In my last two posts, I proposed the palatalization of dental stops before *i in pre-Tangut:

*ti > *tɕi > tʃɨi

But I have never seen an example of a Tangut alveopalatal affricate which can be derived from a dental stop.

Just now it occurred to me that there may be an example of a Tangut alveopalatal fricative which can be derived from a dental stop.

According to my lenition hypothesis, Tangut ʒ- is partly derived from alveopalatals that lenited after a lost presyllable: e.g.,

*CV-tʃ- > *CV-dʒ- > *CV-ʒ- > ʒ-

Example: ʒiw R47 1.46 'junipdf' < *Cɯ-ʃik

cf. Proto-rGyalrong *ɕɔk

gDong-brgyad rGyalrong ɕɤɣ

Somang ɕə́k

Zbu xɕôx

Written Tibetan shug-pa

If dental stops palatalized before *i, some ʒ may go back to *CV-Ti. 

Matisoff (2003: 674) reconstructed Proto-Tibeto-Burman *t(w)i(j) 'water'*.  A hypothetical Tangut cognate of this word would be tʃɨi from *ti. No such word exists.  The basic Tangut word for 'water' is

TT4282 ʒɨəəʳ or ziəəʳ R100 2.85

Sofronov (1968 II: 371) reconstructed its initial as alveopalatal, but Gong reconstructed its initial as z- and Nishida (1964: 194) reconstructed the cluster ʁz-.  The Tibetan and Chinese transcriptions point toward z-:

Tib gzi(H)

Chn 移則 ?*j-tsə

Could 'water' be from *rɯ-tiəə?

*rɯ-tiəə*rɯ-tɕiəə*rɯ-dʑiəə > *rɯ-ʑiəə  > *rʑiəə*rʑiəəʳʒɨəəʳ


*rɯ-tiəə*rɯ-tsiəə*rɯ-dziəə > *rɯ-ziəə  > *rziəə*rziəəʳ > ziəəʳ

with *t > *ts before *i?

Marco Polo romanized the name of the Tangut city

ʒɨəəʳ or ziəəʳ niaa (lit. 'water black')

as Etsina or Etzina implying that 'water' had an alveolar affricate initial. Could the local Tangut dialect have preserved an older affricate initial *ts- or *dz-?  The E- could represent a presyllable.

Tangut -ɨəə or -iəə does not match Matisoff's *-i or the -i of gDong-brgyad rGyalrong tɯ-ci 'water'.  However, the rhyme alternations

-ɨi-ɨə (Grade III)

-i-iə (Grade IV)

are known, even if their origin is not: e.g.,

TT3592 bi R11 1.11 'low, below, down, inferior'

TT3591 bi R11 2.10 'low, below, down, inferior'

TT5019 bị R70 1.67 'to lower, bend, hang down'

TT0501 biə R31 1.30 'low, below, bottom'

(from Gong, "Phonological Alternations in Tangut", p. 805-807)

TT1020 ʃɨi R10 2.9 'go' (stem 1)

TT1052 ʃɨə R30 1.29 'go' (stem 2)

TT3144 si R11 2.10 'die' (stem 1)

TT5480 siə R31 1.30 'die' (stem 2)

It is not clear whether these alternations involve *-ə suffixation or *i ~ ablaut (as -iə could come from an bent upward to *ɨə). In either case, could there have been a parallel alternation of long -ii with long -iəə (whose i automatically becomes ɨ after the normally Grade III initial ʒ)? If so, ʒɨəəʳ or ziəəʳ 'water' could be from the ə-form of a pre-Tangut root *tii.

*Is the ci of gDong-brgyad rGyalrong tɯ-ci 'water' from *ti?

Middle Chinese 水 *ɕwiʔ 'water' may be another example of the t-word for 'water' if it is from Old Chinese *stujʔ rather than *hlurʔ as reconstructed by Sagart (1999: 157-158). WHY ANOTHER TANGUT RAISING HYPOTHESIS CAN-NOT BE CORRECT

In "Why My Tangut Raising Hypothesis Can-not Be Correct", I proposed a chain shift to explain the unusually low frequency of -i after dentals:

1. *a regularly raised ('brightened') to *i after all nonvelar, nonglottal initials unless this raising were partly or fully prevented by the presence of other segments:

*Ta > *Ti (T = cover symbol for dental stops)

2. After original *Ta was lost, Tangut developed new Ta from earlier *Ta with emphatic *T- (or *Ta flanked by 'brightening blocker' segments).

Pre-Tangut could have had emphatic coronals like Hebrew (ט צ), Classical and Modern Standard Arabic (ص ض ط ظ), and Old Chinese (though OC also had emphatics at other points of articulation).

3. Primary and secondary *Ti were palatalized to *Tɕi (later Tʃɨi) whereas *Ti (less frequent than Ti) lost its emphasis and became Ti.

Tangut has more ni than ti, thi, or di because the initial of primary and secondary *ni did not palatalize.

There are several problems with this hypothesis:

1. There is no evidence for emphatics in pre-Tangut. I believe emphatics in Old Chinese were an OC-internal innovation (though *k/q could be very old in cases like OC *khaʔ 'bitter' corresponding to Zhongu Tibetan qhɐ 'id'.)

2. No Tangut alveopalatal affricate : non-Tangut dental stop correspondences have been found. But has anyone looked for them?

3. If Tangut primary and secondary *di became dʒɨi, why does Tangut di outnumber both ti and thi combined? Tangut bi and gi (but not dzi or dʒɨi!) are also much more common than their voiceless counterparts, and I wonder if Tangut voiced initials go back to nasal clusters in some cases: e.g.,

ti < *ti, ta

thi < *thi, *tha


di < *da, *di, *Nda, *Ndi, *Nta, *Nti ... (many more possible sources)

Tangut voiced obstruents are often transcribed in Tibetan with H- implying prenasalization. Perhaps this prenasalization was an allophonic trace of an earlier phonemic distinction: e.g.,

Pre-Tangut */mt/ : */md/ : */nt/ : */nd/ : */d/

Tangut /d/ [nd] ~ [d]

Tangut dʒɨi should be numerous because my hypothesis predicts that it could be derived from

{*m-, *n-, other presyllables} + {*t, *d, ?*dz, *tʃ, *dʒ} + {*-i, *-a}

Yet there are only 7 tangraphs pronounced dʒɨi, whereas there are 19 bi-tangraphs, 11 di-tangraphs, 12 dz(ɨ)i-tangraphs, and 12 gi-tangraphs. My hypothesis fails to predict the low frequency of dʒɨi.

I tried to fix the hypothesis by deriving Tangut ti from a nonemphatic source:

*pti > *tui > ti > tʃɨi

cf. Korean

*tɯj > *ti > ci

and Japanese

*tui > *ti > chi

However, this new hypothesis still does not solve the dʒɨi-problem. Moreover, there is no external evidence for reconstructing diphthongs as a source of Tangut -i after dentals. According to my hypothesis,

TT4743 ti R11 1.11 (negative imperative)

should be from *tui (or *təi?). But Matisoff derived it from Proto-Tibeto-Burman *ta, and its Qiang cognates have no labial vowel. I could claim that this is from *C-ta with a preinitial that blocked the palatalization of *t, but it is simpler to derive ti from *ta.

Here's a fourth hypothesis:

*Cta > *ta > *ti > tʃɨi

(*C- is a preinitial blocking brightening.)

This avoids the problem of reconstructing a spurious *-u- in the negative imperative, but it still predicts a nonexistent large number of alveopalatal -i-syllables. WHY MY TANGUT RAISING HYPOTHESIS CAN-NOT BE CORRECT

TT1491 mi R11 1.11 'not'

is a high-frequency Tangut grammatical word cognate to Written Tibetan ma 'not', Written Burmese ma 'not' and Old Chinese 無 *ma 'not exist'. It was probably monosyllabic in pre-Tangut. But my previous post predicted that it came from pre-Tangut *Ci-mja. Does any language have a sesquisyllabic or disyllabic basic word for 'not'?

Did pre-Tangut *-a regularly become -i unless this raising was blocked or 'bent' in another direction by other segments*? If Tangut mi is from pre-Tangut *ma, then Tangut ma must be from something else: e.g., *Cʌ-ma with a low-vowelled presyllable. There should be fewer Ca than Ci because Ca could only come from complex sources whereas Ci could be from original *Ci (e.g., mi 'Tangut', cognate to Written Tibetan mi 'person') as well as *Ca. And it turns out that a is less than half as frequent as i in tangraphic readings with bilabial initials:

Homophones Chapter I initials p- ph- b- m- Total i : a ratio
R11 -i 4 10 19 6 39 2.17 : 1
R17 -a 3 5 8 2 18

Can this trend be observed after other initial classes?

Labiodentals: Yes.

Homophones Chapter II initials wh- w- Total i : a ratio
R10 -ɨi 1 9 10 2 : 1
R17 -a 1 4 5

Dentals: No. (Did original *Ti become tʃi? Are all Ti-syllables secondary?)

Homophones Chapter III initials t- th- d- n- Total i : a ratio
R11 -i 4 3 11 14 32 0.93 : 1
R11 -wi 1 1 0 4 6
R17 -a 10 10 5 12 37
R17 -wa 1 0 3 0 4

'Retroflexes': Yes, but the number is too small to be meaningful.

Homophones Chapter IV initials ?dʒ- Total i : a ratio
R10 -ɨi 5 5 5 : 0
R17 -a 0 0

Velars: No.

Homophones Chapter V initials k- kh- g- ŋ- Total i : a ratio
R10 -ɨi 1 0 0 0 1 0.84 : 1
R11 -i 1 8 12 0 22
R11 -wi 6 2 10 0 18
R17 -a 24 6 1 8 39
R17 -wa 2 8 0 0 10

Alveolars: Yes.

Homophones Chapter VI initials ts- tsh- dz- s- Total i : a ratio
R10 -ɨi 0 1 4 0 5 2.07 : 1
R11 -i 7 8 8 18 41
R11 -wi 1 0 2 9 12
R17 -a 3 4 4 10 21
R17 -wa 2 1 1 3 7

Alveopalatals: Yes.

Homophones Chapter VII initials tʃ- tʃh- dʒ- ʃ- Total i : a ratio
R10 -ɨi 12 4 7 10 33 2.58 : 1
R10 -wɨi 1 5 2 8 16
R18 3 5 5 3 16
R18 -wæ 1 1 0 1 3

Alveopalatals cannot occur before R17,. so I have substituted R18.

Glottals: No.

Homophones Chapter VIII initials ʔ- h- ɦ- Total i : a ratio
R10 -ɨi 2 0 0 2 0.32 : 1
R11 -i 6 2 0 8
R11 -wi 1 0 0 1
R17 -a 3 8 16 27
R17 -wa 0 4 3 7

Liquids: Yes.

Homophones Chapter IX initials l- lh- r- z- ʒ- Total i : a ratio
R10 -ɨi 17 0 0 1 3 21 1.51 : 1
R11 -i 0 10 (29) 10 0 49
R11 -wi 1 3 0 0 0 4
R17 -a 23 2 (17) 5 0 47
R17 -wa 1 1 0 0 0 2

Since r- almost never occurs before nonretroflex rhymes, I have inserted figures from the retroflex counterparts of R11 and R17.

i-syllables outnumber a-syllables in all initial classes except dentals, velars, and glottals.

Perhaps *a did not raise to i after back initials. Matisoff lists no examples of R10/R11 etyma with back initials. (Gong's ʔjiʳ = my jiʳ 'hundred' appears to be a counterexample, but its retroflexion points toward an earlier initial *r-, not an original initial glottal stop.)

Dentals do not form a natural class with velars and glottals. However, emphatic (pharyngealized or velarized) dentals might have blocked raising. There may have been a chain shift: original plain dentals palatalized and original emphatic dentals became new plain dentals:

Pre-Tangut Tangut: loss of emphatics
Stage 1 Stage 2: raising Stage 3: palatalization
*ti *ti (merger) *tɕi tʃi
*ti ti
*ta ta

*The development of *a (or any pre-Tangut root vowel) could have been influenced by

- a presyllabic initial

- a presyllabic vowel

- a root initial

- a medial (*-j-, *-w-)

- a coda: e.g.,

*-aj > *-ej > -e

*-aw > *-ow > -o

*-ak > *-aɣ > -aa WHAT RAISED PRE-TANGUT *-A?

In "'Brightening' and the place of Xixia (Tangut) in the Qiangic branch of Tibeto-Burman", James Matisoff gave many examples exemplifying five types of 'brightening' (vowel raising) in Tangut. I list them below with my reconstructions instead of those used by Matisoff. Length, tenseness, and retroflexion have been ignored from simplicity. The rhyme groups are Gong's.

1. Rhyme group I

1.1. *-a > -əu (Grade I)

1.2. *-a > -ɨu (Grade III)

1.3. *-a > -iu (Grade IV)

2. Rhyme group II

1.1. *-a > -əi (Grade I)

1.2. *-a > (Grade II)

1.3. *-a > -ɨi (Grade III)

1.4. *-a > -i (Grade IV)

3. Rhyme group VI

2.1. *-a > (Grade I)

2.2. *-a > -iə (Grade IV)

4. Rhyme group VII

3.1. *-a > -e (Grade I)

3.2. *-a > (Grade II)

3.3. *-a > -ie (Grade IV)

5. Rhyme group X

5.1. *-a > -o (Grade I)

(Questionable since Matisoff's derivation of ŋo 'ill' from *na requires a change of initial as well as vowel. Are there any other examples of Tangut ŋ- corresponding to n- in other languages?)

6. Rhyme group IV (little or no raising)

3.1 *-a > -a (Grade I)

3.2. *-a > -ɨa (Grade III)

3.3. *-a > -ia (Grade IV)

There are six basic vowel types in my Tangut reconstruction*:
i ə u
e a o

If Matisoff is correct, *a can become seemingly any vowel in the five other categories. Some historical linguists would be very skeptical about what they would regard as very loose correspondences. What is to stop one from claiming that a Sino-Tibetan word with *a is cognate to a Tangut form with almost any vowel? Moreover, what about regularity? Did *a become other vowels at random? Or is Matisoff's data full of false cognates that happen to share an initial consonant with non-Tangut words?

I don't think such skepticism is necessary. Perhaps the development of *a was conditioned by combinations of presyllable vowels and medials:

Pre-Tangut Grade III/IV: presyllable with high vowel Grade I: at least one presyllable with low vowel
*Ci- *Cɯ- *Cu- *Cʌ-Ci- *Cʌ-Cu- *(Cʌ-)Cʌ- No presyllable
*-a -ɨ/ie (none?) -e -o -a
*-ja -(ɨ)i -iə -ɨ/iu -əi -əu -ɨ/ia (Grade III/IV)

Medial -ɨ- (Grade III) or -i- (Grade IV) was generally conditioned by initials.

The conditioning factor for Grade II may have been some sort of uvular. If I represent it as *X, then Grade II and could be from pre-Tangut *X(-C)i-Cja and *X(-C)i-Ca.

Examples (the numbering is Matisoff's; tones are ignored):

1. 'axe': *si-pja > wɨị (*s- > tense vowel)

3. 'lend': *ri-nja > niʳ

4. 'son' (with intervocalic lenition of *-ts- to *z- before presyllable loss):

*Ci-tsja > zi

*Cɯ-tsja > ziə

*Ci-tsa > zie

10. 'meat': *ki-sja > *ki-sje > *ki-si > *ksi > *kʂɨi > *tʂhɨi > tʃhɨi

(cf. *ks > in Sanskrit)

(Pre-Tangut and maybe even Tangut initial sibilant fricatives might have been aspirated, though this aspiration only became phonemic when they fused with stop prefixes to become affricates.)

15. 'hundred': *Ci-rja > *Ci-rjiʳ > jiʳ

46. 'bird': *Cʌ-Ci-wja > *Cʌ-Ci-wi > *Cʌ-Cʌ-wəi > wəi

47. 'chew': *X-ri-kja > kɪʳ

48. 'hoof': *Ci-kwa > kwe

51. 'arm': *Xi-kwa > kwɛ

58. 'bitter':

*k(ʌ)-ka > *xka > kha

*X-ki-kja > *xkɪ > khɪ

61. 'five': *pɯ-ŋa > *pɯ-ŋə > *pŋə > ŋwə

67. 'hot':

*tsja > tsɨa 'broil; roast' (Grade III is unusual for initial ts-, and may be conditioned by a prefix)

*k-tsja > *xtsja > tshja 'burn'

70. 'ditch': *Nu-kja > *Nu-kju > *ŋkju > giu

?73 'ill': *Cʌ-ku-na > *Cʌ-kʌu-no > *kno > *gno > ŋo

74. 'love':

*Cʌ-Cu-dzja > *Cʌ-Cʌu-dzju > dzəu

*Cu-dzja > dzɨu (Grade III is unusual for initial dz-, and may be conditioned by a prefix)

*Various researchers reconstruct a similar six-vowel system for Old Chinese. If Proto-Sino-Tibetan existed, did it also have a six-vowel system? RGYALRONG RHYMES CORRESPONDING TO R50 AND R53

Guillaume Jacques couldn't find any.

That's it. I'm going to bed.

No, wait. Although his paper lacks gDong-brgyad rGyalrong cognates for Tangut R50 words, it does have cognates for the base forms of R53 words:

Gloss gDong-brgyad rGyalrong Tangut cognate Derived R53 form
rG cook in embers; T burn kɤ pu piu R3 2.3 puo R53 1.51
rG close, T do kɤ pa wɨi R10 1.10 wuo R53 1.51
eat kɤ ndza dzɨi R10 1.10 dzuo R53 1.51
wear clothes kɤ ŋga gwi R11 2.10 gwuo R53 2.44
send on a mission kɤ ɣɤ xpra phi R11 2.10 phuo R53 2.44

In "西夏語中的漢語借詞", Gong found no R50 loans from Chinese and identified only a single R53 loan:

TT5109 dʒuo R50 1.51 'long' < pre-Tangut period NW Chn 長 ?*dʒo < Middle Chinese *ɖɨaŋ
The affricate initial and nonnasal vowel of dʒuo indicate that it was borrowed

- after NW Chinese shifted retroflex stops to affricates

- after NW Chinese lost its nasal vowels

- before NW Chinese shifted voiced obstruents to voiceless aspirates

Are all R53 words derived from bases with other rhymes or borrowed?

I have been assuming that R53 was partly derived from earlier *-o preceded by a high-vowelled presyllable that conditioned upward vowel bending:

*Cɯ-Co > *Cɯ-Cɯo > *Cɯo > *Cuo

I doubt that there no pre-Tangut roots ending in *-o and preceded by *Cɯ-. For now, I will reconstruct pre-Tangut *Cɯ-Co for any R53 word without any known non-R53 base or foreign origin: e.g.,

TT4243 dzuo R53 2.44 'poem' < ?*Cɯ-dzo FROM PERSON TO PERSON: R2~R53 ALTERNATION?

After straining my eyes last night to compare Li Fanwen radicals 118 and 119, I'd like to write a post with almost no reference to tangraphs.

In "The Phonological Reconstruction of Tangut", Gong Hwang-cherng established the following rhyme alternations between R53 and five different rhymes (given here in my reconstruction);

R53 -io (Grade III/IV neutralized) R3 -iu (Grade IV)
R10 -ɨi (Grade III)
R11 -i (Grade IV)
R36 -ɨe (Grade III)
R37 -ie (Grade IV)

I suspect that R53 -io is a merger of Grade III *-ɨo with its less exotic Grade IV counterpart *-io.

The earlier grades of R53 morphemes can be determined from the grades of their non-o counterparts:

TT2413 wɨi R10 1.10 (Grade III) ~ TT3666 wio R53 1.51 < *wɨi-o (Grade III) 'to do, act, make'

TT2062 mi R11 1.11 (Grade IV) ~ TT2015 mio R53 2.44 < *mi-o (Grade IV) 'to hear'

TT5753 dʒɨe R36 2.32 (Grade III) ~ TT4920 dʒio R53 2.44 < *dʒɨo (Grade III) 'to have'

TT1272 sie R37 2.33 (Grade IV) ~ TT0063 sio R53 2.44 < *sio (Grade IV) 'to know; knowledge'

I am not certain whether the last two cases involve e ~ o alternation or the simplification of triphthongs:

*ɨeo > *ɨo > -io (Grade III)

*ie-o > *io (Grade IV)

Although I would expect a Grade III counterpart

R2 -(ɨ)u ~ R53 -io (< ?*-ɨu-o)

to the alternation of Grade IV R3 -iu with R53 -io, Gong lists no examples of R2 ~ R53.

I wonder if these R2 and R53 words for 'person' are cognates in spite of their initials:

TT1087 dʒwu R2 1.2 'person'

TT4105 dzwio R53 2.44 'person'

Perhaps the alveopalatal initial of TT1087 was conditioned by some lost prefix:
?*C-dz- > dʒ-
Compare with how Middle Chinese *dʐ- arose from Old Chinese *r-dz-:

儕 MC *dʐɛj < OC *r-dzəj 'class, category' (that which is equal)

(not attested in early OC; may be a late derivation)

root is 齊 MC *dzej < OC *dzəj 'equal'

As for the rhyme, it looks difficult at first to derive R53 -wio from R2 -wu. However, I could revise my reconstruction of R53 as -wuo which could have originated from R2 -wu + -o. My -i- was based on the -j- in Gong's reconstruction dzjwo R53 2.44. There is no transcriptional evidence for a -j- in this rhyme. Tai Chung Pui (2008: 219) lists five different Tibetan transcriptions of R53:

-o, -oH, -ooH (sic), -wo, -uH

Each only appears once in the manuscripts, so none can be claimed to be typical. None contain i or y, even though -yo is a normal letter combination in Tibetan. TT4105 'person' was transcribed as Hdzwo and bdzo. -wo and -o are not far from my -wuo. The non-w counterpart of -wuo would be -uo (which is what I reconstructed back in April).

If R53 had no palatal element, how can I explain its alternations with palatal rhymes (R3, R10, R11, R36, R37)? Perhaps all high vowels merged to -u- before -o:

Stage 1 Stage 2 Stage 3 Stage 4
*-(w)i-o *-(w)io *-(w)ɨo -(w)uo
*-(w)ɨi-o *-(w)ɨo
*-(w)u-o *-(w)uo *-(w)uo

Gong reconstructed R50 as -jwo which is indistinguishable from his R53 -jwo with a medial labial. Since I no longer reconstruct R53 as -(w)io, perhaps R50 was -io or -yo with a palatal element. R50 occurs only before alveopalatal initials (ʃ- tʃ- tʃh- dʒ-) with two exceptions (Sofronov 1968 II: 33, 57, 256):

TT4278 lio R50 1.48 'winding; crooked' (cognate to Old Chinese 邪 *lja 'crooked'?)

TT1442 lhio R50 1.48 'return; retreat'

R50 was transcribed with sinographs of the Middle Chinese *-ɨaŋ rhyme class that ended in *-jõ in pre-Tangut period NW Chinese (cf. its Tibetan transcriptions -yo, -yong). By the Tangut period, these sinographs may have ended in *-jo.

R50 was transcribed as -o(H) in Tibetan without any -y- (Tai Chung Pui 2008: 218). However, this is not counterevidence against my -i-, because Tibetan spelling conventions do not allow -y- after palatals and l(h)-.

One huge problem with reconstructing R50 as -io is that I cannot explain

- why palatal rhymes plus -o became R53 -uo instead of R50 -io

- why R50 -io did not merge with the *-io that became R53

Another problem is that R50 -io has a Grade IV -i- but is listed before Grade I R51 -o. Arakawa (1999: 41) reconstructed both R50 and R51 as Grade I -o. I am reluctant to follow Arakawa for two reasons:

First, the alveopalatal initials of most R50 syllables are not found in Grade I. (But they are also extremely rare in Grade IV, so my -io reconstruction is barely any better.)

Second, l- can occur before R50 and R51. TT4278 would be lo R50 1.48 in his system which would be homophonous with TT0160 lo R51 1.49 'filter'.

There are not many R50 syllables, so I could try to look at them all in a future post. DOUBLE IM-PERSON-ATION

In "Im-person-ating Radical 118", I cited Tangraphic Sea analyses that apparently equated

Li Fanwen radical 119 (variant of 118?)


Li Fanwen radical 118 'person'


TT1087 dʒwu R2 1.2 'person'

I just noticed that Han Xiaomang (2004: 204 #3527) wrote TT1087 with radical 119 instead of 118 on the right. Yet he wrote its mirror-image counterpart

TT3485 wị R70 1.67 'person'

with radical 118 (2004: 130 #2169) on the left. Both editions of Homophones have radical 118 in TT1087 and 3485 (Sofronov 1968 II: 208, 135). Is there a Tangut text with 119 in TT1087? I wish I could see the original text of Tangraphic Sea to see whether its version of TT1087 has 118 or 119.

The Tangraphic Sea implies that TT1087 should have 119:


TT1087 dʒwu R2 1.2 'person' =

right of TT3485 wị R70 1.67 'person' +

119 = right of its homophone TT2855 dʒwu R2 1.2 'lightning' (phonetic?)

This analysis makes me think TT1087 should be the tangraph for the Japanese superhero Inazuman (< inazuma 'lightning' + Eng man).

On the other hand, the analysis of TT3485 implies it should have 118:


TT3485 wị R70 1.67 'person' =

118 = right of TT4105 dzwio R53 2.44 'person' +

right of TT1087 dʒwu R2 1.2 'person'

If this were not confusing enough, different editions of Homophones have 118 or 119 under a vertical line (Li Fanwen radical 165; meaning unknown) in TT1087 and TT3485. Or maybe I'm seeing a difference that isn't there.

In Precious Rhymes of the Tangraphic Sea 1.23.1602, TT3485 appears to have radical 165 containing 118 instead of 119. I can't find TT1087 in the fragments of PRTS volume 3 that are available online.

Han Xiaomang (2004: 130 #2169, 204 #3527) has radical 165 containing 119

in both TT3485 and TT1087.

Sofronov (1968 II: 302) handwrote radical 165 consistently with 119 except in

TT1080 ziuʳ R81 2.70 'give birth to' (see my Sarah birthday post for near-synonyms)

which appears in both editions of Homophones with 118 inside 165 (Sofronov 1968 II: 266). Nonetheless, Han Xiaomang (2004: 205 #3539) wrote TT1080 with 119 inside 165. Li Fanwen (1997: 1141) filed TT1080 under radical 165 which was consistently printed with 118 inside (at least in left-hand position).

Is radical 165 containing 118 an acceptable variant or a radical distinct from 165 containing 119? I have no idea. TT1080 is not in any extant analyses. I don't even know what the semantic or phonetic values of radical 165 are. IM-PERSON-ATING RADICAL 118

I don't know of any tangraphs with

Li Fanwen radical 119 (meaning unknown)

in non-left-hand position. I have no index of tangraphs by center component, and neither Grinstead 1972 nor Kychanov 2006 have a separate listing for 119 as a right-hand component.

One might expect tangraphs with 119 on the left to have analyses invariably beginning with each other, but two of the available analyses in fact seem to equate 119 with 118 'person':


TT3748 ʃiõ R58 1.56 'base/mean person' =

right of TT1087 dʒwu R2 1.2 'person' (cognate to dzwio 'person'?) +

left of TT3657 dʒu R2 2.2 'timid, weak'


TT3750 R28 1.27 (surname/transcription) =

right of TT1087 dʒwu R2 1.2 'person' +

right of TT0625 nieʳ R79 1.74 'wild animal'

Are these analyses simply wrong? If not, why would radical 118 be changed to radical 119?

Grinstead (1972: 58) asked,

Could the fusion process [used to create tangraphs] have led to duplication of characters, with embarrassing results [e.g., tangraphs with double readings]? I believe that there are non-typical elements, which have been formed out of necessity from the general system of elements ... apparently to make a distinction from [tangraphs for] other words. There seems to be no question of taboo, merely that another character already existed with this form.

He proposed that radical 119 was a "non-typical" element derived from radical 118 'person'. If Grinstead is correct, then I would expect each tangraph with radical 119 to have a near-lookalike with radical 118. I would not expect obvious derivatives of radical 119 tangraphs such as


'earth goddess', 'god' < 'lightning'

to have radical 118 counterparts.

However, the only pairs of lookalikes I have found are

TT3745 riʳ R84 2.72 'pass; before' (with radical 118 twice)

TT3748 ʃiõ R58 1.56 'base/mean person' (with radical 119 on the left)

TT3934 da R17 2.14 'old; aged'

TT3643 tsiʳ R84 1.79 'select; choose'

TT4001 tshwiu R3 1.3 (second half of disyllabic word lhiạ tshwiu 'lightning')

I cannot find any counterparts of

with radical 118 on the left.

But maybe I shouldn't be entirely surprised because the last two may be ultimately derived from tshwiu.

I have my doubts because I find it hard to believe that the graph for the second half of lhiạ tshwiu was derived from the graph for the first half in spite of the Tangraphic Sea analysis:


TT4007 lhiạ R67 1.64 'sparkle; glisten; lightning' =

left of TT4001 tshwiu R3 1.3 +

left of TT2855 dʒwu R2 1.2 'aerolite; lightning' (Kychanov 2006: thunder and lightning; thunderstorm; sparkle')

Note that TT2855 has radical 119 on the right. Is it alone? It looks like it could be a mirror derivative of lhiạ,

which combines with it to form

dʒwu lhiạ 'lightning'

Tangraphic Sea has a very different analysis of TT2855:


TT2855 dʒwu R2 1.2 =

left of TT2852 dii R14 1.14 'thunder' +

right of TT5712 ɣiu R3 1.3 'smoke'

What does 'smoke' have to do with 'lightning'?

And what does 'lightning' have to do with the surname

tʃɨə zwị

whose first half has the Tangraphic Sea analysis


TT3686 tʃɨə R30 1.29 =

right of TT2855 dʒwu R2 1.2 'lightning' +

right of TT3454 (dental initial; pronunciation otherwise unknown) 'draw close to one another; become related to'

Another surname tangraph

TT3750 R28 1.27

has no near-lookalike with radical 118 instead of 119, so why was it written with 119?

Although Grinstead (1972: 148) does list a tangraph looking like TT3750 with 118, it appears in both editions of Homophones (Sofronov 1968 II: 175) with radical 120 (meaning unknown) which looks like a person with a long left leg:

TT3747 khie R37 2.33 'musk'

Han Xiaomang (2004: 160) does not mention any variants.

I wondered if TT3750 could be a mirror derivative of a tangraph with the structure

'beast' + 119

to distinguish it from another tangraph

'beast' + 118 'person'

but no tangraphs with 'beast' plus 118 or 119 exist in Sofronov 1968, Grinstead 1972, Li Fanwen 1997, or Han Xiaomang 2004.

Could 118 be an error for 119 in TT3750? TT3750 appears in both editions of Homophones with 119, but its derivative and homophone


TT1585 R28 1.27 'tail'

left of TT1582 tʃɨi R10 1.10 'meat' +

all of TT3750 R28 1.27 (phonetic)

appears with 118 or 119 depending on the edition of Homophones (Sofronov 1968 II: 195). (The Mojikyo font version contains 118. Han Xiaomang [2004: 19] does not mention the variation.) The Tangraphic Sea analysis implies that 119 is the correct radical since TT1585 contains "all" of TT3750 with 119, though Han Xiaomang (2004: 19) considers 118 to be the correct radical.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2008 Amritavision