin'!)

is a loose equivalent of Jacob < Hebrew יעקב Ya`akov, associated with עקב `akev 'heel' (which has all but the first Hebrew letter of 'Jacob').

khiə is 'foot'. It superficially resembles Old Chinese 腳 *kak 'foot; leg', but the initials and vowels do not match. (2.9.0:02: I expect OC *-k to correspond to Tangut zero after the achromatic vowels schwa and a.)

swiẽ may be a cranberry morpheme unless it is attested as an independent word. Just as cran- differentiates a cranberry from a berry, swiẽ differentiates khiə swiẽ 'heel' from khiə 'foot'. swiẽ may be an old word for 'heel' or 'root' (cf. Chn 跟 'heel', homophonous with/ the same word as 根 'root'). Perhaps swiẽ is from an earlier *Pɯ-seN (N = unknown nasal) which doesn't look like anything else I can think of.

*'The heel of the short mountain'. des = 'of the'.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2009 Amritavision