The Chinese character for 'ox' (as a member of the zodiac) is 丑. It belongs to a pure type B (nonemphatic) phonetic series with mostly nasal initials in Middle Chinese:

GSR number Sinograph Old Chinese Middle Chinese
1076a-d *?...nuʔ *ʈhuʔ
1076e *nuʔ *ɲuʔ
1076h-j *s-hnuʔ < ?*sɯ-snuʔ *suʔ
1076k *nuk *ɲuk

The OC initial cluster of 丑 'ox' is uncertain.

If I were unaware of the rest of GSR1076 and of Taiwanese (see below), I'd reconstruct 丑 as OC *thruʔ or *r-thuʔ since *thr- and *r-th- are the most obvious sources of MC *ʈh-.

However, most of the series has a common denominator *nuK (K = back consonant), so 丑 must have had an *n: *?...nuʔ. The nasalization of Taiwanese thiũ may reflect this original nasal. (Taiwanese also has a nonnasal thiu.)

Without looking at any non-Chinese evidence, I would reconstruct 丑 as *t-hnuʔ (< *tɯ-snuʔ?) with a cluster that fused into an aspirated retroflex stop:

*t-hn- > *t-hl- > *thl- > *thr- > *ʈh-

The problem is that Tai languages that have borrowed the Chinese calendrical animal names have a very different initial for 'ox': e.g., Ahom plaao, Pa-po pao, Tai Lü pau (Nishida 1975: 9). Assuming that plaaw is borrowed and is not an anomalous non-Chinese term in an otherwise Chinese set of terms, how can its pl- be reconciled with the *n- indicated by the other evidence?

Solution 1 (Nishida 1975: 9 with added details)

OC *thn- (aspirated stop-nasal cluster, not a stop-voiceless nasal cluster *t-hn-) became *thl- which was borrowed into early Tai, dissimilating to *phl- and then deaspirating in Ahom and other Tai languages (why?).

I've seen tl-clusters dissimilate to kl-clusters, but I've never seen tl- > pl-. Are there attested examples of that change?

Solution 2 (inspired by my memory of Pulleyblank 1962, which I don't have on hand)

OC *t-hn- became *θn- in a southern dialect, shifting to *fn- or *fl- before being borrowed into early Tai as *phl- (since there is no *fn- or *fl- in Proto-Tai). Hawai'i Creole English bef from English bathe has a similar interdental-to-labiodental shift.

But is there any other evidence for in OC? Pulleyblank first proposed OC in 1962 but has long since abandoned it, and I know of no other current reconstruction with it.

This solution also cannot account for the absence of aspiration in the Tai forms for 'ox'.

Neither solution 1 nor 2 explain why Ahom has -aaw instead of -u.

Solution 3 (entirely mine, but that doesn't make it right)

Perhaps the southern OC word for 'ox' had the same root *nuʔ with an emphatic low-vowel prefix: *-nuʔ.

Emphatic harmony caused the vowel to bend down: *-nuʔ > *pʌ-nɑuʔ.

The presyllabic vowel was lost: *pʌ-nɑuʔ > *pnɑuʔ.

The nasal shifted to a lateral, resulting in the form borrowed by early Tai: *plɑuʔ. Emphasis may have been nonphonemic or lost at this point.

Unfortunately, I know of no past or present southern Chinese words for 'ox' with p-: e.g., Taiwanese has thiũ < *t-hnuʔ. If southern OC *-nuʔ. ever existed, it has left no trace in Chinese, like the *-k variant of Middle Chinese *ʈhuʔ implied by Sino-Korean 축 chuk for 丑. HOW I KNEW THIS YEAR WAS MIUU

'Because I said so' is never an acceptable answer.

In the Tangut-Chinese glossary Pearl in the Palm (1190 AD), the Tangut characters for the Chinese zodiac are listed along with pronunciation keys in 12th century northwestern Chinese.

The Tangut character for the animal associated with the lunar year more or less corresponding with 2009


has the Chinese phonetic gloss

Although 沒 is pronounced mo in modern standard Mandarin, no such language existed back in the 12th century, and the phonetics of 12th century northwestern Chinese are largely uncertain. I am fairly sure it was pronounced with initial mb-, but am less certain that the rhyme was -o. Moreover, the function of the character 輕 'light' to the bottom right of 沒 is unclear. So we need to find more clues.

'Ox' appears in

'heavy lip sounds'

the first chapter of the Tangut dictionary Homophones. This tells us that 'ox' had a bilabial (heavy lip sound) initial consonant: e.g., p-, b-, m-. This correlates with the initial of mb- of the Chinese character gloss 沒. Since there were probably no similar-sounding Chinese characters with the rhyme of 沒 preceded by initial m- or b-,& nbsp;沒 could represent a Tangut syllable beginning with m-, b-, or mb-.

One might guess that 輕 'light' in the Chinese gloss could refer to 'light lip sounds' (labiodentals like f- and v-), but labiodental-initial Tangut characters were placed in the second chapter of Homophones. 'Light' in this context probably meant 'pronounce mb- without b-'. There is an instance in which an mb-Chinese character + 輕 'light' correspond to m- in Tibetan transcriptions: e.g.,

'to move'

was transcribed in Chinese as

(mbo + 'light')

and in Tibetan as mu and muH.

There is no known Tibetan transcription for 'ox', but by analogy with 'move', I could reconstruct its initial as m-:

Gloss Chinese transcriptions Tibetan transcriptions Tangut initial
to move mbo + 'light' mu(H) m-
ox ?mbo + 'light' none

One might think that 'to move' and 'ox' were homophones in Tangut on the basis of their Chinese transcriptions, but the native Tangut dictionary Precious Rhymes of the Tangraphic Sea lists them under two different rhymes:

'to move': rising tone rhyme 3

'ox': rising tone rhyme 6

These two rhymes must have been similar if they were both transcribed in 12th century NW Chinese as mbo. The Tibetan transcriptions for Tangut characters of those rhymes are also similar:

Rhyme Tibetan transcriptions
rising tone rhyme 3 -u, -uH, -yu, -yuH, -o, -oH, -uo (sic); 'to move' also used to transcribe the Sanskrit syllable mu
rising tone rhyme 6 -u, -uH, -yu, and one anomalous case of -wa (intended to be -wu or -wo?; the letter for -u or -o might have been left out or lost); no -o transcriptions but note how 'ox' was transcribed in Chinese with a possible -o character 沒

The difference between the two rhymes is not clear, and there is no agreement on what the difference was:

Rhyme Sofronov 1963 Nishida 1964 Hashimoto 1965 Sofronov 1968 Starostin 1977 Li 1986 Gong 1997 Arakawa 1999 AMR 2009 AMR grade
rising tone rhyme 3 -ǐuɦ -ju (> later -i̭u) -ǐuu -ju -yu -iu IV
rising tone rhyme 6 -i̭u -ǐʊɦ -jo -i̭un (> later -i̭u) -(j)üʔ -ıuu -juu -u:' -ɨuu/-iuu III/IV

Although I have tentatively carried over Gong's interpretation of the distinction in terms of vowel length (u vs. uu), I suspect the distinction was of a different nature. I reconstruct two different medial vowels (-ɨ- and -i-) corresponding to Gong's -j- depending on the initial and the 'grade' (rhyme class)*. Rising tone rhyme 6 is a grade IV (palatal high vowel)** rhyme following labials, so I converted Gong's -j- to palatal -i- instead of nonpalatal -.

This table sums up the logic behind each segment of my reconstruction miuu:

Segment m i u u
Reasoning Chinese mb- with 輕 'light' (= do not pronounce -b-) Grade IV (palatal); Tibetan transcriptions with -y- Tibetan transcriptions with -u carryover from Gong's reconstrruction to distinguish this rhyme from rising tone rhyme 3

A more skeptical reconstruction sticking only to what is almost certain would be mu.

The Tangut word for 'ox' may be onomatopoetic in origin (cf. English moo).

*Tangut had a grade system like that of Chinese (Gong 1995).

**Grade IV in Chinese (and probably also in Tangut) was characterized by maximum palatality (Pulleyblank 1984). HAPPY MIUU YEAR!

If the Tangut still existed, they would regard 2009 as the year of the ox (miuu in their extinct language) and post graphics like this online:

The big character (taken from Li Fanwen's Tongyin yanjiu) is

miuu 'ox' (in the twelve animal cycle; not the regular word for 'ox'; rhymes with English new; see its analysis here*)

The four smaller characters (in the Mojikyo font) at the bottom are read from right to left:

niəə 'two'

təụ 'thousand'

giəə 'nine'

kiew 'year'

niəə təụ jaʳ kiew - the year 2008 - was wonderful. I got to make a number of new friends and renew some old relationships. I hope this year will be even better for you. If we are lucky, Macross won't come true and there won't be any Zentradi invasion in 2009.

I would like to thank Andrew West, David Boxenhorn, Guillaume Jacques, Nathan Hill, Sven Osterkamp, and Viacheslav Zaytsev, whose innumerable contributions make this site possible. Your efforts bring new life to the study of a dead language. I look forward to our adventures in Tangutology in the months to come.

The character

miuu 'ox'

consists of

'top'/'female' (why?)

atop some components of unknown function.

It may be significant that the bottom center and right elements of

miuu 'ox'

are found in only three other characters:

1. mi 'Tangut' (with the left side of

'holy, wise' on the left)

2. second syllable of lhwiẹ dʒiə, another name for the Tangut (with


added to the left of mi 'Tangut' above)

3. miuu 'pool; deep; depths' (homophonous with miuu 'ox'; presumably the shared elements represent the sound miuu, but why is the nonshared element


which has no semantic relationship to 'pool' or 'deep'?) A TALE OF TAILS

Since it's the tail end of the year, I'd like to look at the tangraphs with the long-tailed variant

of 'demon'.

The normal version of 'demon' can only appear on the right side, but the other version supports other elements atop its tail:


Tangut telecode





Elements atop tail





second half of kɪ tswəʳ 'to curse; curse' (but Kychanov 2006: 777 has the syllables in the reverse order)






second half of thaa ʔwɨu


(unknown) + evil





Skt preta (hungry ghost); skull of a dead person






hell; underworld (near-homophony with Middle Chinese 地 *di

'earth' [and first half of 地獄 'hell' = 'earth-prison'] coincidental?)






second half of tʃhɨu mie 'Skt raakṣasa; name of a demon'






second half of lwiu xwəi 'to drive out evil ghosts or demons'

beast ('go' on top)





to strangle

grass ('death' on right)

Why do these tangraphs contain unusual long-tailed demons instead of the more common short-tailed demons? I can answer that question in three ways:

1. Because a mirror image tangraph already exists

TT4465 and its mirror image, TT2222, represent the word

kɪ tswəʳ 'to curse; curse'

I assume the tangraph for the first syllable was devised first with 'demon' in the normal position.

In Tangraphic Sea, each of those two tangraphs is used to analyze the other in combination with 'incantation':



I don't know how 'surpass' is relevant to 'curse'.

One might expect other disyllabic words written with long-tailed demons to contain mirror image tangraphs, but they don't. See below.

( TT4469 has a mirror image

TT5619 dʒ-? R? ?.? 'void'

without an entry in Kychanov [2006].

One might expect TT4469 to be a mirror image of the first tangraph in

tʃhɨu mie 'Skt raakṣasa; name of a demon'

but perhaps it isn't because the rare left-hand phonetic element of the first tangraph

doesn't seem to appear in right-hand position.)

2. Because a tangraph containing 'demon' has been recycled as a whole

This accounts for TT1044 and TT3193:


TT1044 second half of 'to drive out evil ghosts or demons' = 'go' + 'preta'


TT3193 'strangle' = 'demon' + 'calamity'

One could also view TT3193 as born out of necessity since the combination of 'grass' + 'demon' (with the 'demon' radical on the usual right side) already represented 'calamity'.

3. I have no idea

why a short-tailed 'demon' isn't on the right of

since they lack mirror-image counterparts and the elements atop the long tails can appear in left-hand position. I would expect the second tangraph of

thaa ʔwɨu ?'werewolf'

to be a mirror image of the first tangraph. The first graph of

lwiu xwəi 'to drive out evil ghosts or demons'

doesn't even contain 'demon'; its left side is


(cf. Chn 驅魔 'to drive away demons' with 馬 'horse' - but lwiu is an independent verb 'to lose').

Is xwəi an old word for 'demon' that has become a bound morpheme attached to lwiu? I doubt that because Tangut has object-verb order, not verb-object order. Could lwiu xwəi be a compound verb with a second half that was once as independent as the first? MI-NIA MAN-WOLVES?

The most surprising


word in Kychanov (2006) was

thaa R22 1.22 ʔwɨu R2 1.2 'werewolf'

Did the Tangut (Mi-nia) really believe in werewolves? I thought the concept was purely European.

The first tangraph consists of

Li Fanwen radical 317 (meaning unknown)

taken from the left of 'near; relative' (very vaguely like Chn 親 with the left and right elements reversed: 見+亲?) plus the 'demon' right half of jɨu 'demon' ('close to demons' > 'demonic'?):


The second tangraph consists of 'demon' taken from the right of 'oath; swear an oath' plus

Li Fanwen radical 005 (meaning unknown)



taken from the center and right of 'harm; injure; misfortune; obstacle':


12.31.2:20: 'Harm' could be from the left and center of 'heart' plus 'evil' from some other tangraph:


While looking up

TT2387 ?riẽ R43 2.37 ~ rieʳ R79 2.68 'a kind of demon'

in Kychanov (2006) for my last post, I found a number of interesting tangraphs with his radical B285 (= Li Fanwen radical 303) 'demon'

which is reminiscent of the right side of Chn 龍 'dragon' but may be derived from the right half of Chn 鬼 'ghost'.

The first tangraph listed by Kychanov is

TT3142 jɨu R2 1.2 'ghost; devil; demon; Skt raakṣasa'

with an obvious structure: 'death' + 'demon'.

But adding 'grass' to it results in

TT3252 xọ R73 2.62 'natural disaster; calamity'

What does 'grass' have to do with disasters?

Taking 'death' out of 'disaster' results in the second graph of the disyllabic word

tshia R20 2.17 lew R44 2.38 'revenge; magic'

The first graph has 'small' on the left, presumably as a phonetic; cf. the homophonous first half of the disyllabic word

tshia R20 2.17 ʒɨiw R47 1.46 'snub; deceive'

(Li Fanwen [1997: ]: 'bully; humiliate'; the second half can occur in other words)

with a vertical line plus 'person' (why?). Note that

TT1730 tsiə R31 1.30 'small; little; too'

is not read tshia, though its reading is similar.

Adding 'person' to the first half of 'revenge' results in the first half of

ʃɨụ R62 1.59 ʃɨew R46 2.40 'nit; louse'

(which doesn't have the 'bug' radical!)

The second half of that word is written as a combination of 'flesh', 'mouth', and 'finger'.

Adding 'demon' to 'mouth' and 'finger' creates the tangraph

TT0525 ŋwə̣w R61 1.58 'oath; swear an oath'

Other compounds of 'demon' also have inexplicable components.

Why does

TT5059 khwaa R22 1.22 'incantation'

have Nishida's radical 288 'heavy' (cf Chn 重) on top of TT3142 'demon'? According to Tangraphic Sea, that radical is an abbreviation of NIshida's radical 289 'master' taken from the center of

TT4296 ʔwɔ R74 1.71 'surround'

which has 'water' on the left and the mysterious filler ヒ on the right.

Why does 'blue/green' plus 'demon' equal the first graph for the disyllabic word

tshwiu R 1.3 naa R22 1.19 'vow; oath; to vow; to swear an oath'?

'Blue/green' is phonetic, as it appears on the left side of other tshwiu tangraphs, but 'blue/green' itself has a totally different reading:

TT5407 ŋwəʳ R90 1.84 'blue/green'

Note how 'demon' also appears on the right side of TT0525 'oath; swear an oath'. THE DENASALIZATION OF RIER 'HORSE'

Guillaume Jacques pointed out to me that

TT5233 rieʳ R79 1.74 'horse'

is probably cognate to gDong-brgyad rGyalrong mbro < *mraŋ (and hence to Written Tibetan rmang [obsolete], Written Burmese mraŋh, etc.). It may have been *Ci-raŋ in pre-Tangut with a presyllable conditioning raising and bending of *a to ie. At an intermediate stage it could have been *riẽʳ with a nasalized vowel.

I suspect that *riẽʳ could become less marked ̣in two different ways:

- by losing retroflexion and becoming riẽ R43

- by losing nasality and becoming rieʳ R79

The conflicting readings for

TT2387 'a kind of demon'

(cf. Old Chinese 靈 *reŋ 'supernatural'; OC 魑 *tɯ-hraj 'a kind of demon' lacks a final nasal)

implied by Homophones and Precious Rhymes of the Tangraphic Sea could reflect these two paths of development:

Homophones 47A44: riẽ R43 2.37

PRTS 2.15.2407: rieʳ R79 2.68

A related Tangut word for 'horse' only has a denasalized reading:

TT4789 riaʳ R87 2.74

One might expect another reading riã, but according to Gong ̣(2003: 605), nasal vowels "are found mostly in Chinese loanwords". Note that Gong's equivalents of my nasal mid vowels are oral:

Vowel type Rhyme Grade This site Gong
e (rhyme group VIII) R41 I -ẽ -əj
R42 II -ɛ̃ -iəj
R43 III -ɨẽ -jɨj
IV (Gong's III) -iẽ
o (rhyme group XI) R56 I -õ -ow
R57 II -ɔ̃ -iow
R58 III -ɨõ -jow
IV (Gong's III) -iõ

(I excluded tense and retroflex vowels to demonstrate the general pattern.)

So I would say that nonmid nasal vowels are mostly in Chinese loanwords. I presume that pre-Tangut had a complete set of nasal vowels that was reduced to only two types (ẽ and õ) when those Chinese words were borrowed. New nonmid nasal vowels were introduced to approximate Chinese vowel-nasal sequences or nasal vowels:

'monk': 僧 ?*səŋ or ?*sə̃ > TT4631 səĩ R15 1.15

'lute': 琴 ?*khɨm or ?*khɨ̃ > TT0637 khĩ R16 1.16

'body': 身 ?*ʃɨn or ?*ʃɨ̃ > TT3810 ʃɨĩ R16 1.16

'railing': 欄 ?*lɑn or ?*lɑ̃ > TT1298 lã R25 1.24

'mountain': 山 ?*ʃæn or ?*ʃæ̃ > TT1391 ʃæ̃ R26 1.25

'Zen': 禪 ?*ʃɨan or ?*ʃɨã > TT1356 ʃɨã R27 1.26

'winter' 冬 ?*towŋ or ?*tõw > TT4171 təũ R104 1.96

The surviving native nasal mid vowels also appeared in Chinese loanwords:

'sage': 聖 ?*ʃɨeŋ or ?*ʃɨẽ > TT3301 ʃɨẽ R43 2.37

'dragon': 龍 ?*lɨowŋ or ?*lɨõw > TT2559 liõ R58 1.56 and TT0704 liõ R58 2.49 INITIAL R- WITHOUT RETROFLEX RHYMES?

In Gong's Tangut reconstruction, initial r- can only precede retroflex rhymes with a handful of exceptions: the homophone group 47A31-A53.rjɨj R43 2.37, equivalent to my riẽ. This is strange because there is nothing special about R43 that would make it more rhotic-friendly than other lax vowel rhymes. How do we know that this group had initial r- and rhyme 2.37?

All but three members of the group were listed under rhyme 2.37 in Precious Rhymes of the Tangraphic Sea. The exceptions were:

47A34: not listed

47A51: not listed

47A44: listed under R79 2.68 (Gong's -jijr and my -ieʳ)

The inclusion of 47A44 R79 2.68 in a group otherwise consisting of R43 2.37 syllables may indicate that those two rhymes were phonetically similar. Hence reconstructions like Nishida's (1964) are improbable:

R43 2.37 -ǐẽ

R77 (sic!) 2.68 -ʊr

The two rhymes are closer in other reconstructions:

rhyme this site Gong Arakawa
R43 2.37 -iẽ -jɨj -yen
R79 2.68 -ieʳ -jijr -yeq'2

Arakawa regarded R79 as tense rather than as glottalized. I don't know what the -2 after -yeq' means. It does not indicate the tone which Arakawa wrote before initials: e.g., 2ryeq'2 = ?[rjẹʔ] with a second (= rising) tone. Nor does it contrast -yeq'2 with a -yeq' that is absent from his reconstruction.

In Precious Rhymes of the Tangraphic Sea, 47A44 is placed between two members of the R79 2.68 homophone group* 47B18-B43, 47B23 and 47B38.. 47B18, the first member of this group, was transcribed in Tibetan as re (cf. my reconstruction rieʳ). Thus one could reconstruct the entire group - including 47A44 which is implicitly a member according to PRTS - with initial r-**. And if 47A44 is nearly homophonous with 47A31-A53, then that group should also be reconstructed with initial r- since each homophone group is believed to have a single initial (though rhyme and tonal variation is permitted). I suspect that 47A31-A53 (except for 47A44?) riẽ was originally *riẽʳ. Other nasal vowel rhyme syllables may also have been retroflex in pre-Tangut. The twenty syllables ending in R97 -õʳ and the six syllables ending in R98 -iõʳ may be remnants of a much larger class of syllables with nasal retroflex vowels.

I was skeptical of nasal retroflex vowels until I discovered Kalasha (examples from Heegård and Mørch 2004):

Skt paaɳi 'hand' : Kalasha peʳ̃ 'palm of the hand'

Skt maɳi 'jewel' : Kalasha mãʳ(h)ĩʳk 'beads' (< maɳi-ka?)

Skt baaɳa 'arrow' : Kalasha bõʳ 'arrowhead; bullet'

Skt sthuuɳaa 'pillar' : Kalasha thũʳ 'id.'

Note that Kalasha nasal retroflexion originates from an earlier retroflex nasal ɳ, whereas Tangut nasal retroflexion presumably originated from nasal and rhotic segments in the same syllable: e.g., kõʳ R97 2.82 'tooth' may be from pre-Tangut *rkoNH.

*This group contains a single R79 1.74 syllable, 47B35. Other R79 1.74 syllables were listed in the following group (47B44-B48) in Homophones.

**Oddly, Nishida (1964: 136) noted that some member(s?) of this group (his group 16) transcribed Sanskrit le with l- instead of r-. This usage was not noted in his earlier section on transcriptional evidence for rhymes (1964: 62-63).

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2009 Amritavision