In "WOODen SON of PLUM", I proposed that the left side of

TT0971-972 (one tangraph with two readings)

BIG lyịy < ?*C-la 2.54 (native word; cognate to 大 OC *lats 'big'?)

BIG tha 2.14 (loan from 大 Tangut period NW Chn *tha)

was based on 李 Tangut period NW Chinese ?*li 'plum; a surname'.

One might expect

TT0962 shyu 1.2

to mean 'plum', but it actually means 'cool'. The TPNWC reading of 涼 was something like *lyÕ whose initial happens to match lyịy 2.54. (There is even a vague resemblance between TT0962 and the phonetic 京 of 涼. See the TRUE implications here.*)

The actual tangraph for PLUM is

TT0848 ner 2.71 (analysis unknown)

Like 李 'plum', it has an element meaning 'wood' on top. However, neither of the bottom two elements correspond to 子 'child'. The bottom left element is GRASS (Nishida r144) and the bottom right element is THREAD (Nishida r239).

Plant tangraphs usually only have WOOD or GRASS, not both. Since I don't think a plant can be both at the same time, one of the two has to be part of a phonetic - and it is. The bottom parts of TT0848 form

TT3277 萎 WITHER ner 2.71

which is homophonous with TT0848 PLUM ner 2.71. (When GRASS WITHERs, does it become like THREAD? THREAD is not obviously phonetic, as I cannot find any other tangraphs with it in right-hand position which are pronounced like ner.)

If it were not for TT3277, I would have guessed that THREAD was an attempt to imply 子 'child' by resembling the right side of 孫 'grandson'. 系 and THREAD both have more or less horizontal top strokes atop a ム shape atop lines pointing southwest and southeast. (The resemblance is probably not coincidental.**) But this highly contrived explanation would still not account for the presence of GRASS in PLUM.

*9.9.1:20: r169 with a 'horned hat' (Nishida r007) vaguely looks like 京 TPNWC ?*kyẽ. Is this why r169 might have been phonetic in this tangraphs for khyiy 1.36 and Giey 1.34?

TT2083 後代 ?LATER-GENERATION khyiy 1.36

TT2084 眾 NUMEROUS Giey 1.34


TT2086 (kind of animal) Giey 1.34

TT2087 TRUTH Giey 1.34

**The sinograph for 'thread' is 糸 which is similar to 系, the right side of 孫 'grandson'.  THREAD may be based on 糸. WOODEN SON OF PLUM

Here are more examples to show how easy it is to come up with (dubious?) sinographic origins for tangraphic elements. Why does

TT0971-972 (one tangraph with two readings)

BIG lyịy < ?*C-la 2.54 (native word; cognate to 大 OC *lats 'big'?)

BIG tha 2.14 (loan from 大 Tangut period NW Chn *tha)


TT0962 COOL shyu 1.2

on the left?

I just realized that COOL has the WOOD/TREE element (Nishida's r017) on top. Why? Because the whole tangraph is based on the sound and form (but not the meaning) of 蕭 Tangut period NW Chinese *syew 'mugwort'?

The question I can't answer is why

TT5649 WAIST jyiw 1.45

is on the right of BIG.

Last night, I saw a link between TT0962 and 李 Tangut period NW Chinese ?*li 'plum' (and the common surname 'Li') which is not far from Gong's reconstruction of the native Tangut for 'big', lyịy 2.54. I noticed that 子, the bottom of 李 'plum', is equivalent to Nishida's r171 CHILDREN, the left side of

TT1131 CHILD gyi 2.10 (Nishida's r215 [function unknown] is on the right)

(9.8.11:13: r171 vaguely resembles the right side of 孫 'grandson'. But almost anything can vaguely resemble anything!)

If one subtracts the top horizontal stroke from r171 CHILDREN, the result is r169. And just now I realized that 木 'wood', the top of 李 'plum', is equivalent to 卅 r017 WOOD. Summing up:

木 'tree' + 子 'child' = 李 ?*li 'plum'

卅 r017 TREE + r169 (= r171 CHILDREN minus one stroke) = phonetic of TT0971 BIG lyịy 2.54?

Although the phonetic of BIG and TT0962 COOL shyu 1.2 look alike, they may have two different similar-sounding source sinographs (李 ?*li 'plum' and 蕭 ?*syew).


In "The COOLest Component", I demonstrated that the tangraphic element 'r169' was polyphonic: i.e., it had several different types of phonetic values: e.g., a nyi-like value and a syu-like value in

TT5081 嫡 ?PRIMARY-WIFE nyï 1.30

TT0962 COOL shyu 1.2

This is not unlike 兌 Md dui which has different values in combination with other sinographic elements:

dui in 兌綐

duo in 敓莌鮵痥

shuo in 說

shui in 稅蛻帨涗裞

tuo in 脫挩侻

yue in 悅閱

Some sinographs have extreme polyphony in the Japanese script: e.g., 生 depending on context is shou, sei, i, u, o, ki, nama, ha, or fu (excluding even rarer readings).

In both cases, context enables the reader to select the correct reading.

If I saw 兌 by itself, I would pronounce it as dui in Mandarin, but if 兌 were combined with a 'radical' like 言, I would pronounce the resulting character 說 with the appropriate reading (shuo).

Similarly, if I saw 生 by itself in Japanese, I would pronounce it as sei, but if 生 were preceded by 芝 shiba, I would pronounce 生 as fu.




Context-free reading



A different reading with a preceding element

yan + 兌 = 說 shuo

shiba + 生 = 芝生 shibafu

Unlike 兌 Md dui or 生 Jpn sei, r169 has no context-free reading. Moreover, unlike 兌 Md dui, the multiple readings of r169 probably do not have a single phonetic prototype.

Could r169 be like Japanese 生, whose readings are based on a mixture of borrowed Chinese and native Japanese morphemes with meanings related to 'life', 'birth', and 'growth'? Maybe. If r169 meant X, and the Tangut A, Chinese, (and Tangut B?) morphemes for X were A, B, (and C?), then r169 would have two (or three?) readings.

But I think there may be one big difference between tangraphy and the other two scripts:

- All the 兌-derivatives are based on a single phonetic 兌.

- All the readings of 生 are based on morphemes associated with a single graph 生.

- However, can we be sure that r169

- is really one phonetic in all instances?

- has multiple readings all meaning the same thing?

It occurred to me on Tuesday that r169 - the left side of

TT5081 嫡 ?PRIMARY-WIFE nyï 1.30

looked vaguely like 尔 Tangut period NW Chinese ?*zhi, the right side of 你 Tangut period NW Chinese ?*ni.

And it also occurred to me that r169 - the bottom of

TT0962 COOL shyu 1.2

looked vaguely like

素 粟 肅 TPNWC ?*su

束, 术 in 述 TPNWC ?*shu

Are tangraphic phonetic elements polygenetic? Are they modifications of various sinographs which merged into a single form? If a polyphonous phonetic element has reading types A, B, and C, does that mean that element may be a modification of three sinographs X, Y, and Z?

Perhaps, but the probability of chance resemblance is very high. Many sinographs have the basic shape of r169: a 'jellyfish' shape with horizontal lines on top and vertical lines on the bottom. Can I really be sure that

TT0126 (type of horse) po 1.49 (Li Fanwen [1986: 211]: 勻 EVEN)

TT2783 報 REPORT po 1.49 (loan from Chn 報?)

were read as po 1.49 because r169 was derived from the right-hand side (甫) of graphs like 補 for TPNWC ?*pu?

Next: Plum sea-SON. COLD GOLD

In "The COOLest Component", I left out the derivatives of


TT0226 GOLD kiẹ 1.66 = TT5084 =

top of TT0203 RICH lo (tone unknown) +

bottom of TT2087 TRUTH Giey 1.34

TRUTH appears to be an abbreviated phonetic. One might expect


to sound like TRUTH Giey 1.34 or GOLD kiẹ 1.66, but its (Tangut A) reading is tsywar 1.82. If TT5084 is AB and if the components in its 'relatives' are C and D, then:

GOLD: CAB: kiẹ 1.66

TRUTH: DAB: Giey 1.34

A-LITTLE: AB: tsywar 1.82

Were kiẹ 1.66 and Giey 1.34 close to a Tangut B reading of TT5084 (two syllables *ki-e)? If the right-hand element was Tangut B *e, then did, for example,

TT3448 SIX chhyiw 1.46

TT3604 SCHOLAR ryiry 2.68

TT2934 PERSON myiry 2.68

(without the PERSON semantic element which is in SCHOLAR and even SIX!)

end in *-e in Tangut B?

In the surviving volumes of Tangraphic Sea, GOLD appears as a component in

TT0225 TURQUOISE khu 1.1

TT0229 (a personal name) kyĩ 1.16

TT2087 TRUTH Giey 1.34 (which in turn was derived from GOLD)

TT2390 WHITE syu 1.7

The bottom of GOLD is phonetic in TRUTH, and the bottom left of GOLD is phonetic in WHITE.

TURQUOISE is derived from GOLD + WATER. I used to regard this as a semantic compound: 'something that is colored like WATER and valuable like GOLD', but now I'm not sure. The bottom right component has no known semantic value and is not the normal semantic component for WATER, though it is the bottom right part of

TT4282 WATER zyïïr 2.85

(It's the left side of WATER which is the semantic component for WATER. Why wasn't the right-hand variant of this component used in TURQUOISE?)

The tangraph

TT0229 for the name kyĩ 1.16 is derived from the top and left of GOLD plus DOG, the left of

TT2235 (a clan name) shioow 1.57

TT0229 kyĩ 1.16 might have two phonetics based on Chinese:

金 'gold' in Tangut period NW Chinese was something like *?kyĩ

犬 'dog' in Tangut period NW Chinese was something like *?khywã

So even though the left of TT0229 looks like

TT0227 COOL shyuu 1.7

it seems to be an abbreviation of TT0226 GOLD kiẹ 1.66, possibly representing its Chinese translation *?kyĩ. kiẹ 1.66 itself may be a borrowing from Chinese.

Next: Polyphony via polygeny? THE COOLEST COMPONENT

If Nishida's radical 169 doesn't mean 'rare', what does it mean, if anything?

I can only find r169 on the left of five tangraphs:

TT5081 嫡 ?PRIMARY-WIFE nyï 1.30

r169 < right of TT5545 RELATIVES thyu 1.3 (semantic?)

TT5082 (kind of fish) nyi 2.12

r169 < ? (no analysis available)

TT5083 くくりつける FASTEN-TO sywi 1.11 (with WAIST on right)

r169 < left of

TT0970 樑 ?ROOF-BEAM sywi 1.11

TT5084 稍々A-LITTLE, ほゞ ALMOSTtsywar 1.82

r169 < bottom center of TT1879 節 PERIOD tserw 1.87

TT5085 CITY da 2.14 (Li Fanwen [1986: 265]: 壁 WALL?)

r169 < ? (no analysis available)

r169 has a phonetic value ny + high vowel in two cases but its function in the others is unclear.

r169 is beneath a 'horned hat' and on the bottom left of

TT2083 後代 ?LATER-GENERATION khyiy 1.36

('horned hat') + r169 < ? (no analysis available)

TT2084 眾 NUMEROUS Giey 1.34

'horned hat' + r169 < frame of TT2087 TRUTH Giey 1.34

TT2085 祭品 SACRIFICIAL-OFFERINGS Giey 1.34 (9.6.00:31: 'true things'?; cf. TRUTH below)

'horned hat' + r169 < frame of TT2087 TRUTH Giey 1.34

TT2086 (kind of animal) Giey 1.34

('horned hat') + r169 < ? (no analysis available)


TT2087 TRUTH Giey 1.34 =

top of TT1855 PURE sey 1.33

bottom of TT0226 GOLD kiẹ 1.66 = TT5084

It is clear that 'horned hat' plus r169 constitutes a phonetic element. 'Horned' r169 does not occur as an independent tangraph. Note that the analysis of TT2087 implies that 'horned' r169 can be split in two.

r169 occurs as the sole bottom element in two tangraphs derived from each other:


TT0227 COOL shyuu 1.7 =

top of TT0238 HEAD lyụ 2.52 (Li Fanwen [1986: 437]: 腦 BRAIN) +

bottom of TT0962 COOL shyu 1.2


TT0962 COOL shyu 1.2 =

top of TT0881 WOOD syi 1.11 +

bottom of TT0227 COOL shyuu 1.7

These indicate that r169 had a phonetic value shyu(u).

TT0962 itself is phonetic in


TT0963 嚇 FRIGHTEN shyu 1.2 =

all of TT0962 COOL shyu 1.2 +

right of TT3349 CALL ryiry 2.68

However, other tangraphs containing TT0962 on the left did not sound like shyu:

TT0964 SOUTHERN be 2.7

TT0962 < ? (no analysis available)

TT0965 MONASTERY tshə 1.27

TT0962 < TT0971-0972 BIG (see below)

TT0966 NORTHERN-SEA r 2.76

TT0962 < ? (no analysis available)

TT0967 EMPRESS dzow 1.54

TT0962 < TT0971-0972 BIG (see below; semantic)

(BIRD on right is phonetic < abbrev. for TT5717 BIRD jywow 1.56)

TT0968 大 BIG myaa 2.21

TT0962 < ? (no analysis available)

(TT0962 semantic; abbrev. for TT0971-0972, 5660 below?)

The first of these tangraphs seems to contain the second in abbreviated form as a phonetic:

TT0969 (part of a place name) tha 2.14

TT0962 < ? (no analysis available)

TT0971-972 (one tangraph with two readings)

BIG lyịy < ?*C-la 2.54 (native word; cognate to 大 OC *lats 'big'?)

BIG tha 2.14 (loan from 大 Tangut period NW Chn *tha)

TT0962 < ? (no analysis available)

No tangraphs containing TT0962 on the right sounded like shyu:

TT0455 BODY kwər 1.84

TT0962 < right of TT5660 BIG khwey 2.30

(left side may be phonetic; appears in three tangraphs pronounced Gyïr)


TT0962 < ? (no analysis available)

(why is GRASS in the middle of this tangraph?)

TT3476 頭胎 FIRSTBORN pey 1.33

TT0962 < right of TT5771 SEASON lu 2.1

TT5660 BIG khwey 2.30 (loan from Chn 魁?)

TT0962 < ? (no analysis available)

TT5771 SEASON lu 2.1

TT0962 < ? (no analysis available)

r169 in (bottom) right position appears to be phonetic in these three tangraphs:

TT2390 WHITE syu 1.7

r169 < bottom left of TT0226 GOLD kiẹ 1.66

TT4537 DEER syu 1.7 (TT2390 is phonetic)

r169 < all of TT2390 WHITE syu 1.7

TT5545 RELATIVES thyu 1.3

r169 < left of TT5081 嫡 ?PRIMARY-WIFE nyï 1.30

But note that only TT4537 derives r169 from a tangraph pronounced like syu.

These two instances of r169 in right position indicate that it had another phonetic value po 1.49:

TT0126 (type of horse) po 1.49 (Li Fanwen [1986: 211]: 勻 EVEN)

r169 < right of TT2783 報 REPORT po 1.49

TT2783 報 REPORT po 1.49 (loan from Chn 報?)

r169 < right of TT0126 (type of horse) po 1.49

Eight other tangraphs pronounced po 1.49 do not contain r169.

I wonder if r169 in these tangraphs was being confused with

TT5061 赤 RED phọ 2.62 (is the tangraph derived from 赤?)

which is phonetic in three tangraphs pronounced phiow, but not in

TT5082 (kind of fish) nyi 2.12

in which r169 is phonetic.

I have already examined two cases of r169 in tangraphs pronounced tserw 1.87 in "A Rare Hand".

Other examples of r169 in bottom right position are not obviously phonetic:

TT0754 WASH Gor 2.80

r169 < ? (no analysis available)

TT1090 SEVENTH-SON ŋwər 1.84

r169 < bottom center of TT1879 節 PERIOD tserw 1.87

(oddly, this tangraph does not contain SEVEN or SON)

TT5283 (genitive/dative) 'yiy 1.36

r169 < right of TT5545 RELATIVES thyu 1.3 (semantic? 'relatives' > 'belonging' > 'of, to'?; the left element is phonetic)

Summing up, r169 seems to have several types of phonetic values:

7 cases: shyuu 1.7 ~ shyu 1.2 ~ syu 1.7 ~ thyu 1.3

maybe the 2 cases of sywi can be added for a total of 9

2 cases: nyï 1.30 ~ nyi 2.12

2 cases: po 1.49

2 cases: tserw 1.87

When combined with 'horned hat':

5 cases: khyiy 1.36 ~ Giey 1.34

When r169 is combined with WOOD, it looks like TT0962 COOL but in some cases, it may actually be an abbreviation of TT0971-0972 and 5660 meaning BIG with the sound value tha 2.14.That begs the question of why the BIG tangraphs contain COOL (and of all things,

TT5649 WAIST jyiw 1.45!)

(Nishida [1966: 244] defined this element as "representing things having to do with the sun" in one of his lists of radicals but has left this undefined in the body of his dictionary [p. 455].)

And I still have no idea what r169 meant. All I can say is that it wasn't RARE.


I've been talking about 'radicals' a lot lately. One might assume that there are references to 'radicals' in the vast indigenous linguistic literature written by the Tangut themselves. But as Kwanten pointed out, there aren't any.

The Tangut were well-versed in Chinese literature, so they may have been aware of the 540 radicals in Shuowen. (The much more famous 214 radical system would not be devised until centuries later.) Nonetheless the Tangut did not analyze their tangraphs in terms of 'radicals' or the 六書 'six [principles of sinographic] writing'. There is no native Tangut list of tangraphic elements.

Hence these elements have to be identified centuries after the death of the Tangut language by

- examining how the Tangraphic Sea 'cut up' tangraphs into parts

- finding recurring semantographic correlations - if any

It can be difficult to determine what is an element and what is only a part of an element. For example, while writing "Lost among the Leaves", I realized that 乚 may be an independent element in

TT0461 OUGHT dyi 2.10


but may be part of a larger element

TT1100 毒 POISON do 1.49

in the following tangraphs:

TT0721 (a kind of tree) do 1.49

TT0075 SNAKE phio 2.43 (poisonous reptile?)

(I don't know why the top of POISON is written atop 干 in this tangraph but not atop VAPOR or INSECT in the following two tangraphs.)

TT3963 ([?poisonous] insect) gyuu 2.6

TT5796 SCORPION la (tone unknown; poisonous arachnid?)

If these tangraphs are really POISON + some other element rather than combinations of four elements including 乚, I would not expect them to be used as sources of 乚 in Tangraphic Sea analyses.

In fact, TT0075 SNAKE is cited as a source of 乚 in


TT1100 毒 POISON do 1.49 =

center of TT5796 SCORPION la (tone unknown) +

right of SNAKE phio 2.43 (i.e., 乚)

This may imply that TT1100 had two parts:

- its top and bottom left (一+コ+ㄇ)

i.e., Nishida radical 109 無いこと ABSENCE with an extended top horizontal line?

- its bottom right (乚)

If TT1100 were truly a single unit, it might have been analyzed as

without their top or (bottom) left sides.

So far, it looks as if 乚 really is an independent element in POISON. But I still can't figure out what it meant, if anything. Maybe it never had any semantic value. It could have been

- a 'filler' element without semantic or phonetic value inserted to maintain a certain degree of graphic complexity (i.e., to avoid one-element tangraphs)

The problem with the filler hypothesis is the existence of simple tangraphs which have no 'fillers': e.g.,

TT3344 PERSON dzywo 2.44

TT3301 HOLY shyïy 2.37 (borrowed from Chn 聖)

The left-hand element never (?) appears without a vertical line on its right. Hence I consider this a simple tangraph rather than a two-part tangraph.

TT1632 EIGHT 'yar 1.82

TT5649 WAIST jyiw 1.45

Why would these tangraphs lack fillers? Were they created before the 'filler' policy was instituted (or after it was cancelled)? Did one creator of tangraphs have such a policy while another did not?

- a Tangut B phonetic element with zero value in Tangut A

The problem with this hypothesis is, of course, the total absence of any evidence for Tangut B other than the script itself.

I could look at sinography for clues. 乚 is also an infrequent (though not rare) sinographic element (which is why it is in my Windows font). It has no obvious meaning in sinographs such as

乳 TPNWC ?*zhü 'breast, milk'

亂 TPNWC ?*lwan 'disorder'

(TPNWC = Tangut period NW Chinese)

and its phonetic function is obscure without referring to Old Chinese (which the Tangut would not have known about):

札 TPNWC ?*cha < OC *stret 'strip' (with semantic element 'wood')

軋 TPNWC ?*ya < OC *'ret 'crush under wheels' (with semantic element 'cart')

and possibly

禮 / 礼 TPNWC ?*li < OC *ri' 'rites' (but it is unusual to write a *-' word with a *-t phonetic, and I don't know how far back 礼 dates as an abbreviation of 禮)

But 乚 in tangraphy does not appear to have anything to do with the above sinographs. None of the tangraphs containing 乚 sound like 乚-sinographs, or have similar meanings. (If Tangut B existed, perhaps the sound of 乚 was based on the reading of one of those sinographs. Or 乚 could have had some totally arbitrary sound value.)

As a last resort, I could look at the highly restricted distribution of 乚. 乚 only occurs on the right side of tangraphs, and I have not yet found an instance in which a differently-shaped element in another position is derived from 乚.

If 乚 is a semantic element or an arbitrary addition, its location probably won't tell us much.

But always being on the right is significant if 乚 is phonetic. Since Tangraphic Sea analyses build up tangraphs from left to right, this implies that tangraphic elements were meant to be processed - read? - from left to right. Therefore a right-hand element is the last element in its horizontal sequence. And a purely right-hand or bottom element might represent a coda. Conversely, a purely left-hand or top element might represent an onset. And a purely medial element might represent a vowel, or a segment sequence with segments altered by internal sandhi - e.g., -gug- in Korean tagugO 'many languages' < takuk + O. (g cannot occur word-initially or word-finally in Korean; it is an allophone of /k/ between vowels.)

If the purely right-hand element 乚 always represented a word-final segment, then the bottom element in

TT1622 LEAF bạ 2.56

would have been silent: i.e., 卅 WOOD would have been a purely semantic element.

There are two kinds of 'pure' right-hand elements.

The first - the not so pure - have variant forms in other positions: e.g., FIRE in

TT5502 (syllable of a clan name) pyu 2.3 (left-hand version of FIRE + PERSON)

TT3920 (syllable of a clan name) tsya˙1.20 (right-hand version of FIRE + PERSON)

(the use of PERSON in name tangraphs is understandable, but why FIRE?)

This variation may reflect an aesthetic constraint (e.g., no bottom right strokes pointing upward) and/or a phonetic difference. Graphic and phonetic differentiation need not go together:

Greek non-final σ and final ς both represent [s]

Hebrew non-final מ and final ם both represent [m]

Arabic isolated ﻩ, initial ﻫ, medial ﻬ, and final ﻪ all represent [h]

The second kind (e.g., 乚 and ヒ) has no variant forms elsewhere. This may mean that 乚 and ヒ represent phonemes that only occur in word-final position.

If an element can occur in multiple positions, does that necessarily mean that the segment(s) it represents also appear in all those positions? Does it matter when the Tangraphic Sea extracts element X from a tangraph XY as opposed to a tangraph YX? Kwanten proposed that it does.

If X = [A] in left-hand position but X = [B] in right-hand position, the Tangraphic Sea should not derive a right-hand element from a left-hand element, and vice versa. However, derivations like


TT5809 BURN wə̣ 1.68 =

left of TT5807 OUTSIDE dyïr 2.77 +

bottom (left) of TT0993 ROAST pyu 2.3 (TT5502 above is phonetic beneath WOOD)

i.e., AB' < AC + DBC

are frequent, and enable us to equate left- and right-hand variants.

But Kwanten may still be correct if there need not be any correlation between multiple readings and multiple positions: e.g., X = [A] or [B] in any position. So if one wanted to describe a tangraph XY pronounced [A...], one would derive its X from a tangraph in which X = [A] (e.g., ZX [...A]) rather than a tangraph in which X = [B] (e.g., XZ [B...]), regardless of the placement of X.

Using this methodology to describe sinographs, I could write derivations like these:

閱 Md yue = all of 門 Md men + right of 悅 Md yue

稅 Md shui = all of 禾 Md he + right of 蛻 Md shui

(For details on 兌-graphs, see "A Lat of Derivatives".)

Notice how 兌 can have different phonetic values (yue, shui) in the same position (left) and a single phonetic value (yue) in different positions (center bottom and left).

FIRE in tangraphy may have been like 兌 in sinography, apart from one major difference: the lack of history. The seemingly unrelated readings of 兌 once shared a common phonetic core *lot:

Md yue < OC *lot

Md shui < OC *hlots

Could very different readings of tangraphic elements also have shared common phonetic cores when tangraphy was invented? I doubt it, for the Tangraphic Sea was compiled within a century after the invention of the script in the 1030s. It took a millennium for OC *lot and *hlots to become Late OC *ywiet and *shwies. Polyphony (in the sense of having multiple unrelated readings) must have been part of tangraphy from the start. The question is whether this polyphony involved one or two Tangut languages.

Although I started writing this post with the intent to figure out what Nishida's radical 169 really represented, I've gone way off track. I ended up talking about an entirely different element 乚 instead. I'll get back to radical 169 in the following post.

Next: The COOLest component. HOW RARE IS RADICAL 169?

Not very - in two senses.

First, it appears in many tangraphs.

Second, I don't think it means RARE. Nishida (1966: 408) listed three tangraphs under his radical 169 RARE:

TT5070 稀れ RARE, 少ない FEW zyïïr 1.92 (with PERSON on right)

(Li Fanwen [1986: 471]: 微 MINUTE, 稀 RARE)

TT5084 稍々A-LITTLE, ほゞ ALMOSTtsywar 1.82 (with right-hand version of radical 183 [meaning unknown on right)

TT5083 くくりつける FASTEN-TO sywi 1.11 (with WAIST on right)

Obviously TT5070 - the only one of the three that means RARE - doesn't even contain radical 169 RARE. Its left-hand element resembles

TT5068 GRAZING lhew 2.43

with an additional horizontal stroke. (Li Fanwen's version of TT5070 lacks this extra stroke.)

There are only two tangraphs with that left-hand element: TT5070 and

TT5069 ?MATERIAL lhew 1.43 (loan from Chn 料 'material'?)

(Once again, Li Fanwen [1986: 441] omitted the horizontal stroke atop GRAZING in his version of this tangraph.)

The analysis of TT5069 as


all of TT5068 GRAZING lhew 2.43 +

left of TT3290 HERB kyii 1.14

without any reference to the horizontal stroke suggests that this stroke may be little value.

The analysis of TT5070 as


right of TT1053 微 SMALL shyï 2.27 (with MOTION on top - why?)

left of TT3584 SMALL tsəy 1.40

also does not account for the horizontal stroke atop GRAZING.

It seems that Nishida confused TT5070 with TT5081 which actually does have radical 169 with PERSON on the right:

TT5070 稀れ RARE, 少ない FEW zyïïr 1.92

TT5081 嫡 ?PRIMARY-WIFE nyï 1.30

Out of curiosity I looked up 稀 RARE in Li Fanwen's index and found

TT1681 稀 THIN/SPARSE shyi 2.9

TT2396 稀 RARE / 喜 GLAD xyi 1.11

(borrowed from Chinese 稀 'rare' and 喜 'glad' which were nearly homophonous: *xi in Tangut period NW Chinese)

TT5145 異 STRANGE / 稀 RARE ryar 1.82 (vaguely like English rare by coincidence!; Grinstead tr. as SURPASS)

Li Fanwen left out TT5070 RARE from his index of tangraphs that he translated as 稀 'rare'. None of these tangraphs contain either GRAZING or radical 169.

Thus I conclude that radical 169 did not mean RARE.

Next: So what did radical 169 mean? A RARE HAND

The third translation for Chinese 葉 'leaf' in Li Fanwen's (1986: 880) index to Homophones is

TT1610 LEAF tserw 1.87

(See "Lost among the Leaves" for the first two.)

TT1610 appears to consist of Nishida radical 39 HAND + Nishida radical 169 稀 RARE. Radical 169 appears to be based on Chn 希/稀 'rare', though I've always thought of it as resembling Chn 示 'show'. Needless to say, HAND + RARE does not convey LEAF.

The Tangraphic Sea gives the following analysis:


TT1610 LEAF tserw 1.87 =

left of TT1875 闊 ?SPACIOUS xyïy 1.42

center of TT1879 節 PERIOD tserw 1.87

Apparently, RARE in LEAF is an abbreviated phonetic (PERIOD, a homophone of LEAF). But what does ?SPACIOUS have to do with LEAF? Why doesn't LEAF have the WOOD semantic element for plants? Why not analyze LEAF as PERIOD without its 'frame'? Cf. the analysis of

TT1632 EIGHT 'yar 1.82


TT1986 SEVEN shyạ 1.64

without its 'head' (an element unique to this tangraph).

How is anyone supposed to know that HAND in LEAF is from ?SPACIOUS as opposed to PERIOD or the left side of many other tangraphs such as

TT1832 EAR nyu 1.3

TT1834 (a surname) nyu 1.3

(< EAR [phonetic, without ヒ] + SURNAME [semantic])

TT1836 引 PULL shiey 1.34

(analyzed as 'horned hat' from

TT1783 行 GO chyiy 1.35

plus the bottom of LEAD [see below])

TT0008 導 LEAD shio 1.50

(PULL without its 'horns'; analyzed as the bottom of PULL)

TT1830-1831 CYMBALS tsyï 1.30 bạ 1.63 (disyllabic word)


TT1539 FINGER r 2.76

TT1545 HAND lạ

TT1537 HARD 'yiry 1.74

TT1594 COUNTRY lhyịy (tone unknown)

HAND in LEAF could even be from the center or the right side of tangraphs like

TT3537 STRIKE/CHIEF tywï 1.30 (looks like PERSON + HAND)

TT3548 YEAR kyiw 1.45

TT4472 GAME/PLAY 'u 2.1

TT5679 BROAD/VAST wạ 2.56 (looks like WAIST + HAND + WAIST)

TT2622 FINGER/TOE noo 1.52

(same elements as TT1539 FINGER r 2.76 but in reversed order)

TT2700 BEAT ta 1.17

(loan from 打 Tangut period NW Chinese *ta 'beat'; even the graphic structure is similar: both have HAND on the left side and a vertical line on the right)

TT3024 CATCH-FIRE lhew 2.38 (looks like ROPE + BEAT)

TT3990 PLANT tyạ 1.64 (looks like EARTH + HAND)

And how is anyone supposed to know that RARE in LEAF is from PERIOD as opposed to, say,

TT5081 嫡 ?PRIMARY-WIFE nyï 1.30 (with PERSON on right)

TT0962 涼 COOL shyu 1.2 (with WOOD on top - why?)

TT5545 RELATIVES thyu 1.3 (with Nishida radical 185 INTENTION on left)

TT3329 (a kind of plant) tserw 1.87 (analyzed as


left of TT3290 HERB kyii 1.14 [semantic] +

right of TT1879 PERIOD tserw 1.87 [phonetic])

The Tangut B hypothesis would interpret the Tangraphic Sea analysis


as AB 'leaf' = A from CADE '?spacious' + B from CABF 'period'

Next: How RARE is radical 169? LOST AMONG THE LEAVES

Since I just talked about 葉 Old Chinese *lap 'leaf' and its word family - e.g.,

633e 鍱 MC *yiep < OC *lap 'to plate with metal'

633g 牒 MC *dep < OC *lep 'tablet'

633h 蝶 MC *dep < OC *lep 'butterfly' (resembles a pair of flying leaves?)

and perhaps

633j 諜 MC *dep < OC *lep 'spy' (likened to a butterfly? - rather doubtful)

633k 蹀 MC *dep < OC *lep 'trample' (make flat?)

633p 屧 MC *sep < OC *s-hlep 'bottom inlay in shoe; shoe' (flat part of shoe?)

- I wanted to see if Tangut had any cognates. Apparently it doesn't. I can't find anything pronounced *la, *lia, or *lya with 'flat' meanings in Grinstead. (Tangut lost final *-p, so I would expect any Tangut cognates of OC *lap to be open syllables.) And none of the three words for 'leaf' in Li Fanwen's (1986: 880) index to Homophones has initial *l-.

This post will only deal with the first two words which are written very similarly:

TT1622 LEAF bạ 2.56

TT0751 LEAF war 2.73 (defined by Grinstead as BRANCH)

Both contain the element WOOD (卅) either below or atop

TT1621 散 SCATTER mẹ 2.58

(Li Fanwen [1986: 217] reconstructed its reading as be - could it be phonetic in TT1622 LEAF bạ 2.56? - but the rhymes don't match)

No analyses are available for any of these three tangraphs. Both LEAF tangraphs appear to be straightforward semantic compounds: LEAVES are the SCATTERed parts of a plant (WOOD). The positioning of WOOD might indicate that one word meant 'leaves attached to a plant (i.e., not on the ground)' and the other meant 'leaves detached from a plant (i.e., on the ground)'. But perhaps the tangraph with WOOD on top (the default position of that element) represented a more common word for 'leaf' (and if so, why did Grinstead mistranslate it as 'branch'?) and the tangraph with WOOD on the bottom (an unusual, rare position for that element) represented a less common word for 'leaf'.

None of the six other tangraphs with WOOD on the bottom represent plants or plant parts:

TT0573 COOK-WELL lẹ 1.65 (TT0572 BOIL wie 2.8 on top, consisting of FIRE-TRIGRAM plus MEAT [right-hand version])

TT2550 LOCK kywi 1.11 (made out of WOOD? - but the top element is Nishida's radical 28 - METAL!)

TT3081 SWIFT dzyïr 1.86 (what does speed have to do with WOOD?; FEAR [cf. Chn 怖 'fear'] and PERSON are on top)

TT4322 PLOUGHSHARE lyiir 1.93 (Nishida's radical 10 [meaning unknown] on top)

TT5344 BURN lyoor 1.94 (with SMALL and FIRE [right-hand version] on top)

TT5503 BAKE pyọ 1.51 (with FIRE [left-hand version] and PERSON on top)

This would imply that TT1622 LEAF represented leaves that are no longer part of a plant. The trouble is that Homophones has the preceding clarifier tangraph

TT0881 WOOD syi 1.11 (with PERSON twice plus an unknown element beneath WOOD)

implying that TT1622 LEAF is still part of a plant. Maybe the unusual position of WOOD in TT1622 LEAF is simply an arbitrary way of distinguishing it from TT0751 LEAF which contains the same two elements (WOOD and SCATTER). Nonetheless I still expect there to be some correlation between textual frequency of the two LEAF-words and the default/nondefault positioning of WOOD.

SCATTER appears to be a combination of

a rare left-hand element ㄇ of unknown meaning

Nishida's radical 230 SLANT

a rare L-shaped right-hand element 乚 of unknown meaning

We can try to infer the meanings of the mystery components by looking at other tangraphs with them:

TT1619 (syllable of name) xwa 2.14 (defined by Li Fanwen [1986:421] as 濠 MOAT; in Tangut era NW Chn 'moat' was something like *xaw; the mismatch of -wa : -aw eludes explanation)

(The only other tangraph with ㄇ as a pure left-hand element. The right-hand element is Nishida radical 236 HAND, probably an abbreviation of

TT1545 HAND lạ

cognate to 翼 Old Chinese *lək 'wing' [see "When Do 羽 Feathers 立 Stand?"])

TT1623 剛 HARD bạ 2.56 (homophonous with TT1622 LEAF bạ 2.56)

(9.2.18:40: The translation is Li Fanwen's. Jacques [2007: 153] translated it as 飢/饑 'famine', which would have been read as *ki in Tangut period NW Chinese. ㄇ seems to be a modification of 几 TPNWC *ki, indicating to the reader that TT1623 was the translation equivalent of a TPNWC word pro)

(Is the vague structural similarity to 剛 coincidental? The vertical line in Tangut is HAND [Nishida radical 39] and only graphically resembles Chn 刂 'knife'.)

(18:49: It is not clear how HAND and WOOD could function semantically in TT1623. One might expect something like EMPTY + STOMACH or MOUTH + NOTHING. On the other hand, the Chinese equivalents 飢 and 饑 have a transparent structure:

食 'eat' [semantic] + 几/幾 [phonetic])

TT1624 SPINACH bạ 2.56 (extended use of TT1622 LEAF bạ 2.56; right side appears to be an abbreviation of

TT1790 VEGETABLE naa 2.19)

Obviously, ㄇ+卅 is a phonetic element bạ (an abbreviation of TT1622 LEAF bạ 2.56?), but the meaning of ㄇ is still unclear.

TT0748 REAPHOOK kiə 2.28 (with WOOD on top; does this imply that ㄇ was pronounced kiə? This is more evidence for deriving ㄇ from 几 TPNWC *ki.)


TT3021 CAGE sywïy 1.62 =

left of TT3020 UTENSIL gyu 2.3 +

left of TT1619 MOAT xwa 2.14

Both CAGE and UTENSIL have Nishida's radical 5 (繩 ROPE) atop HAND. What does ROPE have to do with cages (as opposed to nets) and utensils?

I don't see any common semantic or phonetic value for ㄇ.

乚 might as well stand for 'lost', which describes how I feel. Without consulting Grinstead, I can find only two other tangraphs besides TT1621 散 SCATTER mẹ 2.58 with a pure right-hand 乚:

TT0461 OUGHT dyi 2.10


(homophonous with and similar to OUGHT, except for the arrangement of the left-hand elements [Nishida radical 261 (meaning unknown) and Nishida radical 201 WRITTEN-CHARACTER)

center and right-hand elements are phonetic in

TT1779 TIMID dyi 1.11 (with the mysterious 'horned hat' [Nishida radical 7] on top)

乚 appears to be part of a greater element in the following tangraphs:

TT1100 POISON do 1.49 (ult. borrowing from 毒 Late Old Chinese*douk 'poison'?; a combination of three unknown components - phonetic elements spelling the Tangut B word for 'poison' - also borrowed from Chinese?)

phonetic in

TT0721 (a kind of tree) do 1.49 (with WOOD on top)

TT0075 SNAKE phio 2.43 (poisonous reptile?; Nishida radical 98 域 REGION on left - why?; is it an abbreviation of

TT0625 BEAST shywow 1.56 [loan from 獸 Late Middle Chinese or TPNWC *shiw?])

TT3963 (kind of insect) gyuu 2.6 (poisonous insect?; Nishida radical 216 氣 VAPOR on left - why?)

TT5796 SCORPION la (tone unknown; poisonous arachnid?; Nishida radical 217 虫 INSECT on left)

How could the Tangut B hypothesis account for these tangraphs?

Tangraphs which have no clear semantic element are probably combinations of phonetic elements: e.g.,

TT1619 濠 MOAT xwa 2.14

TT1621 散 SCATTER mẹ 2.58

TT1100 毒 POISON do 1.49

TT1100 would be structurally comparable with the hangUl spelling 독 tok 'poison' (< Middle Chinese *dok), a stacking of the letters




Whether the components of TT1100 represent segments or syllables is not clear.

TT1100 may represent a Tangut B borrowing from Chinese like *d-o-k, though a letter for a final *k (or a syllable like *ku with a vowel added - cf. Jpn doku 'poison', yet another loan from Chinese) should have a higher frequency. But perhaps there were many ways to write *k(u).

The tangraphs

TT1622 LEAF bạ 2.56

TT0751 LEAF war 2.73

may indicate the Tangut B sound sequences

ABCD (translation equivalent of Tangut A bạ 2.56)

DABC (translation equivalent of Tangut A war 2.73)

sharing a common root ABC (homophonous with ABC 'scatter', or even affixed forms of a root 'scatter'? - the use of a single symbol for the prefix and suffix implies they were homophonous)

or a single Tangut B word ABC equivalent to both Tangut A bạ 2.56 and war 2.73. The position of the WOOD element indicates which Tangut A reading was intended. There is no reason to expect each and every Tangut B word to have one and only one Tangut A translation equivalent. In Japanese, there are many cases of one-to-more than one correspondences such as

native Japanese atsu- 'hot', written with two graphs for unrelated Chinese non-homophones 暑 or 熱

The Tangut B-A 'leaf' correspondence could be described in a similar manner:

Tangut B ABC 'leaf', the basis for two graphs for unrelated Tangut A non-homophones TT1622 and TT0751

Why "the basis" instead of "written"? Because there are no known Tangut B texts. The only evidence for Tangut B is tangraphy, whose structure may imply a language distinct from Tangut A.

I have wondered if some Tangut A texts are really Tangut B texts. If the two languages had convergent grammars, it might be difficult to distinguish them. However, surely they would not have identical grammars.

If Tangut B existed at all, it was an (artificial?) 'insider' language that the Tangut used to deliberately complicate their script.

Here is an analogy that may clarify this idea:

- In the far future, both Chinese and Japanese are forgotten.

- Only Japanese texts are discovered.

- Two spellings of atsu- 'hot' are discovered: 暑 and 熱.

- Their Sino-Japanese readings sho and netsu are not discovered.

- They are interpreted as semantic compounds:

暑 < 日 'sun' + 者 'person' (why?)

熱 < 埶 'to sow' (why?) + 灬 'fire'

- But in reality, these graphs were semantophonetic compounds reflecting Old Chinese, not Japanese, pronunciation*:

暑 OC *sta' < 日 'sun' + 者 OC *tya' 'person'

熱 OC *ngyet < 埶 OC *ngets 'to sow' + 灬 'fire'

- Similarly, tangraphs that appear to be semantic compounds are really semantophonetic compounds (or purely phonetic compounds) reflecting Tangut B, not Tangut A pronunciation.

*The semantophonetic structure of the two graphs for 'hot' is less obvious with Sino-Japanese or Mandarin pronunciation:

暑 SJ sho, Md shu < 日 'sun' + 者 SJ sha, Md zhe 'person'

熱 SJ netsu, Md re < 埶 SJ gei, Md yi 'to sow' + 灬 'fire'

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