The first half of

1055 4061 1tʃɨu 2riuʳ 'fine steed'

has an unusual structure. Although it shares a central vertical line (Boxenhorn alphacode: bae) with 218 other tangraphs (Tangut characters), this common component is flanked by unique left- and right-hand components (Boxenhorn alphacodes: tis and tir).

In the Tangraphic Sea dictionary, tangraphs are analyzed in terms of other tangraphs. If 1055 contains elements absent from all other tangraphs, how was it analyzed?


1055 1tʃɨu (first half of 1tʃɨu 2riuʳ 'fine steed') =

'frame' of 1013 1tʃɨu 'the surname Chu; transcription for Chinese *tʃɨu' +

bottom of 1054 1thiu (reduplicated in 1thiu 1thiu 'truth'; does it ever occur alone?)

( 1054 is an error for 1053!)

I suspect that 1055 is simply 1013 plus extra strokes rather than a fusion of two tangraphs. 1054 may be an unrelated tangraph which happens to have lots of horizontal strokes in its lower half.

Why regard 1055 as derived from 1013? The second half of 'fine steed' is homophonous with the surname tangraph

5489 2riuʳ

5489 is used to write parts of surnames but may not represent a monosyllabic surname. If there was a Tangut surname 'Rur', then perhaps 'fine steeds' were called 'chu-rurs', named after two families (the Chu and the Rur) associated with horses. Calling horses 'chu-rur' would be like calling handguns 'smith-wessons'.

There are at least three problems with this hypothesis:

1. There is no independent evidence to suggest that the Chu and Rur families bred horses, etc. (Is anything known about non-aristocratic Tangut families?)

2. I am not certain there ever was a surname Rur. Although the Homophones dictionary defined 5489 as a tribal name, that doesn't necessarily mean it could stand alone.

3. If the first half of 'fine steed' is the tangraph for the surname Chu with additional strokes, why isn't the second half the tangraph for the surname Rur with additional strokes? There is no graphic resemblance between 4061 and 5489:


4061 (2nd half of 'fine steed') vs. 5489 'the surname Rur''

Why not just add 'horse' to 5489 instead of creating a totally unrelated character 4061 for the second half of 'fine steed'? (And why not just add 'horse' to 1013 for the first half of 'fine steed' instead of arbitrarily adding extra strokes?)

Next: What Lies Upon the Back of a Fine Steed DIVIDING A FINE STEED

In part 3 of "Kenning Camels", I asked,

... what if 1tʃɨu 2riuʳ 'fine steed' is a compound rather than an indivisible disyllabic word?

David Boxenhorn proposed that 1tʃɨu meant 'steed' by itself. Although no such monosyllabic word is attested,

1055 4061 1tʃɨu 2riuʳ 'fine steed'

might be an old phrase that later became an unanalyzable noun:


noun 1tʃɨu, adjective 2riu, phrase 1tʃɨu 2riuʳ (with Tangut noun-adjective order) all coexist


noun 1tʃɨu extinct; 1tʃɨu 2riuʳ reanalyzed as a noun; does adjective 2riu still exist?

1tʃɨu would be like the were- of werewolf, a once-independent word that is now only part of a compound whose second half (wolf) is still an independent word.

Was there any nearby language with a 1tʃɨu-like word for 'horse'?

There are four tangraphs other than the second half of 'fine steed' pronounced 2riuʳ. None are adjectives and none are the sort of noun I'd expect in second position: e.g., 'king': 'horse-king' = 'king of horses' = 'fine steed'.

2470 2riuʳ (half of

2470 2470 2riuʳ 2riuʳ 'bat' (reduplicative word)

2470 0564 2riuʳ 1dwiaʳ 'bat'

0564 1dwiaʳ is 'skin' in isolation

Neither word is attested outside dictionaries, but I'm guessing both are Common Tangut (CT).

5276 2riuʳ (second half of

5272 5276 2tʃhɨə 2riuʳ 'sore, ulcer' (CT))

5489 2riuʳ (part of the three surnames

1. 3230 5489 ?lhiẹ 2riuʳ 'Lherur'

Lhe is the first syllable in at least six surnames.

2. 5489 1073 2riuʳ 1kə 'Rurky'

8.20.3:38: 1073 1kə is the first half of

1073 1300 1kə 1lã 'aerolite'

which is not attested outside dictionaries and might be RT. Is 1073 purely phonetic in 5489 1073, and if it is, why not create a special tangraph for 1kə in family names? Why are family names written with a mixture of created and recycled tangraphs?

3. 4589 5489 2kiə̣ 2riuʳ 'Kyrur')

I don't know if there was a monosyllabic surname Rur.

5736 2riuʳ 'to ferment' (CT)

It's possible that both halves of 1tʃɨu 2riuʳ 'fine steed' were once independent words that became extinct without a trace in later Tangut, but I'd rather treat the word as indivisible.

8.20.4:09: Could 1tʃɨu 2riuʳ be a borrowing from Khitan? Unfortunately, there is no known Khitan word *juru or *curu 'horse', and the closest known word *cur means 'two'.

(Khitan 'voiced' obstruents may have been voiceless nonaspirates that contrasted with aspirated voiceless obstruents: *j = *[tʃ] vs. *c = *[tʃh], etc.) KENNING CAMELS (PART 3)

I didn't expect to continue this series, but David Boxenhorn came up with a kenning that might underlie the Ritual Tangut word 2miəə 1tʃɨu for 'camel':

Ritual Tangut syllables Proposed Common Tangut source words

2miəə boundary
1tʃɨu 1tʃɨu steed

David suggested the equation

'boundary' = 'borderland' = 'desert'

Hence 'boundary steed' = 'desert steed' = 'camel'.

However, all examples of 'boundary' in Li Fanwen (2008) are in Buddhist texts and none seem to involve literal physical boundaries or deserts.

2miəə shares a consonant with Middle Chinese 漠 *mak 'desert', but the vowels do not match.

(The odd correspondence of Tangut schwa with MC *a is also found in 'camel':

T 2la 2diə : 駱駝 MC *lakda)

In part 2, I objected to half of a word - the first half of 1tʃɨu 2riuʳ 'fine steed' - in a kenning, but David suggested that 1tʃɨu may be an abbreviation for 1tʃɨu 2riuʳ 'fine steed'. That reminded me of how 駝, the second half of Tangut period Chinese 駱駝 *lotha 'camel', became the source of Tangut

3973 1tha 'camel'

and is an abbreviation for 'camel' in 駝背 'hunchback' (lit. '(ca)mel-back') and 駝鳥 'ostrich' (lit. '(ca)mel-bird'; usually now written 鴕鳥 with 鳥 'bird' instead of 馬 'horse' in the first character).

The use of monosyllables as abbreviations of disyllables is well-known in Chinese, though perhaps not fully understood. Was there a similar practice in Tangut?

In any case, what if 1tʃɨu 2riuʳ 'fine steed' is a compound rather than an indivisible disyllabic word?

Next: Dividing a Fine Steed KENNING CAMELS (PART 2)

If disyllabic Ritual Tangut (RT) words consist of kennings, then I would expect the RT word for camel

0704 1260 2miəə 1tʃɨu 'camel'

should be homophonous with a sequence of Common Tangut (CT) words that obliquely refer to a camel: e.g., 'golden mountain', 'desert horse', 'walking hill', etc.

There are two tangraphs pronounced 2miəə:

0163 2miəə 'to call' (RT CT?), 'boundary' 'region' (CT) (details on the corrected glosses)

4272 2miəə (first half of

4272 4306 2miəə 2khɛw 'a kind of tree';

only attested in dictionaries; might be RT)

I cannot see what either has to do with camels. This isn't looking good.

There are seven tangraphs pronounced 1tʃɨu:

1013 1tʃɨu 'the surname Chu; transcription for Chinese *tʃɨu; CT)

8.18.1:43: 1013 is the source of the left side of 0660, the first half of 0660 3996 2la 2diə 'camel' (CT) - coincidence?

1055 1tʃɨu (first half of

1055 4061 1tʃɨu 2riuʳ 'fine steed' - CT)

8.18.1:17: graph derived from 1013 with extra strokes?

1736 1tʃɨu 'to do, to work, to manage' (CT)

1882 1tʃɨu (second half of

5412 1882 2lwəʳ 1tʃɨu 'shaman, sorcerer';

only in dictionaries?; RT?; CT equivalent is 3439 piẹ)

3272 1tʃɨu (second half of

1297 3272 2ka 1tʃɨu 'urn, earthen jar, bottle';

only in dictionaries?; RT?; synonym

2000 5196 1ŋwə 2liẽ

also only in dictionaries but probably CT since 5196 2liẽ 'bottle' is widely attested)

5012 1tʃɨu 'the surname Chu; to do, to work' (alternate spelling of 1013 and 1736; was the 5012 Chu family different from the 1013 Chu family?; CT)

5054 1tʃɨu 'have to' (CT)

Andrew West pointed out that 1055 is the only one whose semantics even come close to 'camel', but it's only half of a word. And even if 1tʃɨu turned out to be an earlier Tangut monosyllabic word for 'horse', I would then have to explain

- the first syllable of 'fine steed'

- the first syllable of 'camel' (RT)

So does this one counterexample invalidate the kenning hypothesis? Not necessarily. Perhaps other RT words are kennings, whereas the RT word for 'camel' may be an old native word for 'camel' that was displaced in non-ritual contexts by two waves of Chinese loans:

0660 3996 2la 2diə 'camel' < Middle Chinese 駱駝 *lakda 'camel'

and the newer

3973 1tha 'camel' < second half of Tangut period Chinese 駱駝 *lotha 'camel' KENNING CAMELS (PART 1)

Although there are about 6,000 different tangraphs (Tangut characters), the vast majority of Tangut texts only contain about 3,000 different tangraphs with the exception of dictionaries which include pronunciations and defintions for 'extra' 3,000 tangraphs. What are these 'extra' tangraphs for?

As I mentioned in "How Many Tangut Words for 'Camel'?", Grinstead (1972: 199) noted a large number of Tangut equivalents of a single English word. Some of these equivalents turned out to be monosyllabic halves of disyllabic equivalents rather than true monosyllabic equivalents. Nonetheless, a wealth of synonyms remains.

In 1996, Ksenia Kepping announced her discovery of odes containing words written with the 'extra' tangraphs. She called this vocabulary 'Ritual Tangut' (RT). David Boxenhorn has suggested the term 'Odic Tangut'.

RT has two unusual characteristics:

1. RT words are not obviously related to their Common Tangut (CT; i.e., normal) equivalents: e.g.,

0704 1260 2miəə 1tʃɨu 'camel' (RT)

sounds nothing like

0660 3996 2la 2diə 'camel' (CT)

However, there is some degree of graphic overlap between the spellings of the two words. I will write about this in a later post.

2. RT words are disyllabic, whereas CT roots are often (not always) monosyllabic. In this particular case, the RT word 2miəə 1tʃɨu corresponds to both a monosyllabic CT word

3973 1tha 'camel' (from the second half of Tangut period Chinese 駱駝 *lotha)

(note again the partial graphic resemblance to the spelling of 2miəə 1tʃɨu)

and a disyllabic CT word 2la 2diə 'camel' (possibly from Middle Chinese 駱駝 *lakda).

I thought RT might be based on a substratal language - possibly even the 'Tangut B' language that may be implied in the otherwise inexplicable structure of many tangraphs. However, on Monday (my time), Andrew West proposed that RT could be based on kenning: e.g., the RT word for 'camel' could be a combination of two CT words (2miəə, 1tʃɨu) indirectly referring to a camel.

Next: Testing the Kenning Hypothesis DISSECTING TANGUT CAMELS

The Tangut characters for 'camel' are derived from each other.

The analyses of 0660 and 0704 are circular:


0660 2la (first half of the regular word 2la 2diə 'camel') =

left of 1013 1tʃɨu 'the surname Chu; transcription for Chinese *tʃɨu) +

(Were camels associated with the Chu family?)

right of 0704 2miəə (first half of the ritual word 2miəə 1tʃɨu 'camel')


0704 2miəə (first half of the ritual word 2miəə 1tʃɨu 'camel') =

left of 0346 2ʃɨa (first half of 2ʃɨa 2vəi 'bliss') +

(Were camels associated with bliss?)

right of 0660 2la (first half of the regular word 2la 2diə 'camel')

The analysis of the second graph for the regular word

0660 3996 2la 2diə 'camel'

is unknown, but it is a component in the circular analyses of the other two tangraphs for 'camel':


1260 1tʃɨu (second half of the ritual word 2miəə 1tʃɨu 'camel') =

bottom left of 3996 2diə (second half of the regular word 2la 2diə 'camel') +

center and right of 3973 1tha 'camel'


3973 1tha 'camel' =

left of 3996 2diə (second half of the regular word 2la 2diə 'camel') +

center and right of 1260 1tʃɨu (second half of the ritual word 2miəə 1tʃɨu 'camel')

Grinstead (1972: 112, 206) identified two other tangraphs for 'camel'. Their analyses do not suggest that they meant 'camel':


3985 kiaa 'foundation; burden' =

right of 2679 2niə 'to arrive' +

right of 1603 nia (tone unknown) 'to repair'


1670 1swie (second half of 2swio 1swie 'to grind') =

top left of 0398 2swio 'to grind' (independent word; 2swio 1swie is a reduplication) +

bottom right of 4174 2miu 'to move' +

right of 3576 2swie 'clear'

They do, however, share a right-hand element pak (in David Boxenhorn's alphacode) which in turn resembles the 它 on the right of Middle Chinese 駱駝 *lakda 'camel'. Is that a coincidence?


3996 2diə (second half of the regular word 2la 2diə 'camel')

simply be a distortion of the second half of 駱駝? The left-hand element giu (in David Boxenhorn's alphacode) vaguely resembles Chinese 馬 'horse'. There is a tangraphic element for 'horse' based on 馬:

alphacode: hin

Why wasn't hin used instead instead of giu? hin appears in tangraphs for words related to horses: e.g.,

0764 1rie 'horse' (hin cannot stand alone, so it's accompanied by the mysterious right-hand filler ヒ)

1115 1gie 'horse' (Earthly Branch)

0803 2riaʳ 'horse' (in compounds; cognate to 0764)

0520 1dzee 'to ride' (1886 'person' on right)

2407 2dzee 'to ride' (2541 'person' atop 0764 'horse'; cognate to 0520)

hin is never combined with pak, so one cannot argue that 3996 giubaepak (bae being the alphacode for a vertical line) was necessary because the combination hin(bae)pak was already taken.

1603 nia (tone unknown) 'to repair'

looks like hinpak in my graphic but is actually hiopak with a different left element hio whose second vertical line is as tall as the フ on the top left.

And what is the point of bae in 3996 or in 585 other tangraphs such as 0520 'to ride'? It has no known semantic or phonetic function.

Finally, if 3996 (the second half of 2la 2diə)  is based on the second half of 駱駝, why isn't 0660 (the first half of 2la 2diə) based on the first half of 駱駝?

I'm out of time. More tomorrow.


Grinstead (1972: 199) wrote in the preface to his "English-Tangut Word-List",

One rather surprising result from this list is the frequency of multiple Tangut solutions [translations] to a simple English word. Given that the Tangut vocabulary was about 5800 characters, this richness of synonyms is worth further investigation.

Grinstead implicitly equates Tangut characters (tangraphs) with words. He thought the Tangut script had only "about 5800 characters". However, Tangut must have had more than 5800 words. Although the Tangut vocabulary could be written with about 5800 (perhaps more like 6000+) characters, there is no one-to-one correlation between characters and words. Each tangraph represented a syllable of Tangut, and Tangut words could consist of one or more syllables. Therefore Tangut polysyllabic words were written with two or more tangraphs.

Grinstead (1972: 112, 206) listed six different tangraphs for 'camel':

Do these tangraphs represent six different Tangut monosyllabic words for 'camel'?

The first tangraph

3985 kiaa 'foundation; burden' (two unrelated homophones written with the same tangraph?)

has nothing to do with camels according to Li Fanwen (2008: 640).

The third tangraph represents the second half of a disyllabic word

0398 1670  2swio 1swie 'to grind'

Did Grinstead find a text in which 3985 and 1670 were used for 'camel', or did he confuse them with

3996 (see below)

which has the same right side (pak in David Boxenhorn's alphacode).

The other four tangraphs represent parts of three different words for 'camel'. Andrew West pointed out the first two words to me:

1. The basic word

0660 3996 2la 2diə 'camel'

This is the most widely attested word for 'camel' and the only word that appears in Tangut texts other than dictionaries.

2la 2diə vaguely resembles Middle Chinese 駱駝 *lakda 'camel', but the rhymes of the second syllables do not match. If this word was borrowed from Chinese, the difference in rhymes needs to be explained.

Grinstead listed the tangraphs for 2la 2diə in reverse order: 3996, 0660. Was he unaware that 0660 and 3996 formed the spelling of a single word?

2. The ritual word

0704 1260 2miəə 1tʃɨu 'camel'

is only attested in dictionaries and may be part of the Tangut ritual (or 'odic') language found only in odes and dictionaries. Neither Andrew nor I can identify any cognates.

1260 is not in Grinstead's list of 'camel' tangraphs.

Although tangraphy is often thought to be a semantically-based writing system, the spellings for the basic and ritual words only share two graphemes in common:

bae and gol in David Boxenhorn's alphacode

3. The Chinese loanword

3973 1tha 'camel'

is probably from the second syllable of Tangut period northwestern Chinese 駱駝 *lotha 'camel', just as

2852 1tha 'Buddha'

is probably from the second syllable of Tangut period northwestern Chinese 佛陀 *fətha 'Buddha'.

Oddly, the first syllable of TPNWC 佛陀 *fətha was also borrowed as

0418 1xwɨə 'Buddha' (Tangut had no native f)

which never combined with 2852 to form a hypothetical


*0418 2852 1xwɨə 1tha 'Buddha' (asterisks = unattested)

corresponding to both syllables of TPNWC 佛陀 *fətha 'Buddha'.

If the basic word 2la 2diə is also from Chinese, it must have been borrowed before *d shifted to *th in TPNWC.

Next: Dissecting Tangut Camels

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2011 Amritavision