The three tangraphs

TT3543 riẽ or l(d)iẽ R43 2.37 'three; third'

TT3544 tswəʳ R90 1.84 'to milk'

TT3306 niạ < *sɯ-na R67 1.64 'spirit; spirit-guardian; divinity'

cf. Written Burmese nat 'spirit'

are difficult to distinguish. Only their left-hand components are different:

Li Fanwen radical 119 (meaning unknown)

Li Fanwen radical 118 'person'

Li Fanwen radical 200 'holy'

There is no way to guess the meaning of TT3544 'to milk' by looking at it. It seems to consist of two 'people' around a vertical line and it lacks 'liquid', 'breast', etc. But according to Tangraphic Sea, the 'person' on the left is an abbreviation for 'breast':


TT3544 tswəʳ R90 1.84 'to milk' =

left of TT3649 new R44 1.43 'breast' +

looks like 'person' + ('meat' + 'ornament'); the latter combination is not an independent tangraph

left of TT1545 lạ R66 1.63 'hand' (used when milking?) +

right of TT3405 thọ R73 2.62 'grow up' (why?)

resembling two 'people' surrounding Li Fanwen radical xxx 'vexation' (< right side of Chn 煩 'vexed'?)

'Spirit' looks like 'holy' + 'person', though the latter component is an abbreviation for 'to protect, defend, support' according to Tangraphic Sea:

I cannot believe that tangraphic users could keep track of what 'person' represented in a thousand tangraphs.

I have no idea what 'person' or any of the other parts of TT3543 'three' signify.

Li Fanwen (1997: 1134) lists five tangraphs other than TT3543 with radical 119 on the left, and I found two others:

Tangraph Li Fanwen 1997 number My reconstruction Rhyme Tone.rhyme Li Fanwen 1997 gloss; notes
2459 ʃiõ R58 1.56 base/mean person; radical 118 'person' on right
2460 R28 1.27 surname and transcription tangraph; radical 199 'beast' on right
2646 da R17 2.14 surname tangraph; the right side looks like 'bird/waist' with a dot; this radical (Kychanov B235) is in only three tangraphs
3664 tʃɨə R30 1.29 first syllable of a surname; radical 139 'finger' on right
3665 lhiạ R67 1.64 sparkle; glisten; lightning; also first half of disyllabic word lhiạ tshwiu 'lightning'; (see below); radical 367 'lightning' on right
3642 tshwiu R3 1.3 second half of disyllabic word lhiạ tshwiu 'lightning' (see above)
0121 də̣i R68 1.65 god; Kychanov 2006: divinity of the Earth; the great goddess mother Earth; 'top' on top; resemblance to Middle Chinese 地 *di 'earth' coincidental
(10.19.20:26: Why would the name of an earth goddess be written with 'top' instead of 'earth' or 'bottom'?)
4953siR112.10god; deity; the function of radical 036 'horned hat' on top of 'lightning' (see above) is unknown

(10.19.1:09: Added LFW3642 which is not in Li Fanwen's list of radical 119 tangraphs. TT0121 is listed under radical 188 'top'.)

(10.19.20:21: Added LFW4953, also not in Li Fanwen's list of radical 119 tangraphs.)

Three represent syllables of surnames and two represent 'lightning', but the rest have no shared semantics and none have shared phonetic features.

Since Kychanov (2006) does not include Li Fanwen radical 119 as a right-hand radical, that element may have been interfiled with his radical B210 'person' (= Li Fanwen radical 118). It's also possible that radical 119 never appears in non-left positions, though I am not sure since I can barely distinguish it from 'person'.

Next: Impersonation. LI FANWEN RADICAL 295 / KYCHANOV RADICAL B082

best known as the left side of

TT1718 sọ R73 1.70 'three'

has no known semantic function, but it does seem to have multiple phonetic values:

Left and center positions Both left and right-hand positions Center position All positions Center and right positions Right-hand position
KU (is ʔəụ a subtype?; cf. velar-glottal alternations in Chinese), lA wA n + nonlow vowels PA rI

Capital letters signify sound classes. A may represent achromatic vowels (a, ə) or *a brightened to various degrees (a ~ ie ~ i). wA may ultimately be a subset of the PA group with initial lenition due to prefixation: *CV-PA > wA.

Some values are unique to one position.

In left-hand position (complete list?)

Li Fanwen 1997 number My reconstruction Rhyme Tone.rhyme Li Fanwen 1997 gloss; notes
5820 ʔəụ < *?q- R61 1.58 cover; shelter; rely on
5890 R4 1.4 loose ('pine' is an error based on 松 'pine', also used to write 鬆 'loose')
5930 2.4 princess
5889 R28 2.25 steal; pilfer; rob; cf. Old Chinese 偷 *hlo 'steal' (but vowels don't match)
5923 1.27 first syllable of the surname ləŋəu
5887 nee R38 2.34 post, carry along
5931 noo R54 2.45 send; post; mail (cognate to 5887 above via mid vowel alternation)
5865 sọ R73 1.70 three
5920 tsə R28 1.27 transliteration tangraphs for Tangut period NW Chinese *tsz̩; cf. premodern Sino-Korean for this Late Middle Chinese rhyme category (though SK was based on an eastern LMC dialect predating TPNWC)
5974 wiẹ R64 2.54 send; dispatch (cf. semantics of noncognate 5887 and 5931 above)

In center position (incomplete list; there is no convenient way to find all tangraphs with the same central element)

Li Fanwen 1997 number My reconstruction Rhyme Tone.rhyme Li Fanwen 1997 gloss; notes
0735 dʒie R37 1.36 cool
1897 pha R17 1.17 follow
2365 difference
1898 kiəʳ R92 1.86 to fly
2818 khə̣ R77 1.68 to rock; sway; shake
2101 bəuʳ R80 1.75 gather; collect
2756 bəiʳ R82 1.77 獎賜, 奉賜; Kychanov: 'grant, bestow'
2551 lɨa R19 1.19 hide; conceal; cf. Kychanov 0895 'secret'
2614 (a surname)
2668 kə̣u < ?*sʌ-ku R62 1.58 loose; cognate to 5890 above
3473 giu ? sleep
2745 zəiʳ R82 2.71 sunshade; umbrella
3221 niooʳ R103 1.95 abundant
3862 siə R31 2.28 fetters

In right-hand position (complete list?)

Sofronov 2006 number My reconstruction Rhyme Tone.rhyme Sofronov 2006 gloss; notes
0886 thwəʳ R90 1.84 leap; jump
0887 phiẹ R64 1.61 curved; slanting; crooked; slant
0891 phie R37 1.36 rebel; escape; evade
0889 bi R11 2.10 escape; avoid; leave a place (cognate to 0889?)
0888 xɨə R30 2.27 callous; heartless; inanimate; trouble; disaster; talk in one's sleep
0890 diaa R24 1.23 transcription sign; a fanqie character (though B082 is taken from Kychanov 5055 lɨa R19 1.19 rather than an -iaa R24 tangraph!)
0892 waʳ R85 1.80 desperate situation; dire; defeated enemy
0902 wiạ R67 2.57 send; appoint; grant a divorce; surrender a town (cognate to Li Fanwen 5974 wiẹ R64 2.54 above)
0903 name of a plant
0893 ʃiõ R58 1.56 resistance; defense; the Chinese surname 尚
0894 rəʳi < *Cʌ-ri R82 2.71 tortoise shell; watchtower
0896 board; plank
0895 dwəəu R6 2.5 secret
0897 tsəu R1 2.1 arise; appear; gush
0898 tsạ R85 1.80 disorder; entangle; turn over escape; flee (cf. semantics of 0889 and 0891 above)
0899 jieʳ < *Cɯ-re R79 2.68 kill; put away
0900 a surname
0901 jiʳ < *ri R84 2.72 kill (cognate to 0898 above)
0904 ? a kind of bird
0905 nie R37 2.33 go over; the other way around; hide


TT1718 sọ R73 1.70 'three'

be a phonetic representation of a disyllabic Tangut B word for 'three': e.g., something like


(The possible values of the first syllable are taken from the recurring phonetic values of Li Fanwen radical 295 in the tables above. Presumably the value in 'three' is the same as in the Tangut B reading of 'fetters'. The value of the second syllable is inferred from the various tangraphs sharing their right sides with 'three', but that component could have had a second, as yet unknown, Tangut B sound value.)

No combination of any of the known values of radical 295 with dzɨw sounds anything like the words for 'three' in the region:

Uyghur üq [ytʃ]

Written Tibetan gsum

Tangut period NW Chinese *sã < Middle Chinese *sam

Does this mean that the disyllabic interpretation of 'three' is wrong? Or does it mean that Tangut B had a word for 'three' unlike its neighbors? Attested Tangut had three such words which I will examine later.

David Boxenhorn has suggested that tangraphs represent abbreviations of mnemonic phrases. Thus the Tangraphic Sea analysis


TT1718 sọ R73 1.70 'three' =

center of TT2981 siə R31 2.28 'fetters' +

right of TT2631 dzɨw R46 2.40 'width'

might tell us that the phrase for 'three' contained one word from the phrase for 'fetters' and another word from the phrase for 'width'. What would these words have been? I have no idea. If David is right, the phrases could have been riddles whose wording may never be reconstructed. THE WIDTH OF FETTERS

Guillaume Jacques asked me if I had any ideas about the relationship between

TT1718 sọ < ?*s(ʌ)-som R73 1.70 'three'

cf. Written Tibetan gsum 'three', Written Burmese suṃḥ, Old Chinese *səm or *s-lhəm

and two tangraphs sharing its right side (Kychanov radical B202):

TT2458 dzɨw R46 2.40 'glue; to link'

with Li Fanwen radical 216 'nose' (why?) on the left

TT2631 dzɨw R46 2.40 'width'

with Li Fanwen radical 139 'finger' (used to measure width?) on the left

I know of two more tangraphs with B202:

TT5313 dzɨw R46 2.40 'miserly; frugal'

with TT2458 dzɨw R46 2.40 'glue; to link' as phonetic

beneath Li Fanwen radical 185 'cherish' (< top right of 惜 'cherish'?) on top

TT2586 tʃhwæ R18 1.18 'iron arrow with three tips; triangular arrow head'

with Li Fanwen radical 293 'metal' on top

and Li Fanwen radical 085 'steep' on the right

(Is this the only tangraph with B202 on the left side?)

Are there any other tangraphs with B202?

B202 is clearly a phonetic symbol for dzɨw R46 2.40. But I am hesistant to regard it as a semantic symbol for 'three', even though it appears in 'iron arrow with three tips'. It does not appear in the other tangraphs for 'three' or 'third':

TT3543 riẽ or l(d)iẽ 'three; third'


TT0870 lheʳ or reʳ (giu) 'three; third'

(I'll discuss the uncertain reconstruction of these words later.)

So for now I can only say that B202 is an abbreviation for sọ 'three' in 'iron arrow with three tips'.

The Tangraphic Sea analysis of TT1718 sọ 'three' makes no sense:


TT1718 sọ R73 1.70 'three' =

center of TT2981 siə R31 2.28 'fetters' +

right of TT2631 dzɨw R46 2.40 'width'

The function of the left side of TT1718 (Li Fanwen radical 295) is unknown.

(10.17.1:27: The irrationality of such analyses made me sympathetic toward the Tangut B hypothesis. Is there any language in which 'three' is phonetically similar to 'glue', 'width', or 'miserly'?)

I've been baffled by TT1718 ever since I memorized it back in 1996. Is the fact that all tangraphs with B202 on the right

are in the Mixed Categories volume of Tangraphic Sea a key to the mystery?

10.17.1:23: I would expect the three dz-initial B202 tangraphs to be in Mixed Categories, since all (?) dz-initial tangraphs were put into that volume for some unknown reason. However, it is not clear why some non-dz-initial tangraphs (e.g, TT1718 sọ) are in that volume and others aren't.

The fanqie for TT1718 sọ is

TT3437 s R70 1.67 + TT3228 tḥ R73 1.70

Both spellers are listed in the level tone volume of Tangraphic Sea. If Mixed Categories tangraph readings had a unique phonetic characteristic, why would the reading of TT1718 sọ be indicated with non-Mixed Categories tangraphs? MC tangraphs also don't seem to share any unique graphic characteristic. Could MC just be a volume of leftovers? That is unlikely since I cannot imagine how all dz-tangraphs - and high-frequency tangraphs like TT1718 sọ 'three' - could have been accidentally excluded from the first two volumes. The rationale behind MC is as obscure as the structure of TT1718. WHEN I + I = I (OR刂?)

Last night, I proposed "a constraint against two adjacent vertical lines in the same tangraph." I should have specified that I meant two adjacent vertical lines which

- are both the same height

- result from the juxtaposition of two separate elements: e.g.,


Li Fanwen radical 200 'holy' + the left side of TT1244 bii R14 2.12 'common people'

- are not parts of a more complex radical: e.g.,

Li Fanwen radical 079 'sound'

Obviously there are many tangraphs with

Li Fanwen radical 023 'not'

but note that its first vertical stroke is slightly shorter than its second. 'Not' is not identical to a Roman numeral II.

I suggested that two adjacent vertical lines (of the same height from different elements) merged into one in TT3316 'Tangut':


Guillaume Jacques interpreted my proposal as a sort of graphic 'haplology'. He suggested that the vertical line in 'holy' is 'dissimiliated' (i.e., shortened) in

TT3722 miee R40 2.35 'name'

to distinguish it from the vertical line to its right. Although I am not certain what is going on in 'Tangut', I am confident that Guillaume is correct about 'name'.

No analysis of 'name' is known, but it could be a combination of


the frame of TT3724 miee R40 2.35 'praise' (phonetic? extended use of 'name'?) +

an unknown tangraph containing 'holy' (presumably semantic, but why?)

One could propose a new element that looks like 'holy' but has a shorter vertical line. However, I prefer to view the center bottom element of 'name' as a conditioned allograph of 'holy' instead of a distinct element with its own semantic and/or phonetic value.

Four other tangraphs which appear to have the same frame as 'name' and 'praise' are not pronounced miee:


TT3719 dʒwiõ R58 1.56 'publicize; propagate; declare; spread; to name' =

frame of TT3722 miee R40 2.35 'name' (semantic) +

right of TT3667 to R51 2.42 'be born; rise' (semantic)

TT3704 nii R14 2.12 'someone; a certain person'

with the frame surrounding Li Fanwen radical 148 'absence'

'name' + 'absence' = 'an undetermined person'?

TT3721 l(w)a R17 2.14 'to disgrace'

with the frame surrounding Li Fanwen radical 267 'evil'

'name' + 'evil' = 'disgrace'?


TT3696 dəəu R5 1.5 'slave; servant' =

left of TT3807 miəə R33 1.32 'woman; female' +

cf. 女 'woman' on the left side of Chn 奴 'slave'

all of TT2897 thie R37 2.33 (phonetic symbol), 'a word for counting saddles and teams of horses'

cf. the shape of Tangut period NW Chinese 定 *thjẽ

The first three tangraphs ('publicize', 'someone', 'disgrace') are semantic compounds but thie is neither phonetic nor semantic in dəəu 'slave'.

There is a vague similarity between

thie and the phonetic dəəu (Li Fanwen radical 153; meaning unknown)

Replacing thie with radical 153 and reversing the order of elements results in

TT0359 dəəu R5 2.5 'angry'

'Angry' and 'slave' are both loanwords from Tangut period NW Chinese *ndu (with different tones in both languages). The Tangut mid vowel may have been conditioned by a Tangut low vowel presyllable:

*Cʌ-nduu > *Cʌ-dəəu > dəəu

Could dəəu 'slave' be a mirror-image derivative of its homophone dəəu 'angry'


but with additional distortion (why)? The only other mirror-image pairs I can think of are not homophones and lack drastic distortion: e.g.,


TT4105 dzwio R53 2.44 'person'

TT3769 niee R40 1.39 'heart'


TT0311 kiiʳ R101 2.86 'rob; steal'

TT1726 kwiəəʳ R100 1.92 'steal; pilfer' < ?*P-k-

(Gong placed R100 and R101 into separate rhyme groups, but I wonder if they should be in the same group. Are there any other instances of R100-101 alternation?) THE TANGUT: A HOLY PEOPLE

Tonight I realized that the left side of

TT1244 bii R14 2.12 'common people'

is also the right side of

TT3316 mi R11 2.10 'Tangut'

cf. Written Tibetan mi 'person'

Until now I had assumed that 'Tangut' was


Li Fanwen radical 200 'holy' +

Li Fanwen radical 054, second variant (meaning unknown)

but now I wonder if it is really


Li Fanwen radical 200 'holy' +

left of TT1244 bii R14 2.12 'common people'

with a vertical line shared by both radicals. (Perhaps such sharing is inevitable due to a constraint against two adjacent vertical lines in the same tangraph.)

My earlier analysis can still be salvaged if radical 054 was taken from the center of TT1244. (But then what would the vertical line [Li Fanwen radical 003] signify?) In either case, I think 'Tangut' and 'common people' are related tangraphs.

I suspect the analysis of 'common people' is circular:


TT1244 bii R14 2.12 'common people' =

(Li Fanwen radical 003 from unknown tangraph? +)

right of TT3316 mi R11 2.10 'Tangut' +

right of TT2329 sie R37 2.33 'people'

The only other tangraphs I can find with the left side of TT1244/right side of TT3316 are

TT3739 dʒɨə ?R30 or dʒɛ ?R35 'Tangut'

also the second half of

lhwiẹ R64 1.61 dʒɨə/dʒɛ ?R30/35 'Tangut'

obviously TT3316 'Tangut' with Li Fanwen radical 118 'person' added

TT0183 miuu R7 2.6 'ox'

'Ox' has no obvious semantic overlap with 'Tangut' or 'people', and I suspect its bottom consists of two radical 054s separated by radical 003 rather than radical 054 plus the left side of TT1244. PEOPLE WITHOUT PEOPLE


TT1244 bii R14 2.12 'common people'

a loan from northwestern Late Middle Chinese 民 *mbin 'people' (transcribed in Tibetan as Hbyin and Hbyïn)?

Vowel length may compensate for an earlier lost coda:

*bin > bii?

Cf. how Sofronov (1968 II: 136) reconstructed *-jeC for R14 in Tangraphic Sea Tangut.

Oddly, TT1244 does not contain

Li Fanwen radical 118 'person'

I have no idea what the components of TT1244 mean.

No other tangraph has its combination of left-hand elements ('person' + Li Fanwen radical 054) on the left.

Its left and right sides may be variations of the same element (a distortion of Chinese 民 'people'?).

TT2329 sie R37 2.33 'people'

could be the source of its right side. (The right sides of TT1244 and 2329 are not identical in the Mojikyo font, but they are the same in the editions of Homophones reproduced in Sofronov [1968 II: 106, 197].) BORN REPENTANT

While looking for a tangraph to translate 'birthday' in my second recent post for Sarah, I learned that

TT5000 lhi R11 2.10 '生 give birth to'

also means '悔 repentance'. Why? Its right-hand element

Li Fanwen radical 372 (meaning, if any, unknown)

is a phonetic for lhi, but what is the function of its left-hand element, Li Fanwen radical 324, which is also an independent tangraph

TT4990 dʒɨə R31 1.30 'skin'

There is no obvious connection between 'skin' and either 'give birth to' or 'repentance'.

And why is

Li Fanwen radical 118 'person'

added to 'give birth to/repentance' to create

TT lhi R11 2.10 'source'


Li Fanwen radical 196 'waist/bird'

instead of 'person' results in

TT5754 wị R70 1.67 'end'

which has nothing to do with waists or birds and does not even sound like lhi. Its Tangraphic Sea analysis makes no sense:


TT5754 wị R70 1.67 'end' =

left of TT5728 bəi R8 2.7 'sun' +

all of TT5000 lhi R11 2.10 'give birth to/repentance'

Even if bəi 'sun' is phonetic, how could 'give birth to/repentance' be semantic?

Placing 'person' on the other side of 'give birth to/repentance' results in

TT5001 lhi R11 2.10 'tranquil; peaceful; slowly'

What does a 'person' on the right have to do with peace?

And taking away 'skin' from 'tranquil' results in

TT1220 lhi R11 2.10 (second half of the surname

ŋwəu R1 2.1 lhi R11 2.10

with 'person' in each tangraph though in a different position)

Kychanov (2006: 433) also glosses TT1220 as 'small', but this gloss seems to be based on tsẽ R41 1.40 'small', another reading of the first half of ŋwəu lhi. ŋwəu/tsẽ is written as


'person' + 'small' (Li Fanwen radical 143)

and has the Tangraphic Sea analysis


TT3584 tsẽ R41 1.40 'small' =

right of TT5070 ziəəʳ R100 1.92 'few; little; tiny'

all of TT1730 tsiə R31 1.30 'small; little; too'

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