I wish



mə-wəəi niəə bɛɛ-reʳ

lit. 'birth day happy'

in Tangut.

means 'clean' and is written as 'water' plus 'light'.

To means 'end' but Grinstead (1972: 55) glossed it as 'lightning' since it appears in many characters having to do with the sky.

Mi is one of the two phonetic characters for the Tangut imperial surname. THE GOLDEN GUIDE: LINE 40: TANGRAPHS 196-200

40: Subject (196-197) + Object (198 199) + Verb (200)

Tangraph number 196 197 198 199 200
Li Fanwen number 3753 4884 2806 4902 0182
My reconstructed pronunciation 2kõ 2ni 2zəuʳ 1ŋwəəu 1vɨạ
Tangraph gloss sovereign; government official; public suffix 'person' imperial edict speech; word to allow; to say
Word ruler or government official?
Translation The ruler (or official?) pronounces the words of an imperial edict.

196: 3753 consists of 'sage' (geo) plus 'hand' (pik).

Kychanov (2006: 455) translated 3753 as 'sovereign' and 3753-4884 as 'emperor', whereas Li Fanwen (2008: 605-606) translated 3753 as 'government official' and 'public' and 3753-4884 as  'government official'.

Perhaps both are right. 2kõ could be a merger of an old loan from Middle Chinese 公 *koŋ 'duke; public' and a newer loan from Tangut period northwestern Chinese 官 *kõ 'government officer' (cf. the later Hphags-pa Chinese reading gon [kɔn]).

197: 4884 consists of a 'horned hat' (function unknown), 'temperament' (why?), and a member of the large ヒ family of radicals. I have no idea why they add up to 'person' (or if one prefers to translate a suffix as a suffix, '-er').

198: I have no idea what the left radical (gii 'waist; bird; sun ...') is doing in 2806. The right radical (yos) is a phonetic for 2zuʳ and appears as a semantic element in 2805 'to command' and 5459 'supervise'.

2806 as a whole combines with 1586 'sound' to form 0774 'om', even though 'om' is has nothing to do with imperial edicts and does not sound like 2806 2zuʳ.


I would expect 0774 to be pronounced ʔõõ, the closest possible match to Sanskrit om [oom], but there is no Grade I long nasal -õõ in my reconstruction (though Grade II -ɔ̃ɔ̃, Grade III -ɨõõ, and Grade IV -iõõ exist). Tangut does have short -õ, but not before a glottal stop.

Gong reconstructed 0774 as 0a (0 = unknown tone), implying that the Tangut borrowed 'om' from Middle Chinese 唵 *ʔam or a later *ʔã rather than from Sanskrit (or Written Tibetan om).

The fanqie for 0774 is

1xi + a

implying 0xa with x- and without a nasal vowel. I can't believe the Tangut would not try to pronounce om as accurately as possible. What's going on here?

199: 4902 has an analysis with three source tangraphs:


4902 1ŋwəəu 'speech; word' (biodexbelcin) =

4980 1tʃhɨeew 'to speak; to say' (biofoubaebil) +

3195 2gwi 'word; phrase' (dexbaxbaecem) +

1593 1thie 'dagger' (cuotexbilbelcin)

How is anyone supposed to figure out that belcin in 4902 is from 1593 as opposed to 29 other tangraphs, and what does 'dagger' have to do with 'speech'? A more relevant tangraph with belcin is

1487 1tʃhɨi 'to advise; to urge' (boabaxbelcin)

which has 'language' (boa) on the left ... unlike 4902 or 4980. 3195 has baxbaecem, a variant of cun, the right-hand variant of boa. Why are semantic radicals used so inconsistently?

200: 0182 is another language-related tangraph without 'language' (boa), though it does contain 'sound' (bos) and a speech radical of some sort (qin):


0182 1vɨạ 'to allow; to say' (bosqin) =

1586 1ɣɪ̣ 'sound' (bostal) +

4520 2bi 'command; directive' (bukqin; buk = 'mouth') THE GOLDEN GUIDE: LINE 39: TANGRAPHS 191-195

39: Subject (191) + Object (192) + Location (193) + Verb 1 (194) + Verb 2 (195)

193 and 194 may be a disyllabic verb 'report before (the emperor)'.

Tangraph number 191 192 193 194 195
Li Fanwen number 2384 2492 1778 5402 5026
My reconstructed pronunciation 2mieʳ 2dza 2ʔiu 2khie 1mi
Tangraph gloss minister to survey; to measure; plan front; forward; before; sovereign to reach; to arrive; to report to hear
Translation Ministers report their plans before (the emperor) and listen (to him).

191: Compare 2384 ('hand'+ 'person') with 3508 in line 32.

192: 2492 has a radical with an unknown meaning (if any) between two 'people'. It could have an analysis like


2492 2dza 'to survey; to measure; plan' (dexbuldex) =

2785 2tew 'scheme, strategy' (dexbuldumcin; semantic) +

2549 1dza 'miscellaneous; mixed' (jerdex; phonetic) < Chn 雜

though 'person' would be practically useless as a phonetic since it is in 1,187 tangraphs with all sorts of readings.

Another possible source of the right radical (and center radical?) is

0293 2dzəi 'to consider' (buldexcok or buldexcok)

193: The left side of 1778 ('light' in Grinstead 1972: 15) could be from

0567 1riʳ 'front'

The right side of 1778 appears in tangraphs for 'to look' and could be from Chinese 看 'to look', but it may be in 1778 because it also vaguely resembles Chinese 前 'front' minus刂.

194: 5402 has a rare left radical (teo) that I will look into later. The right radical (jor) is six times more common.

Adding 'person' to 5402 results in a tangraph for a partial synonym


5402 2khie 'to arrive; to report' > 2679 2niə 'to arrive' (but not 'to report')

195: 5026 is the only tangraph in this line with a known analysis:


5026 1mi 'to hear' (biopokdex) =

4681 1niu 'ear' (biobaebescin; semantic) +

0615 2dweʳw 'consciousness; to smell' (pokdex; cryptosemantic)

I call 0615 'cryptosemantic' since its semantic function is only clear if one knows about the double meaning of one of its Chinese translations: 聞 'to smell; to hear'. THE GOLDEN GUIDE: LINE 38: TANGRAPHS 186-190

38: Subject (186-187) + Location (188-189) + Verb (190)

(Less likely: a long compound noun: 'crown prince tower game'.)

Tangraph number 186 187 188 189 190
Li Fanwen number 4456/4457 0276 2829 4440 1097
My reconstructed pronunciation 2tha / 2liẹ 2no 1dʊ 2riẽ 2ʔəu
Tangraph gloss great; large child storied building building with two or more stories to jump; recreation; game; theatrical performance; to play
Word crown prince tower
Translation The crown prince plays in the tower.

186: One of a few tangraphs with two readings. See line 32.

186-187 'great child' is a calque of Chinese 太子 'crown prince'.

187: fus, the right radical of 0276, is a phonetic for 2no.

0276 and the previous tangraph in this line combine to form 2628:


2628 1gooʳ 'man; male' (giigir) =

4456 2tha /4457 2liẹ 'large' (wungii) +

0276 2no 'child' (fougirfus)

188: 2829 and 2865 are the only two graphs in which the right radical heu is a phonetic for 1dʊ:


2829 1dʊ 'storied building' (gaiheu) =

1890 2bie 'high' (semantic; gai)

2865 1dʊ 'peaceful' (phonetic; dexjolheu)

189: 4440 2riẽ belongs to a small category of r-syllables with nonretroflex, nasal rhymes. (The vast majority of r-syllables contain retroflex rhymes.) I'd like to investigate this category later.

4440 looks like the mirror image of 2852 and is derived from it:


4440 2riẽ 'building with two or more stories' (bordex)

2852 1tha 'Buddha' (dexbor) < northwestern Chinese 陀 *tha < 佛陀

2862 1ni 'family; house' (dextiiheu)

I suspect that the 丰 bor in 4440 is intended to resemble a tall building in 4440 (or the 婁 of Chinese 樓 'storied building') whereas the丰 bor in 2852 is based on 弗 from Chinese 佛 'Buddha'.

190: I couldn't find 'to play' for 1097 in Kychanov (2006) or Li Fanwen (2008), but I thought that would be a more appropriate translation than 'to jump'. I later discovered that Nishida (1996: 386) also translated 1097 as 遊ぶ 'to play'.

Why is part of 'meaning' in 1097?


1097 2ʔəu 'to play'? (duubaehom) =

0797 1phi 'meaning; idea' (beeduucin) +

3306 2miu 'to dance' (gerbaehom) < Chinese 舞 TANGRAPHIC RADICALS 8: HES FEL-LEN MASTER?

The radical fel resembling 永 'eternal' from the last tangraph of line 37 of the Golden Guide only appears in two other tangraphs:

Tangraph Li Fanwen number Reading Gloss Notes

1151 2ʔo leisure; free time variant of 1296 with left radical hes, a variant of her; hes is only in this tangraph; was the long leg of hes by analogy with the long leg of dia in 3099 below?

1296 Nishida (1966: 244) defined the left radical her as 'tilt', but it appears in 228 other tangraphs with varying semantics

3099 1dʒɨee to live; to stay; to exist I think of the left radical dia as the long left legged version of dex 'person'*.

One sinograph corresponding to 3099 is 住, a combination of 亻 'person' and 主 'master'. If dia corresponds to 'person', could fel be a distortion of 'master'? Or is 'person' in 3099 and 住 just a coincidence? 'Person' is so common in tangraphs that its presence is almost meaningless and I would expect many coincidences with sinography.

Could fel in 1151/1296 be a distortion of the 木 'tree' on the right side of the sinograph 休 'rest'? The 亻 'person' in 休 'rest' is also in 傾 'tilt', which could have been converted into the 'tilt' radical her/hes in 1151/1296.

*The analysis of 2495 implies that dia is equivalent to dex:


2495 1dʊ 'to exist; to have; to place' (dexjoldex) =

left and center of 2865 1dʊ 'peaceful' (phonetic; dexjolheu) +

left of 3098 2diẽ 'to stop; to cease' (diajix) THE GOLDEN GUIDE: LINE 37: TANGRAPHS 181-185

37: Subjects (181, 182) + Location (183-184) + Verb (185)

Tangraph number 181 182 183 184 185
Li Fanwen number 0510 4447 4204 1892 3099
My reconstructed pronunciation 1ŋwəʳ 1dzõ 2kəụ 1mii 1dʒɨee
Tangraph gloss emperor; heaven empress central room house; residence to live; to stay; to exist
Word inner court
Translation The emperor and empress live in the inner court.

181: 0510 is a source of 1623 in line 35. It's a phonetic-semantic compound:


0510 1ŋwəʳ 'emperor; heaven' (qesgirdex) =

left of 0654 2ŋwəʳ 'dark green' (qesgir; phonetic) +

left of 3950 1tshwiu '☰ heaven trigram' (girdexgie; semantic)

182: 4447 looks like half of 'big' from plus 'bird'. The right side is phonetic:


4447 1dzõ 'empress' (wunyul) =

left of 4456 2tha /4457 2liẹ 'large' (wungii; semantic) +

right of 2584 1dzõ 'bridge' (ciryul; phonetic)

2584 in turn has 4447 in its derivation:


2584 1dzõ 'bridge' (ciryul) =

left ('water') of 3058 2ziəəʳ 'water' (cir) +

right of 4447 1dzõ 'empress' (wunyul)

I think the real source of the phonetics of 2584 and 4447 is

2262 dʒwɨõ 'bird' (giigirwur)

with a dot added atop wur (to signify a different initial?)

If dʒ-tangraphs can serve as sources of phonetics in dz-tangraphs, can tʃ(h)-tangraphs serve as sources of phonetics in ts(h)-tangraphs?

183: 4204 is derived from 4396 which was in line 36.


4204 2kəụ 'central room' (boxcirpok) =

top ('wood') of 4396 2riəʳ 'hall' (boxhasgii; semantic) +

all of 3085 2kəụ 'to gargle; to rinse' (cirpok; phonetic)

184: 1892 is the other source of 1623 in line 35. (See 181 for the other half.) Guess what's in its Tangraphic Sea analysis ...


1892 1mii 'house' (bexpeucok) =

right ('house' radical + ?) of 1623 2vạ 'palace' (qesbexpeu) +

right of 0289 2vəi 'wall; city; castle' (beldexcok)

185: 3099 has an analysis that makes no sense.


3099 1dʒɨee 'to live' (diafel) =

left of 3098 2diẽ 'to stop; to cease' (diajix) < loan from Late Middle Chinese 停 *deŋ

right of 1296 2ʔo 'leisure; free time' (herfel)

3099 has a rare radical fel that I'll examine tomorrow night. It's long reminded me of 承 'receive from above', though now I see it also resembles 永 'eternal'. THE GOLDEN GUIDE: LINE 36: TANGRAPHS 176-180

36: Location 1 (176-177) + Location (178-179) + Verb (180) with an implied subject.

Tangraph number 176 177 178 179 180
Li Fanwen number 3406 4396 3294 3819 2396
My reconstructed pronunciation 2piụ 2riəʳ 2məəi 2ləu 2dzəəu
Tangraph gloss palace room; hall; main building wise seat; place; official title to sit; to reside
Word imperial throne
Translation .The emperor sits on the imperial throne in the palace hall.

176: See here.

177: 4396 is a straightforward semantic-phonetic compound:


4396 2riəʳ 'hall' (boxhasgii) =

'wood' from the top of 4204 2kəụ 'central room' (semantic; boxcirpok) +

all of 1453 2riəʳ 'besides' (phonetic; hasgii)

178: I doubt 3294 was derived from one of its probable derivatives:


3294 2məəi 'wise' (giibil) =

left of 2449 2bəi 'sun' (semantic and phonetic?; giigeo - what is gii doing in 'besides' above?)

was the b- of 'sun' [mb] when 3294 was created?

right of 3416 2məəi 'relief; terrain' (balgiibil)

3294 is phonetic in 3416; the meanings of bal and bil, if any, are unknown

179: The analysis of 3819 is unknown. It contains a radical

alphacode cue

reminiscent of the 广 of 座 'seat', The radical cue is also in

3830 2nie 'king' (see line 32).

180: Gong proposed that 2396 is a calque of Chinese 坐 'to sit' with the Tangut 'person' radical instead of Chinese 人 and Tangut 'bottom' instead of Chinese 土 'earth'. That is a more likely origin than the Precious Rhymes analysis:


2396 2dzəəu 'to sit' = 2994 2vaʳ 'to gather; numerous; to sit' minus its top (cou)

The Combined Homophones and Tangraphic Sea analysis of 2994 confirms that its botttom is from 2396:


2994 = top of 3015 2lhiụ 'to capture; to gain' + all of 2396

Gathering involves capturing and gaining but being numerous and sitting do not, so I wonder if 2994 represents three unrelated homophones.

Although 2dzəəu 'to sit' has the same initial as Middle Chinese 坐 *dzwaʔ 'to sit', the rhymes don't match very well, so the shared initials may be a coincidence. THE GOLDEN GUIDE: LINE 35: TANGRAPHS 171-175

From now on, I'll diagram the structure of each line:

35: Subject (171 + 172) + Object (173 + 174) + Verb (175):

Tangraph number 171 172 173 174 175
Li Fanwen number 3372 1623 2544 4573 2634
My reconstructed pronunciation 2kəụ 2vạ 2ʃɨẽ 1bi 1dʒwɨõ
Tangraph gloss inside palace sage light; brightness to publicize; to propagate
Translation The inner palace propagates the light of the sages,

171: Notice how the source of the left side of 3372 doesn't quite match - it has an extra vertical stroke:


3372 2kəụ 'inside' (dexceubilpok) =

3215 1kee 'house; room'(dexceubaedim; dim = bil + Alan Downes' alphacode 'se') +

2983 2ʔəu 'locative suffix' (ciapok)

There is no tangraph with the same left side (alphacode dexceubil) as 3372.

172: The left side of 1623 (alphacode qes; resembling 'eight' with an added horizontal stroke on top) has an 'imperial' meaning here and elsewhere, though it may also refer to bodies (e.g., in 1042; compare its shape to Chinese 身 'body') and have other functions.


1623 2vạ 'palace' (qesbexpeu) =

0510 1ŋwəʳ 'emperor' (qesgirdex) +

1892 1mii 'house; residence' (bexpeucok)

Nishida identifiedㅑ (alphacode bex) as 'house' but has no glosses for .qes or peu (the right side of 3372).

173: 2544 has a dubious analysis:


2544 = 0480 1lha 'sage' (bolgeo) without its left side

(1lha may be cognate to Written Tibetan lha 'god'; the different semantics make me doubt borrowing)

See "Tangraphic Radicals 7" for a different origin proposal.

174: What is 'eight' doing in 4573, 4645, 3226, and 3228? It seems to do double duty as 'light'.


4573 1bi 'light' (jeudil) =

4645 2riẽ 'radiance' (jeujuo)

3226 1niəə 'bask; to shine' (diljeu; reverse of 3226; cognate to 2440 2niəə 'sun; day')


4645 2riẽ 'radiance' (jeujuo) =

4573 1bi 'light' (jeudil) +

5120 1swew 'bright' (dosbou; note how the ソ of dos is joined to bou to form juo)


3226 1niəə 'to shine' (diljeu)

4573 1bi 'light' (jeudil) +

3228 1thioo 'beautiful; wonderful' (dexjeu)


3228 1thioo 'beautiful; wonderful' (dexjeu) =

0541 2ʃwɨo 'dignified' (bumdex) +

2665 2məi 'shining; beautiful' (dexjeudil, probably analyzed as 3288 + 4573, taking us back to the start of the chain; could 2məi be a loan from Chinese 美 'beautiful'?)

175: Notice how the right side of 'to rise' replaces 'sage' in 'name'. This begs the question of what 'sage' is doing in 'name'.


2634 1dʒwɨõ 'to propagate' (dexbaswarbae) =

2639 2miee 'name' (dexbasgeobae; cognate to Chinese 名) +

3678 1to 'to be born; to rise' (dexwar)

To propagate something is to raise a name (cf. Chinese 揚名 'raise name' = 'become famous').

6.22.8:08: David Boxenhorn's Tangut Search enables me to find all nine tangraphs with the structure bas...bae. The structures of 2637, 2638, 2639, 3834, and 4570 are partly or wholly opaque to me.

Tangraph Li Fanwen number Reading Gloss Notes

2631 2nii what < 2639 'name' + 'not' - i.e., nameless

2633 2la foolish; stupid 'person' + 'evil' (Nishida 1966: 245); from 'name' + 'evil' - the foolish have bad names (reputations)?

2634 1dʒwɨõ to publicize; to propagate < 'name' + 'rise'; see above

2637 1dəəu slave; servant 'person' (from 3168 'woman') + 2639 (which isn't homophonous; why?)

2638 2miee name 'person', 'tiger' < Chn 乕 (what does this abbreviate?); why two tangraphs for 2miee 'name'?

2639 'person', 'sage' (why?)

3834 2lhew to catch fire 'hand'; why is there no 'fire'?

3836 2thie transcription tangraph; based on mirror image of Chinese 定?; 2833 in line 24 is also based on 定, but looks different and reflects an earlier Chinese pronunciation with *d- which had not yet shifted to *th-

4570 scheme; stratagem only bas...bas tangraph with 'eight' on left instead of 'person'; < 'eight' from 3228 'wonderful' (see above; why?) + phonetic 3836 TANGRAPHIC RADICALS 7: ARE SAGES EAR-IE?

One of the first tangraphs I learned 14 years ago was

2544 2ʃɨẽ 'sage' < Chinese 聖 (alphacode geo)

It undoubtedly represents a loanword from Chinese, but it looks nothing like 聖. For years I assumed it had something to do with

1886 2dzwio 'person' (alphacode dex)

but why should the last stroke of 'person' be arbitrarily changed, and what does the extra vertical line represent?

Last night, I realized that 'sage' might be from 耳 'ear', the top left component of 聖 'sage':

- minus the top horizontal stroke and the left vertical stroke

- with the three remaining nonvertical strokes separated from the right vertical stroke

- with a short vertical stroke added to the bottom of the third nonvertical stroke

I vaguely recall once proposing - either on this blog or in my own notes - that 'sage' might be from this variant of 聖 (䎴 = 耳 + 井)with 井 simplified to a single vertical stroke. Now I wonder if 井 may not be relevant.

I don't see any plausible derivation from any of the other 55 variants of 聖 without subtracting components.

In any case, the tangraph for 'sage' is simpler than its sinographic counterpart in spite of tangraphy's reputation for complexity.

6.21.21:57: Notice that 'sage' is at the bottom of

1546 2lɨụ 'body (alphacode bixgeo; see line 34 of the Golden Guide)

corresponding to Chinese 身 'body' ... resembling 耳 'ear', my proposed source of 'sage'.

6.22.0:22: Notice how some variants of 身 'body' have a 亠 radical separate from the rest of the character. The bix radical atop 1546 may be based on a variant with 亠. THE GOLDEN GUIDE: LINE 34: TANGRAPHS 166-170

I meant to do a record three lines in one day, but oh well ...

Line 34 doesn't parallel line 33 in any way apart from shared subject matter (clothing).

Tangraph number 166 167 168 169 170
Li Fanwen number 1546 4729 1141 5977 3782
My reconstructed pronunciation 2lɨụ 1lhwəu 2kiʳ 2zeew 2zəu
Tangraph gloss body clothing waist official clothes; uniform; cape; cloak; overcoat; belt to tie up; to wrap around; shackles; fetters
Translation Wrapping clothing around the body and tying belts around the waist.

166: 1546 has a typical dubious analysis:


1546 2lɨụ 'body (alphacode bixgeo) =

1245 1ʔie 'self' (alphacode bixbeedoe; contains bee 'volition' [see 169]) +

1042 2giu 'body' (alphacode qeswek; wek = bixgeo)

I assume was created after 1546, not the other way around.

Could 1546 be based on Chinese 身 'body'? I doubt 'sage' on the bottom is relevant.

I wonder if the 'self' in the analysis is influenced by the fact that 身 contains a shape similar to 自 'self'.

2liụ < *S-luH 'body' is cognate to Written Tibetan lus 'body'.

167: 4729 looks like a 'horned hat' atop

0847 1ʃɛ̣ 'lameness' (with 'horse' on the left)

but its analysis gives a completely different and at least partly unlikely origin:


4729 1lhwəu 'clothing' (alphacode biohinheu) =

4838 2phə 'brown shirt' (alphacode biohinqae)

I suspect this is based on 4729, not the other way around) +

4898 2ziʳ 'armor' (alphacode texdexdaiheu)

168: I wonder if 1141 and/or its bottom component

3087 1dʒɨew 'waist'

are based on the right side of 腰 'waist'. Li Fanwen (2008: 191) lists the following analysis in parentheses from the Combined Homophones and Tangraphic Sea:


1141 2kiʳ 'waist' (alphacode baxgii) =

1178 1mia 'end' (alphacode baxbelgus; how is this relevant to 'waist'? I'm tempted to regard this as a loan from Middle Chinese 末 *mat plus a lost high vowel prefix: *Cɯ-mat > *Cɯ-miat > 1mia)

3087 1dʒɨew 'waist' (alphacode gii)

I assume that the parentheses indicate a reconstruction based on surviving parts of tangraphs.

169: I think Nie and Shi's translation 'belt' for 5977 is the most likely object of 'to tie'. But maybe the Tangut also spoke of tying (i.e., wearing) uniforms and coats.

I have no idea what the left half

alphacode pum

of 5977 means, if anything. It is a phonetic with three types of sound values: zeew, ta(ʳ)/təʳ, tiọ.

(6.21.1:51: pum looks like a stacking of


alphacodes buu + bee

'feces' (Nishida 1966: 242) + 'volition' (Nishida 1966: 244), 'temperament' (Grinstead 1972: 15)

but may not have a nongraphic (i.e., a semantic or phonetic relationship) to either.

Nishida (1966: 244) identified the right half

alphacode pai

as 'finger'. No analysis is known.

170: I presume the noun 'shackles, fetters' is derived from the broader verb 'to tie up': 'that which one is tied with'.

6.21.1:36: The alphacode ger for the left side of 3782 corresponds to dex + bae.


3782 2zəu 'to tie up' (alphacode gerpas) =

5876 2kəụ 'to tie up' (semantic; 'tree' on left; alphacode tarbaedex) +

3757 2dziew 'wooden handcuffs' (semantic/phonetic; alphacode baepas)

3782 and 3757 may share a common root *dz-w. The initial z- of 3782 may be a *dz- that lenited after a lost prefix.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2010 Amritavision