The last tangraph (Tangut character) in last night's entry on ovine characters was

2969 2ŋəʳ 'sheep, goat'

It was unique among tangraphs for sheep, but it has a near-lookalike among tangraphs for goats:

1-2. 1189 1104 2gwã 2te  'wild goat' (Li Fanwen 2008; Nishida 1964: 'a kind of sheep')

The top component of those two tangraphs

is shared with 121 other characters without any single common phonetic or semantic denominator: e.g.,

0745 2vɨe 'the surname syllable Ve'

0900 2vɨi 'under, below, bottom'

0901 2nɤẽ 'dirt, filth'

0911 2kɑ 'arduous, difficult, hard, tough'

0916 1tshu 'conceited, restrained'

The analysis of 1104 is unknown, but 1189 is derived from the top of 1422 1sy' 'deer' plus all of 5181 1tsɑʳ 'beast, animal':


5181 in turn has a circular derivation with 5167 1ɬa 'deer':


Why derive 'goat' or 'sheep' from 'deer'?

The center component of 2969 'sheep, goat' and 1104, the second half of 'sheep' or 'wild goat', is shared with


3. 2367 1tshə 'goat' =

left and center of 3454 2tshe 'sheep' +

right of 1neʳ 'wild animal'

whose right side has the same source as the left side of


2969 2ŋəʳ 'sheep, goat' (and probably  the second half of1189 1104 2gwã 2te  'wild goat' or 'a kind of sheep' as well)

3454 2tshe may be from 2367 1tshə plus *-j.

3454 and 2367 share their left and center components with

4. 3768 2tʂɨụ 'goat, lamb' (analysis unknown)

whose right side is in only two other tangraphs:

1852 2tʂɨụ 'to be able, to dare' (analysis unknown)


3805 1tʂɨụ 'that' =

right of first half of 5883 2tsha 1ʐɨiw 'to bully and humiliate' (the second half is 'to invade' by itself) +

right of 3768 2tʂɨụ 'goat, lamb' (analysis unknown)

Since all three are nearly homophonous, their shared component

is clearly phonetic. OTHER OVINE TANGRAPHS

Yesterday I wrote about the tangraph (Tangut character) for the calendrical term 2mø 'sheep':


Given that tangraphy (the Tangut script) is supposed to be largely semantic, it is surprising that the other ten 'sheep' tangraphs do not share any single element with each other or 2mø:

Tangraph Li Fanwen number Reading Gloss Notes Type
3452 2ʔe sheep left side shared with 2550 2ʔe 'banquet'; phonetic A
3470 1mə 1tʂɨa 1mʌ 'patron god of sheep' left side phonetic; 1mʌ is homophonous with 3513 1mʌ and may be an adjective 'heavenly'
1588 1tʂɨa B
3454 2tshe sheep left side phonetic/semantic?; also in
2367 1tshə 'goat'
2910 1tshə 'lamb'
3127 1tshə (second half of 2ɮəʳ 1tshə  'waterfowl'; 2ɮəʌʳ is 'water', so is the word literally 'watergoat' or 'waterlamb'?)
5959 1pʌʳ 'lamb'
0939 2tẹ right side shared with 2910 1tshə 'lamb'
sounds like a possible *S-prefixed derivative of the second half of 1189 1104 2gwã 2te  'wild goat'
5557 2ga old sheep (only in dictionaries) left from 2910 1tshə 'lamb' (semantic)
right from 0590 2ɮø̣ 'longevity' (semantic)
5095 1nɪʳ one-year-old sheep (only in dictionaries) looks like 5557 plus a right-hand component, but analyzed as:
left from 3452 (semantic; above)
center from 0590 2ɮø̣ 'longevity' (semantic)
right from 1996 2de 'to drink'  (why?)
5004 1gyʳ no semantic components shared with 5095 in spite of being a synonym
top and bottom right from 5007 2gyʳ 'to lie down' (phonetic)
bottom left from 0276 2no 'child' (semantic)
4925 1nwew six-year-old sheep (only in dictionaries) top and bottom left from 4971 1ʂwɨi 'year' (semantic)
bottom right from 3194 ɬʌ (tone unknown) 'full' (semantic)
2969 2ŋəʳ sheep, goat left from 0558 1neʳ 'wild animal' (semantic; also at right of 2367 1tshə 'goat')
right from
all of 1153 1dʐə 'skin' (semantic; according to Precious Rhymes of the Tangraphic Sea) or
the second half of 1189 1104 2gwã 2te  'wild goat' (semantic; according to Combined Homophones and Tangraphic Sea)

Types A-C have right-hand components similar but not identical to the right side of 5504:




The significance of the different top parts is unknown.

Types B-D share a component

(coincidentally?) resembling Chinese 羊 'sheep' with 48 other tangraphs which lack ovine semantics: e.g.,

0079 ~ 0676 1vɨe 'to go'

0123 dʐɨi (tone unknown) 'to go'

0561 2dʐɨo 'to help' (< Chn 助)

0787 1kɑ̃ 'to expel' (< Chn 赶)

Type E tangraphs have a 'horned hat' in common, but according to Tangraphic Sea analyses, that 'hat' has two different sources:

=+ 5095 1nɪʳ 'one-year-old sheep' = 5007 2gyʳ 'to lie down' + 0276 2no 'child'
=+ 5004 1nwew 'six-year-old sheep' = 4971 1ʂwɨi 'year' + 3194 ɬʌ (tone unknown) 'full'

Type F is unique.

1.3.1:05: Or is it? I will look at 'goat' tangraphs next. A EWE-NIQUE COMPONENT?

2015 is the year of the 2mø 'sheep':

whose tangraph (Tangut character) consists of two halves of unknown function:


The second half is so rare that I don't think it's in my Tangut radical fonts. I thought it was unique to 5504 'sheep', but it's also in one variant of 3007 2Taʳ 'net' (dental initial unknown):


The first variant of 3007 with 亠 at the top right has a unique right side:

Are these rare right-hand components single units or compounds whose parts are abbreviations of other tangraphs?

=(< ?)+(< ?)?

=(< ?)+(< ?)?

The bottom half of those components, Kychanov and Arakawa's radical B299, is in 23 tangraphs (their 5096-5118). They have no obvious single phonetic or semantic common denominator, and none sound like 5504 2mø 'sheep' or 3007 2Taʳ 'net', so B299 cannot be phonetic in those two tangraphs.

B299 looks like the right-hand version of Li Fanwen (1987) radical 71 (in Andrew West's numbering):

which is in non-right-hand positions in eight tangraphs:

Tangraph Li Fanwen number Reading Gloss Notes
0083 1vɪ dragon semantic component in 4234 (below)
1188 2ŋɑ egg only in dictionaries
4234 1vɪ dragon plant longan (< Chinese 'dragon eye'); 'dragon' with 'wood' on top
5441 2swi mother-in-law center and right from 1188 'egg' (above)
5496 2bə cheek variant of 5510 which has 'flesh' on the left instead of LFW radical 71; I assume 'flesh' is original
5529 2ʂwɨi the surname Shi share a common phonetic
5530 1ʂwɨi first half of 5530 4761 1ʂwɨi 1ʂwɨa 'in a soft low voice'
5546 2ʂwɨi second half of 3278 5546 1tshi 2ʂwɨi 'food'

Complimentary distribution is no guarantee of allography. Conversely, the facts that

- none of the LFW 71 tangraphs appear in known analyses of B299 tangraphs

- none of the B299 tangraphs sound like ʂwɨi or (except maybe 5696 1bɛ 'smallpox')

does not necessarily rule out a relationship between two: e.g., the two radicals might be interchangeable in tangraphic analyses that have not yet been (re)discovered. (If only we could see what Nevsky saw before his texts were taken away!)

Nishida radical 241


the left side of 2mø 'sheep', can appear in initial, medial, and final position in 109 other tangraphs. None of the 109 others sound like 2mø, so radical 241 must be semantic and be an abbreviation of some other tangraph ... unless it is a Tangut B phonetic component in 5504.

Next: Other ovine tangraphs. THE GOLDEN GUIDE: LINE 103: TANGRAPHS 511-515

103. I can't remember what I was supposed to blog about next, so I'm going to stick with tradition and finish the year with a line from the Golden Guide. Three lines to the last surname ...

Tangraph number 511 512 513 514 515
Li Fanwen number 3601 0914 4305 3801 0775
My reconstructed pronunciation 2khew 1gɤe' 1tsũ ?kø̃ 1gy
Tangraph gloss opening, entrance a place name Ge the surname 宗 Zong (*tsũ) 'ancestor' second half of 1kə ?kø̃ 'dung beetle' transcription character
Word the surname 寇 Kou (*khew) the surname 崖 Ya (*ŋgɤe) the surname 姜 Jiang (*kø̃) the surname 虞 Yu (*ŋgy)
Translation Kou, Yi(ng?), Zong, Jiang, Yu

511: 3601 is a borrowing of Tangut period northwestern Chinese 口 *khew 'mouth' (albeit with a specialized meaning), but its character is a much more complex semantic compound:


3601 2khew 'opening' =

left of 2627 2lɨə̣ 'earth' +

all of 4866 1kɤe 'incomplete' (why?)

The presumably semantic function of 4866 is unclear. Would 'complete earth' be solid and lack an opening?

512: I presume the sources of 0914 somehow describe Ge, wherever that was: e.g., Ge was high and its earth looked as it had been kneaded by the gods.


0914 1gɤe' 'a place name Ge' =

top of 1166 1tswə 'to knead, rub' +

right of 0464 2so 'high'

Oddly 0914 appears as a transcription of Chinese 櫻 *ʔĩ in the Timely Pearl, which may be why Nie and Shi (1995) regarded 0914 as a transcripiton of the homophonous Chinese surname 嬰.

513: I presume 5635 is phonetic in 4305 despite the mismatch in initial voicing:


4305 1tsũ 'ancestor' =

top of 4250 1si 'wood' +

all of 5635 2dzu, second half of 2me 2dzu 'steed'

Unlike other 'horse' words,

1379 5635 2mɛ 2dzu 'steed' (only in dictionaries; a 'ritual' word?)

was written without the 'horse' radical:

Why wasn't that radical in all equestrian tangraphs?

514: The analysis of 3801 is unknown, but it looks like a transparent compound of the radicals for 'insect' and 'small':


I am uncertain about the tone of 3801. It is listed in the second tone volume of Tangraphic Sea, even though its placement in Homophones implies that it is a first tone tangraph.

515: 0775 is a straightforward semantophonetic compound:


0775 1gy 'transcription character' =

left of 1586 1ɣɤị 'sound' +


Mentioning Marathi व्ह् <vh> last night reminded me of a topic I've had in mind since I briefly mentioned Verner's Law two weeks ago: the segmental traces of accent in Avestan.

Avestan has an h before rk and rp that doesn't correspond to any consonant in other languages: e.g.,

'wolf': A hrk-; cf. Sanskrti vṛ́ka, Russian volk

'body': A hrp-; cf. Latin corpus, Sanskrit kṛ́p- 'beauty'

Moreover, Avestan sometimes has a (< *hrt?) where an rt is expected: e.g.,

'battle': A ana-; cf. Sanskrit pṛ́tanā

Did Avestan preserve an h - a laryngeal? - lost elsewhere in Indo-European? No, it turns out that h and were conditioned by an accent preserved by Sanskrit but unwritten in Avestan:

*ṛ́ > *ə́r > əhr before *k, *p

*ṛ́t > *ə́rt (> *əhrt?) > əṣ 

*r also became hr (and merged with *t to become ṣ) after other vowels: e.g.,

*árt (> *ahrt?) > aṣ in maya- 'man'; cf. Sanskrit mártya-

See chapter III, section 2 in Beekes (1988) for problematic cases in which Avestan accent does not agree with Sanskrit accent; the Sanskrit accent may not necessarily be conservative.

What is the phonetic motivation for those changes? I suspect hr was voiceless [r̥]. Why would *r devoice after a stressed vowel?

Similarly, why would Germanic spirants (e.g., þ) remain voiceless after an originally accented vowel but become voiced elsewhere?

'brother': *bhréʕtēr > Gothic broþar; cf. Sanskrit bhrā́tar-

'father': *pʕtḗr > Gothic fadar; cf. Sanskrit pitár-

If the Proto-Indo-European accent was a high tone, there may have been a correlation between high tones and voicelessness vaguely reminiscent to that found in Cantonese. However, higher Cantonese tones are associated with historically* voiceless initials that precede them, whereas Indo-European high tones were associated with voiceless consonants that followed them. Moreover, the Cantonese correlation involves all historically voiceless consonants, whereas the Indo-European correlations only apply to a limited number of consonants - just one in the case of Avestan! Lastly, there is one other accent-related shift in Avestan that does not involve voicing - or does it? *h (< *s) became before y and an accented vowel:

'to do homage': nəmax́ya-; cf. Sanskrit namasyá-

Could the spirant sign transliterated as have been voiced [ɣʲ] (assimilating to voiced palatal y [j]), just as Germanic spirants voiced before accented vowels?

*Modern Cantonese has many voiceless initials before vowels with low tones. Those initials are historically voiced: e.g., 唐 tʰɔːŋ ˨˩ <  *dɑŋ 'Tang (Dynasty)'. WHY IS ENGLISH V BORROWED AS <VH> IN MARATHI?

I was surprised to see that the Marathi version of the name Silver is सिल्व्हर <silvhar> with an -h- absent from Hindi सिल्वर  <silvar>. Similarly, Victoria is व्हिक्टोरिया <vhikṭoriyā> in Marathi but विक्टोरिया <vikṭoriyā> in Hindi, so vh is not just in medial position. And I found a few Google results for ग्रूव्ह <grūvh> 'groove' with -vh in final position.

I have been transliterating Marathi in angle brackets because I don't know how it's pronounced.

Wikipedia's article on Marathi phonology distinguishes between /ʋ/ <v> and murmured /ʋʱ/ (presumably written <vh>). (The latter is in a table of consonants but not explictly listed as a phoneme.)

On the other hand, Pandharipande (2003: 701) listed only /w/ <v> in his table of the Marathi alphabet. (There is no separate table of phonemes.)

In either case, Marathi <vh> is a combination of the approximant <v> (/ʋ/ or /w/) which sounds like English v and <h> which is a fricative like English v. Is <vh> pronounced [v], particularly in the speech of Marathi-English bilinguals?

Does <vh> also occur in native Marathi words, and if so, what is its origin?

(12.29.0:49: Yes. The Wikipedia entry for Marathi mentions केंव्हा  <keṃvhā> 'when'.

12.29.1:23: As for the origins of <vh>, I found the following Sanskrit-Marathi correspondences in Turner's A Comparative Dictionary of Indo-Aryan Languages:

apahelā : <avher> 'disrespect'

deva- + ghara- : <devhārā> 'shrine'

yavākāra 'barley-shaped' : <dzavhār> 'millet'

lauha- + aṅgika 'iron-bodied' : <lavhã̄gī> '?"

vibhīdaka : <vheḷā> 'a kind of plant'

śubha- + kāla- : <sovhāḷā> 'festive occasion'

<vh> has four types of sources in that list:

1. lenited stop + h

2. v + lenited aspirated stop (e.g., gh or bh)

3. v or bh + lenited k (cf. how *k became a source of aspiration in Korean and probably also in Tangut)

4. u + h

But there are other <vh> that seem less motivated: e.g., Sanskrit laba (without any aspirates or k) 'quail' became Marathi <lavā> (with b-lenition as I would expect) and <lavhā> with an irregular <h>.)

I once knew a native Marathi speaker. I wish I could ask him about <v> and <vh>! CAN ENGLISH C BE FROM TH?

I was surprised to read in Wikipedia that Eustace is equivalent to both Greek Ευστάχιος <Eustákhios> and Ευστάθιος <Eustáthios>. I suppose Greek kh was borrowed as [k] and then softened to [s] before a front vowel that was later lost in speech (though not in writing). But is there any other case of English c corresponding to Greek th? Maybe correspond is the wrong word. The kh ~ th variation goes back to Greek. Can it be explained in terms of phonological differences between dialects? Or is it a random change that occurred as the legend of the saint spread? Did the th-version make it to England? Was there ever an English form like French Eustathe that went out of use in favor of Eustace?

Judging from Wikipedia article titles, the Catholic/Protestant world favors the kh-version of the name, whereas the Orthodox world favors the th-version:


Catalan Eustaqui

Czech Eustach (but the entry for the given name is titled Eustác with c [ts]!)

Dutch, Finnish, German, Hungarian, Norwegian, Swedish Eustachius (as in Latin)

French Eustache (but note Eustathe also exists; see above)

Italian Eustachio

Polish Eustachy

Portuguese Eustáquio

Spanish Eustaquio


Bulgarian Евстатий < Evstatij>
Romanian Eustațiu

Russian Евстафий <Evstafij> (Greek th was Russianized as f)

Serbian Евстатије <Evstatije>

It is not surprising that English Eustace follows the Western pattern.

Tangut fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2014 Amritavision