According to various secondary sources (Sofronov 1968 II, Li Fanwen 1986, 1997, 2008, Shi et al. 2000), Tangut 1449 'servant' has rhyme 2.40, yet its fanqie in a primary source (Mixed Rhymes of the Tangraphic Sea) indicates it has rhyme 2.80 -or1:


1449 2chhor1 = 5886 1chhon2  'to steal' + 1237 2vor1 'calf'

Perhaps the incorrect rhyme 2.40 originated with Sofronov (1968 II: 375) who may have been the first to identify the rhymes of the final speller 'calf' and its homophone

0387 2vor1 'the surname Vor'

as 2.40 -i̭eɯ (= my -ew3, not my -or1). Was 4 a typo for 8? In any case, the Precious Rhymes of the Tangraphic Sea lists 'calf' under rhyme 2.80. (Unfortunately, 'servant' is in the Mixed Categories section of the Precious Rhymes of the Tangraphic Sea which is organized by initial class* and not by rhyme, so its rhyme has to be inferred from its fanqie.)

Gong regarded rhyme 2.80 as Grade I. Normally Grade I rhymes do not follow Class VII initials like chh-, yet 'servant' has the unusual combination chh- + 2-or1 even though it is not a known loanword or transcription character (i.e., a character for an un-Tangut syllable).

I would expect Grade II/III/IV rhyme 2.81** (as listed in Kychanov and Arakawa 2006: 289) after chh-. Gong's reconstruction in Li Fanwen's 1997 and 2008 dictionaries contains rhyme 2.81 -jwor (= my -or2 and -or3), even though it is next to the rhyme number 2.40 for Gong's -jiw (= my -ew3).

Could the fanqie of 'servant' be an error that tells us rhymes 2.80 and 2.81 were similar? Or did 'servant' truly have an exceptional reading 2chhor1 instead of 2chhor2 or 2chhor3?

*2.1.2:18: 'Servant' is in the Mixed Categories section for syllables with tone 2 and Class VII initials (ch-, chh-, j-, sh-). For some unknown reason, j-syllables are only in Mixed Categories and not the other two volumes of the Precious Rhymes of the Tangraphic Sea. A random syllable with a Class VII initial in Mixed Categories is likely to have j-. Perhaps that is why Sofronov (1968 II: 298) reconstructed 'servant' with initial ndź- (= my j-). But 5886 'to steal', the fanqie initial speller of 'servant', has chh-. That initial can be confirmed by the Tibetan transcription chi [tɕʰi] for

3465 1chhi3 'meat'

the fanqie initial speller of 'to steal'. I do not know why 'servant' with chh- was placed in Mixed Categories instead of the Precious Rhymes of the Tangraphic Sea volume for the second tone.

*2.1.2:45: Grade II rhymes like -or2 are normally distinguished from Grade III/IV rhymes like -or3 and -or4. I do not know why they were grouped together:

Rhyme 96 1.91 1-or2 1-or3 1-or4
2.82 2-or2 2-or3 2-or4

Such clustering may indicate that Grades II-IV were more phonetically similar to each other than to Grade I: e.g., in Gong's reconstruction, Grade II had medial -i- and Grade III (= my Grades III and IV) had medial -j-. There are no rhymes combining Grade I with other grades.

There are minimal pairs of rhyme 1.91 syllables such as


4798 1mor2 'transcription character'*** : 1543 1mor4 'true'

which Gong interpreted as a distinction between Grade II mior and Grade III mjor (he did not recognize Grade IV). There are no such minimal pairs of rhyme 2.81 syllables, though that is probably due to chance. Generally there are fewer tone 2 syllables than tone 1 syllables in a rhyme category. (Nonetheless there are a few rhyme categories consisting solely of tone 2 syllables: e.g., there are no tone 1 syllables with rhyme 23.)

***2.1.2:51: According to the Tangraphic Sea, 4798 1mor2 is for transcribing mantras. However, Grade II tangraphs are almost never used in transcribing Sanskrit. That implies Grade II had a phonetic quality absent from Sanskrit. A rare exception is the transcription character

0310 2mu2

for Sanskrit mū. What Sanskrit syllable was 4798 1mor2 devised to write? 單于 *DAR-UGHA? PART V: THE SPOUSE SUPREME

Having just mentioned the Central Asian title qaɣan and Old Chinese (OC) 烏 *qˤa 'crow' in my last post, I thought I should write something about 閼氏, the transcription of the Xiongnu title for the wife of the supreme ruler (單于; see parts I-IV below).

The phonetic of 閼 is 於 *qa 'in', a near-homophone of  烏 *qˤa 'crow';  both 烏 and 於 are drawings of crows). Baxter and Sagart (2014: 40) reconstructed 閼 as *qˤat and 氏 as *k.deʔ. It is tempting to conclude that *qˤat k.deʔ transcribed the Xiongnu prototype of the Central Asian title qatun 'queen', particularly if *k.d- had been simplified to *d- in the Han Dynasty.

However, the late Eastern Han scholar 蘇林 Su Lin phonetically glossed 閼氏 as 焉支 *ʔɨen kie, implying an alternate earlier OC reading *ke for 支. Guangyun (1008) lists the Middle Chinese reading *tɕie < OC *ke for 氏 in 閼氏. The modern Mandarin reading yānzhī for 閼氏 is a reflex of Su Lin's *ʔɨen kie. (OC *qˤat k.deʔ would have become Mandarin *èshì.)

Moreover, Old Chinese *q- had already shifted to a glottal stop by the time 閼氏 first appears in the Records of the Grand Historian (c. 109 BC) in which 'Alexandria' was transcribed as 烏弋山離. If 烏 transcribed a foreign word with a zero initial, then 閼 (which shared an initial with 烏 *qˤa) must also have transcribed a foreign word with a zero initial (or initial glottal stop or, less likely, ʕ-), then Former Han 閼氏 *ʔˤat deʔ / ke could not have transcribed a Xiongnu word like qatun (or even *χatun).

Would alternate readings of 閼 help salvage the qatun hypothesis? 閼 has four Middle Chinese readings in Guangyun:

MC *ʔɑt  < OC *qˤat (my *qat, Zhengzhang's *qaːd)

MC *ʔɨat  < OC *qat (my *Cɯ-qat, Zhengzhang's *qad)

MC *ʔɨen  < OC *qran or *qren (my *Rɯ-qan or *Rɯ-qen, Zhengzhang's *qran)

MC *ʔen  < OC *qˤen (my *qen, Zhengzhang's *qeːn)

The OC readings other than OC *qˤat are not in Baxter and Sagart (2014) but are my attempts to project the MC readings back into OC using their system.

The last two MC readings are listed for 閼氏 in Guangyun.

All readings would have had an initial glottal stop during the Former Han. *ʔˤat, *ʔat, *ʔran  ~ *ʔren, *ʔˤen.

The final *-t or *-n could have been an attempt to transcribe Xiongnu *l. Vovin (2002: 392) identified 閼氏 as a transcription of the Xiongnu cognate of Proto-Yeniseian *ʔal/r'it 'wife'.

If Old Chinese still had *-r (and the transcription of 'Alexandria' may indicate otherwise'*), then the use of a *-t/-n graph instead of an *-r graph points to *l instead of *r in the Xiongnu word. Otherwise, *-t/-n could indicate either *l or *r in Xiongnu.

Other segments may be problematic for the *ʔal/r'it hypothesis:

Proto-Yeniseian *a does not match *e in the OC ancestors of the readings prescribed by Guangyun. But maybe the first Xiongnu vowel was higher than *a: e.g., *æ, *ɛ, or even *e.

The final *-t of the Proto-Yeniseian word matches the *d- of 氏 *deʔ but not the *k- of the reading *ke that would be the source of the readings later specified by Su Lin and Guangyun.

Moving on to semantics, did Xiongnu elevate 'wife' to 'empress' (cf. how Proto-Indo-European *gʷēn 'woman' was elevated to English queen), or did 'empress' become 'wife' (cf. the downward shift of status of English lady)? Or was the word 'wife' in Xiongnu and interpreted as 'empress' by the Chinese?

If the Xiongnu title transcribed as 閼氏 is not the source of qatun, where did qatun come from? I am not convinced that the Xiongnu title transcribed as 護于 is the source of qaɣan (see part IV), but I do favor Vovin's (2007: 184) speculation that qatun is qan 'ruler' with a Rouran feminine infix *-tu-. Alas, as Vovin noted, "[a]lmost nothing is known about the Ruan-ruan [= Rouran] language", so that hypothesis cannot be tested.

(The above statement does not rule out a Yeniseian origin for qaɣan and/or qan. Perhaps there was a Xiongnu phrase 'great ruler' that sounded like qaɣan, but that phrase was unrelated to the Xiongnu word underlying 護于 and was not an official title in the Xiongnu language [or it somehow escaped the notice of the Chinese]. The Rouran could have been the first to make qaɣan a title.)

*Baxter and Sagart (2014) reconstructed 山 corresponding to -xan- in 'Alexandria' with final *-r. I expressed my doubts here. But if they are correct, then Early Old Chinese *-r must have shifted to *-n in the Former Han dialect underlying 烏弋山離 unless the transcriber misheard 'Alexandria' as 'Alexaria'. 單于 *DAR-UGHA? PART IV: QA-UGHA

I thought of one more reason not to reconstruct Old Chinese 單于 as *dar-ʔu-ɣa. If 護于 'crown prince' (with the same second character) is a transcription of the Xiongnu prototype of the Central Asian title qaɣan, then it is unlikely that 護于 was *ʁwah-ʔu-ɣa, unless a trisyllabic Xiongnu word was compressed into a disyllabic word in Turkic and Mongolic.

Moreover, if Vovin (2007: 180) is correct, qaɣan is from the Xiongnu cognate of Proto-Yeniseian *qɛʔ 'great' plus the Xiongnu cognate of Proto-Yeniseian *qʌ̄j ~ *χʌ̄j 'ruler' with allophonic medial voicing and a Mongolic suffix -n; no *u would be expected in such a compound (unless it was a Xiongnu linking vowel). Vovin (2007: 181) suggested that medial *-w- could have been lost in 'ruler' during the two millennia separating the Xiongnu from the modern Yeniseian languages.

Vovin regarded 'ruler' as being the second half of 單于 whose first half he identified as the Xiongnu cognate of Yeniseian tɨr-words for 'lower reaches of the Yenisei, north'. (He did not provide a Proto-Yeniseian source for those words.) He translated 單于 as 'Ruler of the North' (as opposed to the Chinese in the south).

That brings to mind two questions.

First, is there another system of titles in which the 'ruler of a region' outranks a 'great ruler'?

Second, how did 'great ruler' come to outrank 'ruler of the north' (which may have been the prototype of early Turkic tarxaːn and Written Mongolian <darqan>) if not Written Mongolian <daruγ-a>?

A third question is why neither 護于 nor 單于 contain Han Dynasty uvulars corresponding to the q of *qʌ̄j, qaɣan and <darqan>. By that time, the original Old Chinese uvular stops had been lost, and new uvulars had developed from velars: e.g.,

*q- > *ʔ- (e.g., 烏 'crow'; more here)

*Cʌ-k-, *k- + lower vowel > *q-

*ɢʷ- > *ʁw- (e.g., 護)

*Cɯ-ɢʷ- > *ɣw- (e.g., 于)

The first certain attestation of qaɣan in Chinese transcription has such secondary uvulars: 可寒 *qʰɑˀʁɑn ~ *qʰɑˀɢɑn (Book of the Later Han, 5th century AD; Baxter and Sagart reconstructed 可 and 寒 with velars in Old Chinese, though I think 可 may have been uvular all along).

Did Xiongnu *q weaken to *χ? Does the variation in later forms of the titles reflect borrowing and transcription before and after the shift or from different dialects of Xiongnu? Even if Xiongnu had both *q and at different times and/or in different dialects, that still doesn't explain the voiced initial of 護于 *ʁwah-ɣwa.

I noted on Tuesday that 單于 had type B syllables corresponding to what are type A syllables from a Chinese perspective in <daruγ-a>. tarxaːn and <darqan> also have type A syllables. How did a Xiongnu BB word become an AA word in 'Altaic'? Conversely, if the word were AA in Xiongnu, why was it transcribed as AA in Chinese?

護于 has a different kind of mismatch; 護 is type A and 于 is type B, but qaɣan is a sequence of type A syllables (as is its later Chinese transcription 可寒). Was the Xiongnu title harmonized in Turkic and Mongolic? Harmonizing (*AB > AA) is understandable (and is the key to my Old Chinese reconstruction), but a complete flip-flop (*BB > AA) is not. 單于 *DAR-UGHA? PART III: THE OPPRESSOR

I thank Andrew West for pointing out the biggest problem with my already discredited Old Chinese 單于 *dar-ʔu-ɣa idea: the fact that Written Mongolian ᠳᠠᠷᠤᠭ᠋᠎ᠠ <daruγ-a> 'chief' has an internal etymology: it is from the verb <daru-> 'to press' plus a suffix <γ-a> that "expresses an unfinished action which started in the past and continues into the present" (Poppe section 362; the hyphen in transliteration indicates a break in the traditional Mongolian script, not a morphemic boundary). If <yabu> is 'to go' and <yabuγ-a> is 'someone who started going and is still going', then a <daruγ-a> is 'someone who started pressing and is still pressing': i.e., an oppressor (Dashdondog 2010: 105).

Although I assume the Xiongnu empire included Mongolic speakers, it is unlikely that its non-Mongolic rulers would adopt a title for their supreme ruler from the Mongolic language of the ruled. Therefore the resemblance of 單于 *dar-ɦwa (as reconstructed by Baxter and Sagart 2014: 260) and <daruγ-a> is coincidental.

Pulleyblank (1963: 257) thought that <daruγ-a> might have been a direct borrowing from Xiongnu (unlikely for the reasons I just stated unless the derivation from 'to press' is a folk etymology) unlike early Turkic tarxaːn (borrowed into Written Mongolian as <darqan>; see Clauson 1972: 539-540) which might have been an indirect borrowing (perhaps via the Rouran). If the two words are unrelated, perhaps the Turkic form is the true successor of 單于. Although Turkic t- does not match Old Chinese *d-, Written Mongolian <darqan> may be from an unattested Turkic variant with *d-. (Clauson listed no Turkic forms with d-.) 單于 *DAR-UGHA? PART II: THE XIONGNU SON OF HEAVEN

Last night I left out one other argument why 于 could not have been something like *ʔu-ɢa in Old Chinese. If Late Old Chinese *-w- had developed from *Cu-presyllables, I would expect a lot of word families with zero ~ *-w-alternations like those of Tangut: e.g.,

1671 1ne < *Cɯ-ne 'red' : 2765 1nwe < *Pɯ-ne 'to turn red'

See Gong (2002: 45-46) for more examples. However, Gong noted that "such examples are exceptional in Chinese": e.g.,

熱 Middle Chinese *ɲiet 'hot' : 爇 Middle Chinese *ɲwiet 'to burn'

Baxter and Sagart (2014: 47) reconstructed this pair as *C.nat and *not which are equivalent to *Cɯ.nat and *Cɯ.not in my system*. What was originally an alternation of root vowels became an alternation of medials; there is no need to explain it in terms of a *Cu-presyllable: e.g.,

*Cu.not > *Cɯ.not > *ɲwiet

*Cu.nat > *Cu.nwat > *ɲwiet

There is only one zero ~ *-w-alternation in phonetic series 0097 于: 荂 has two Middle Chinese readings, *kʰwæ and *xɨə. All other 于-graphs have *w. Is it likely that all 于-readings were once *Cu-Qa (later becoming Middle Chinese *(C)w-syllables) with the sole exception of *Cɯ-qʰa which became *xɨə? It is simpler to reconstruct 0097 as a *Qʷ-series and regard the second reading of 荂 as an outlier which may be from a Middle Chinese dialect whose ancestor lost *-w- under certain conditions.

Worse yet, I know of no word families with zero ~ *-w-alternations written with phonetic series 0097: e.g., there is nothing other than the transcription 單于 suggesting that 于 'to go' had a root *ɢa. And there is no guarantee that the Xiongnu word underlying 單于 was like its much later Written Mongolian descendant ᠳᠠᠷᠤᠭ᠋᠎ᠠ <daruγ-a> 'chief'; 于 could have transcribed Xiongnu *ɣwa with *-w-.

I would not be surprised if Xiongnu had a cluster *ɣw reduced to <γ> in Written Mongolian because Xiongnu *Tr- in the full title of the supreme ruler

𢴤黎孤塗單于 *Tˁraŋ rˁi kˁwa lˁa dar ɣwa 'heaven son chanyu'

was simplified to <t> in Written Mongolian ᠲᠩᠷᠢ <tngri> 'heaven'.  (The initial of 𢴤 could have been *tˁr-, *tʰˁr-, or *dˁr-. See Pulleyblank 1963: 241.)

The initial of 孤塗 *kˁwa lˁa, the Chinese transcription of the Xiongnu word for 'son', was probably [q], as Early Old Chinese *q had become a glottal stop by this point (as in Egyptian Arabic). Could Ket qalek' 'younger son, grandson' be cognate to Xiongnu *qwala? As Pulleyblank (1963: 245) wrote,

Being a word for a fundamental human relationship, it ['son'] is unlikely to be a loanword in Yenisseian and unless it is an extraordinary coincidence it creates a presumption that the Hsiung-nu [Xiongnu] belonged to that language group.

*Given that the phonetic of 熱 and 爇 is 埶 *ŋ̊et-s (= my *Cɯ-ŋ̊et-s or *Hɯ-ŋet-s) with a velar nasal, perhaps their root could also be reconstructed with a velar nasal as *ŋjat ~ *ŋjot. The *-j- would condition the palatalization of *ŋ-. There is no *-j- in Baxter and Sagart's 2014 reconstruction. 單于 *DAR-UGHA?

It would be interesting to reexamine all of the Late Old Chinese transcriptions of Xiongnu words from Pulleyblank's seminal 1962-1963 article on Old Chinese consonants.

Baxter and Sagart (2014: 260) reconstructed one of those words as

單于 Han *dar-ɦwa 'Xiongnu ruler' (< Early Old Chinese *[d]ar + *ɢʷ(r)a)

The word is better known by its modern Mandarin pronunciation chanyu (shanyu in Giles' 1892 dictionary). It survives in Written Mongolian as ᠳᠠᠷᠤᠭ᠋᠎ᠠ <daruγ-a> 'chief'. (Has the word been found in Khitan?):

Three notes:

1. Both 單 and 于 are type B syllables. Norman (1994) proposed that the type A/B distinction in Old Chinese was one of pharyngealization: type A syllables were pharyngealized, and type B syllables weren't. Norman drew parallels with 'Altaic' languages, and I have wondered if the phenomenon spread from Chinese to 'Altaic', as it is reconstructible at the Middle Old Chinese level (or even at the Early Old Chinese level according to Baxter and Sagart): i.e., it existed in Chinese during a period long before the first attested 'Altaic' language.

It is commonly assumed that Xiongnu was a typologically 'Altaic' language. If it was, and if 單于 was an accurate transcription of the Xiongnu word for 'ruler', then that word should have contained type B (i.e., nonpharyngealized syllables) in Xiongnu as well as in Late Old Chinese. However, Written Mongolian <daruγ-a> contains the equivalents of type A syllables. Did the word shift from type B to type A during the many centuries separating 單于 and <daruγ-a>? Or is 單于 simply an inaccurate transcription?

2. Did 單 still end in *-r during the Han Dynasty? According to Starostin (1989), *-r became *-n in what he called 'Classical Old Chinese' prior to the Han Dynasty. I have not yet seen any Han Dynasty rhyme evidence for *-r. Nonetheless there is independent evidence for *-r-retention in some dialect(s) at a much later date: e.g., Muromachi Period (!) Japanese soroban still resembles 算盤 Old Chinese *[sˁ]orʔ-s [bˁ]an 'abacus'; the *-r- corresponds to *-n- in prestigious Middle Chinese *swanʰ ban from centuries earlier. Perhaps *-r was retained in nonprestigious dialects on the margins: e.g., the presumably eastern dialect that was the source of soroban. The transcription 單于 could be based on a northern dialect of Chinese spoken near the Xiongnu state.

3. Written Mongolian has <uγ> corresponding to *ɦw (or *ɣw?) in the Chinese transcription. The labial and fricative segments are in the opposite order.

Unlike Baxter and Sagart, I don't think Old Chinese originally had a distinction between pharygealized and nonpharygealized uvulars. Their nonpharygealized *ɢʷ- corresponds to my *Cɯ-ɢʷ- (details here). *ɯ is my cover symbol for an unknown high vowel. What if 于 was *ʔu-ɢa which later underwent metathesis?

Old Chinese *ʔu-ɢa > *ʔu-ɣa > *ʔu-ɣua > *ɣua > Middle Chinese *wuo

Then 單于 would be *dar-ʔu-ɢa (or ...ɣa?) which would be a near-perfect match for Written Mongolian <daruγ-a>.

There are several problems with the metathesis hypothesis:

First, Baxter and Sagart (2014) did not reconstruct presyllables with *ʔ-. Perhaps some of their *C- are glottal stops, but that remains to be confirmed.

Second, according to my theory of the origins of the type A/B distinction, 單 should also have a presyllable since I reconstruct them to shift *a-syllables from type A to type B:

Old Chinese *Cɯ-dar > *Cɯ-dɨar > *dɨan > *dian > *dien > Middle Chinese *dʑien > Mandarin chan ~ shan

Yet such a presyllable would correspond to zero in Written Mongolian <daruγ-a> which I assume did not undergo apheresis.

One possibility is that the voicing of the initial is secondary and is from a nasal prefix that was absorbed before 于 lost its presyllable:

*Nɯ-tar > *Nɯ-tɨar > *Ntɨar > *dɨar

The normal reading of 單 was *Cə.tˤar (= my *Cʌ.tar) with *-t-.

The trouble with this scenario is that the presyllable had to condition vowel warping before its unstressed vowel was lost. Did the Xiongnu word have a diphthong like *ɨa in its first syllable?

Another possibility is that dental initials like *d- were normally type B unless a conditioning factor (a low-vowel presyllable?) was present (cf. how I think uvulars were normally type A unless a high-vowel presyllable was present).

Third, the labiality of 于 'to go' was probably in the root if it is cognate to Written Tibetan Hgro 'to go' and Written Burmese ကြွ <krva> 'to go (honorific)' (Gong 1994: 81) or to wa-type words for 'to go': e.g., Tangut

0676 1ve3 'to go' < *CE-wa(ŋ)

So the Old Chinese root initial must have been labiouvular rather than simply uvular.

Could Written Mongolian <ruγ> be a simplification of a Xiongnu cluster like *rɣw that was faithfully reproduced in the Chinese transcription? TANGUT RHYME 41: SMALL PERSON, BIG PROBLEM

I have long been troubled by my reconstruction of Tangut rhyme 41

3798 1tsen1 'small' (which looks like 'person' + 'small' but is from 'few' + 'small')

because the evidence points in different directions:

1. Internal evidence

There are only fourteen known rhyme 41 tangraphs in eight homophone groups: seven in the first ('level') tone volume of the Tangraphic Sea and one with dz- in the Mixed Categories volume:

Source Homophone group Circle Reading
Tangraphic Sea 'level' tone volume 1    1phen1
2 1den1
3 1ben1
4 ○   1tsen1
5 1len1
6 1ven1
7 ○  1lwen1
Mixed Categories of the Tangraphic Sea 8 1dzen1

The low frequency of this rhyme and the absence of a 'rising' tone counterpart *2-en1 imply that it could not have been something simple like -e.

The circles divide some but not all homophone groups. The reasoning for the implied grouping of, for instance, 1phen1 and 1den1 as distinct from 1ben1 is unknown. (I would understand if 1phen1 and 1ben1 were grouped together since both had Class I [i.e., labial] initials.

The coexistence of v- and l- with dz- indicates that rhyme 41 must have been Grade I. They do not normally coexist in any other grade:

Grade v-, l- dz-

(The table above only shows the general pattern. There are exceptions.)

Tangut rhymes normally form sequences with the following order:

- Grade I-IV V

- Grade I-IV V'

- Grade I-IV Vn

Rhyme 41 is where I would expect 1en1:

- Rhymes 34-37: -e1, -e2, -e3, -e4

- Rhymes 38-40: -e1, -e2, -e3/e4

- Rhyme 41: -en1?

- Rhymes 42-43: -en2, -en3/en4

Yet as we will see, no other evidence supports a nasal vowel. (My -n indicates nasalization; it is not a consonant.)

2. Chinese transcription evidence

As I already noted in my entry on line 104 of the Golden Guide,

1720 1ven1

was a transcription character for the Chinese Grade III (not I!) surname 隗 *2wi3 in the Tangut translation of Sunzi.

1720 also transcribed Chinese Grade I  外 *3wai1 in Sunzi and the Timely Pearl and Grade III (not I!) 偉 *2wi3 in the Ganying Pagoda inscription.

Sofronov (1968 II: 30) listed 1720 as a transcription of Chinese Grade I 磑 *1we1 ~ 3we1.

None of those sinographs would have been read with nasal vowels in the dialect known to the Tangut (unless the nasality of *ŋ-, the former initial of 外 and 磑, spread to the vowel - but 隗 and 偉 never had nasal initials).

No other rhyme 41 tangraphs were used to transcribe Chinese to the best of my knowledge.

In the Timely Pearl,

3798 1tsen1 'small'

was transcribed as 栽 *1tse1, which may indicate that the nasality of rhyme 41 was lost by the end of the 12th century in the author's dialect. I do not know of nay other

3. Tibetan transcription evidence

All three Tibetan transcriptions known to me lack nasals:

dwi, dwe (Nishida 1964: 53; frequency of each unknown; note the absence of -w- in the reconstructions!)

-eH (Tai 2008: 216; initial consonant unknown)

4. Sanskrit transcription evidence

As far as I know, rhyme 41 tangraphs were never used to transcribe Sanskrit. That implies rhyme 41 was unlike anything in Sanskrit: e.g., it was not short i, long ī, or long e [eː] with or without a following nasal or nasalization. (Sanskrit has no short e.)

5. Comparative evidence

Guillaume Jacques (2014: 186) compared

3798 1tsen1 'small'

to Japhug xtɕi < 'id.' without a nasal, but noted the unexpected initial correspondence (cf. Somang kə-ktsî 'id.' with ts-). He derived rhyme 41 from pre-Tangut *-ij without a nasal. Is it possible that Japhug and Somang lost a final nasal?

6. Conclusion

For the time being, I weigh the internal evidence over the external evidence and write rhyme 41 as 1-en1, but I remain uneasy about nasality.

An alternative is to follow Gong and write rhymes 41-43 with -i or -y instead of -n:

41. -ei1 (instead of -en1; cf. Gong's -əj)

42. -ei2 (instead of -en2; cf. Gong's -iəj)

43. -ei3/-ei4 (instead of -en3/-en4; cf. Gong's -jɨj)

I would then change my -on rhymes to -ou or -ow (cf. Gong's -ow) so that there are no nasalized mid vowels.

However, I do not understand why Gong reconstructed final glides in those rhyme groups. GSR 0289

In line 104 of the Golden Guide, the Tangut character


may have transcribed the Chinese surname 薛 which was pronounced something like *4se4 in the dialect known to the Tangut) and as ɕie in the modern northwestern dialect of Xi'an. Those readings lack a labial segment present in the modern standard Mandarin reading Xuē [ɕye]. The [y] of Xuē corresponds to nothing in the prestige Middle Chinese dialects preserved in Chinese traditional phonological sources and in the reading traditions of Japan, Korea, and Vietnam:

Middle Chinese *siet

Phags-pa Chinese ꡛꡦ <see>

Sino-Japanese setsu

Sino-Korean sŏl; idealized Middle Sino-Korean syə́rʔ (this y is IPA [j], not [y])

Sino-Vietnamese tiết < *siət

Baxter and Sagart (2014)'s reconstructions of Grammata Serica Recensa series 0289 also lack labial segments:

GSR Sinograph Old Chinese Old Chinese (this site) Middle Chinese (this site) Mandarin Gloss
0289a *s.ŋat *sɯ.ŋat *siet xuē to control, correct, govern
0289d-e spec. of plant; place name (i.e., where the plant grows?)
0289g *ŋ(r)at * *Cɯ.ŋ(r)at
(was *C- = *s-?)
*ŋɨet niè concubine’s son
0289j (shoots from) tree stump

(Karlgren [1957: 89] wrote,

The alternation s- :ng- in this series is probably a trace of some Archaic [i.e., Old Chinese] initial consonant combination.

and Baxter, Sagart, and I would agree.)

Hence I thought the [y] of 薛 Xuē [ɕye] might be a local Mandarin innovation, but it isn't. Forms with labial vowels coexist with ɕie-type forms in all branches of Chinese: e.g.,

Jin: 并州 Bingzhou ɕieʔ (lit.), ɕyəʔ (colloq.)
Wu: 常山 Changshan ɕyʌʔ

Hui: 祁門 Qimen syɐ̆

Gan: 都昌 Duchang siol

Xiang: 雙峰 Shuangfeng ɕya (lit.), se (colloq.)


Southern: 雷城 Leicheng soi

Northern: 石陂 Shibei sye

Yue: 高要 Gaoyao sit (new), syt (old)

Ping: 南寧 Nanning ɬyt

Hakka: 惠州 Huizhou syet

Unclassified languages also have a mix of labial and nonlabial forms: e.g., in 富川 Fuchuan, the 七都 Qidu dialect has si but the 八都 Badu dialect has suɐi.

I don't see any obvious pattern here. In Bingzhou the labial form is colloquial (i.e., likely to be native or at least from an earlier layer of borrowing), but in Shuangfeng, it is literary. (I suppose that if the labial form is an innovation, it must originate outside Shuangfeng.) Labiality is so widespread that it must have been present in some earlier prestige dialect(s), albeit not those recorded in the mainstream phonological tradition.

Nonlabial forms once (?) existed in Beijing Mandarin itself. Giles (1892: 449) listed under the reading* xiē and gave xiě (with a different tone and xuē as alternate readings. Are xiē and xiě now extinct? Do any 薛 families today call themselves Xiē and Xiě?

*I converted Giles' romanization to pinyin for ease of comparison.

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