15.2.4.16:32: AN ARMY COMMANDER WHO WAS SMALL LONG AGO

(Small [smɔɰ] and ago [əʹgoɰ] almost rhyme in my dialect.)

Andrew West posted the second part of his series on two Tangut families. (Last month I wrote a response to his first part.) Part 2 concerns 小李鈐部 Xiaoli Qianbu (1191-1259) who was commemorated on a stele where his name appears as

3799 1141 1531 2805 2sew1 2li3 1ga4 2bu'4

His surname was also sinified as 昔里 Xili (*Sili in Old Mandarin; the characters are phonetic symbols normally meaning 'long ago' and 'village' or 'Chinese mile'). Andrew explained the variation:

A more plausible explanation is that their family name was originally Tangut, pronounced something like Sili [which would be meaningless in Chinese], and that after Xiaoli Qianbu moved away from Hexi he sinified it to the similar-sounding [and meaningful] Chinese characters xiǎolǐ 小李 ['small plum'; *sew li in Old Mandarin], and invented the story of his ancestors being Shatuo Turks who had been given the Tang royal family name [Li].

I think there is another possible explanation. The alternation of *sew and *si in the Old Mandarinizations of his name remind me of how Tangut -ew often appears as -i in Tibetan transcriptions. If Tibetans heard -ew, they could have transcribed it as -eHu (which is how they transcribed similar Chinese rhymes a couple or so centuries earlier). Yet the Chinese transcriptions of Tangut in the Timely Pearl clearly indicate a final *-w: e.g.,

0100 1lew1 'one' (which shares rhyme 44 with 3799 2sew1 though the tones are different)

transcribed in Tibetan as kli (x 3), gli (x 1), gliH (x 1)

transcribed in Chinese as 婁 *1lew1

The Tibetan and Chinese transcriptions reflected two kinds of Tangut dialects, one that shifted -ew to -i* and another that retained -ew. Xili and Xiaoli may be Sinifications of the same surname in those two types of Tangut dialects:

(*2sew1 2li3 >) 2si1 2li3 > Old Mandarin 昔里 *sili (= modern standard Mandarin Xili)

2sew1 2li3 > Old Mandarin 小李 *sewli (= modern standard Mandarin Xiaoli)

Perhaps one Tangut dialect was Xiaoli's native dialect and the other was the prestige dialect.

This kind of double transcription still exists today: e.g., someone may be known in English as both Wong and Huang, Cantonese and Mandarin versions of the same family name 黄.

As for 鈐部 Qianbu (Old Mandarin *kem pu), it too has variants: 甘卜/敢不 *gam bu and 紺部 *gam pu. The *-mb-* and *-mp- sequences reflect the prenasalization of the initial b- [mb] of

2805 2bu'4 'to command'

Andrew proposed that 鈐 *kem has a double function: it means 'to stamp a seal' (and is hence reminscent of Qianbu's Mongolian title daruγači which could be interpreted as 'presser'; see part III of my series on the title) and also sounds like

1531 1ga4 'army'

plus the [m] of 2bu'4 [mb ...]. (I am not certain about how the rhyme -u'4 was pronounced.) The Old Mandarin vowel *e may reflect the frontness of Tangut Grade IV a which may have been low front [a] or [æ] (as opposed to Tangut Grade I a which may have been low back [ɑ]).

Andrew interpreted

2893 2khwe1 'great'

following Xiaoli Qianbu's name on the stele as a noun 'great man' (i.e., 'official') rather than as an adjective (which was my interpretation).

In either case I agree with Li Fanwen (2008: 474) who regarded it as a loan from Chinese 魁 *1khwe1 with a similar semantic range: 'great' ~ 'great man'. The native Tangut word for 'great' is

4457 2leq3

which may be cognate to Old Chinese 大 *lats 'big' and 太 *l̥ats 'great'.

*A more exotic possibility is something like -iɰ with an unrounded glide coda (cf. Sofronov's -eɯ for rhyme 44). -ɰ had no equivalent in either Tibetan or Chinese and would be ignored in transcriptions.

However, I would not be surprised if -w was simply lost, as Tangut had already lost all other codas. A Tangut dialect that lost -w would have a simple C(w)V syllable structure (assuming that the mysterious quality indicated by ' was not a coda).

15.2.3.23:27: AN AGNOSTIC APPROACH TO TANGUT GRADES

Perhaps the most distinctive feature of the complex version of my Tangut transcription system is my use of numbers for grades. I got the idea from Arakawa Shintarō, who uses syllable-final numbers (e.g., the -2 of his -eq2) to distinguish between similar rhymes (but not grades):

Note there is no Grade I -eq in Arakawa's transcription corresponding to his Grade I -eq2.

I also use syllable-final grade numbers in my notation for Tangut period northwestern Chinese for ease of comparison with Tangut.

In theory one could even rewrite other reconstructions using grade numbers to further facilitate comparisons: e.g.,

Numerical notation enables us to look at reconstructions in terms of categories rather than phonetic details. All three generally* agree on the categorization of rhymes 17 and 18, but disagree on whether 19 and 20 are the same grade (Gong), variants of the same grade (Arakawa), or different grades (this site).

I have changed my mind about the phonetic interpretation of the grades several times, but I am almost certain that Grade II did not have -i- or -y-. Grade II tangraphs were not used to transcribe Sanskrit CyV-syllables. As far as I know, the only Grade II tangraph used to transcribe Sanskrit was

3144 1pa2 (= Gong's 1pia and Arakawa's 1pya)

and it stood for pā, not pyā. Moreover, there was another transcription for Sanskrit pā which was Grade IV:

3425 1pa4 (= Gong's 1pja and Arakawa's 1pa:)

That tells me Grade II syllables had some phonetic quality without any Sanskrit parallel. -2 indicates that quality without specifying what it was. For years I thought that quality was lowering and backing and more recently I thought it was a medial -ɤ-, but now I am not so sure. That quality originated  at least in part from  pre-Tangut medial *-r-, but Tibetan transcriptions of Grade II syllables do not contain any r, so *-r- must have become something else in Tangut by the early first millennium.

As for the other grades, I once thought that Grades I, III, and IV were characterized by (partly) lowered vowels, medial -ɨ-, and medial -i-. However, my reconstruction could not account for facts such as these:

- Sanskrit i was transcribed as

0932 i3 (which I reconstructed as ʔɨi, a diphthong that does not exist in Sanskrit) instead of i4 (which I reconstructed as i).

- Sanskrit -a and -ā were often transcribed with -a4 (which I reconstructed as -ia) as well as -a1 (which I reconstructed as -a): e.g., Skt pā (not pyā!) as 3425 1pa4 (see above).

- Similarly, Sanskrit -e was often transcribed as -e4 (which I reconstructed as -ie) instead of -e1 (which I reconstructed as -e).

Until I figure out how to reconstruct the grades in a manner that fits what is known about the Tangut transcription of Sanskrit - and the Tibetan transcription of Tangut and the Chinese grade system that influenced traditional Tangut phonological analysis - I will use numbers for grades in lieu of specific phonetic symbols.

*Arakawa (1997: 128-129) distinguished between two types of rhyme 17: -a and -a2 (whose 2 does not indicate a grade). There are a few minimal pairs such as

3755 1tsha 'mixed' (< Chn 雜): 5969 1tsha2 'hollow bag' (in Arakawa's notation)

which are not distinguished in Gong's or my notation. 3755 and 5969 had different fanqie in Tangraphic Sea and were in widely separated homophone groups in Tangraphic Sea (see 23.212 and 25.122 in Andrew West's online edition) and Homophones (29B31 and 35A58). I should indicate this type of rhyme-internal distinction somehow in my database: e.g., as -a1a and -a1b. The problem with using -a and -b is that it may imply that all rhymes ending in the same letter have something in common. Does 5969 1tsha1b have something in common with, say,

4917 1tshu1b 'shovel'

which was somehow different from

0916 1tshu1a 'conceited'?

Arakawa's notation does not distinguish between those two syllables which have different fanqie and are not homophonous in Homophones (32B16 and 33B37). In Tangraphic Sea, 4917 1tshu1b immediately follows 0916 1tshu1a (see 6.171 and 6.162 of Andrew West's edition), implying that the two may have been similar in a way that 3755 1tsha1a and 5969 1tsha1b were not.

15.2.1.23:57: TANGUT PHONETIC DATABASE VERSION 1.0

1586 1ghiq2 'sound'

I have been posting reconstructed Tangut readings on my blog for years, but I have never provided a complete list of readings until now. Strictly speaking, the forms in version 1.0 of my Tangut phonetic database (downloadable in three formats: .htm / .pdf / .xls ) are transcriptions rather than phonetic reconstructions. The letters and numbers symbolize phonetic distinctions but do not necessarily precisely represent them.

The database has sixteen columns:

A. LFW: Tangut character numbere from Li Fanwen's 2008 dictionary. L stands for Li Fanwen. Characters L5995-6000 from his 1997 dictionary are also included as L5995a-L6000a.

B. Simple: Transcription for lay use. No tone numbers (see column F), -q (see column K), apostrophes (see column M), or grade numbers (see column N).

C. Complex: Full transcription for scientific use. Includes tone numbers (see column F), -q (see column K), apostrophes (see column M), and grade numbers (see column N).

Question marks indicate missing information: initials, vowels, and grades.

D. Class: The nine initial consonant types from Homophones and/or Mixed Categories of (the Precious Rhymes of the) Tangraphic Sea:

Voiced obstruents may have been prenasalized: e.g., b- may have been [mb], etc.

The Class II consonant v- may have been [w], [ʋ], or [v]. There may have been another Class II consonant f- which is not included in the database. Five characters with h(w)- according to Tangraphic Sea fanqie are listed as Class II in Homophones. They may have had initial f-.

There is no consensus on the reconstruction of Class IV. This database tentatively follows Gong and has two Class IV initials which are identical to Class III n- and Class VII j-. Li Fanwen (1986) reconstructed a distinct palatal nasal. Nishida (1964) and Arakawa (1999) reconstructed a series of Class IV consonants.

Class VII consonants may have been retroflex [tʂ tʂʰ dʐ ʂ], alveopalatal [tʃ tʃʰ dʒ ʃ], or palatal [tɕ tɕʰ dʑ ɕ].

Class VIII glottal stop is not written in either the simple or complex transcription. The absence of an initial consonant in the transcription (a 'zero initial') indicates an initial glottal stop [ʔ].

What appears to be Initial w- in the transcription is actually a Class VIII zero initial plus medial w-sequence indicating the glottal stop-glide cluster [ʔw], not a simple glide [w].

Class VIII fricatives may have been velar [x ɣ] or glottal [h ɦ]. They may have had uvular allophones [χ ʁ].

Class IX lh- may have been voiceless [l̥] or a lateral fricative [ɬ].

Class IX z- may have been a lateral fricative [ɮ].

Class IX zh- may have been retroflex [ʐ], alveopalatal [ʒ], or palatal [ʑ].

Tai (2008) has made a convincing case for a sixth Class IX consonant ld- on the basis of Tibetan transcriptions of Tangut. I have not yet included this consonant in my database. It may have been a lateral affricate [dɮ]. I speculate it may have a voiceless counterpart lt- [tɬ].

E. Rhyme: The 105 rhymes of Tangut without regard for tones (see column F).

F. Tones: Translations of Tangut tone names which are in turn translations of traditional Chinese tone names. The names may not have described the contours of the tones: e.g., the 'level' tone may not have been level.

0. Tone unknown. (There is no tone zero.)

1. 平聲 'Level' tone.

2. 上聲 'Rising' tone.

(3. Reserved for the Tangut equivalent of the Chinese 去聲 'departing' tone. No tone 3 syllables are in the Tangut primary sources that I have on hand.)

4. 入聲 'Entering' tone. Rare tone in eleven syllables from a partial page of the Precious Rhymes of the Tangraphic Sea. More examples must have been on the rest of the page.

G. Subrhyme: Rhyme number taking tone into account. For example, subrhyme 1 of tone 1 is the first tone 1 rhyme, etc. There are 97 tone 1 subrhymes and 86 tone 2 subrhymes. Most rhymes have two subrhymes, but some have only one: e.g., rhyme 6 has no tone 2 subrhyme and rhyme 23 has no tone 1 subrhyme.

H. Initial consonant: See the list for column D above.

I. Medial consonant: -w- or zero. When no consonant precedes w- in the transcription, w- represents the cluster [ʔw], not [w].

J. Vowel: There are six vowel symbols:

These may have symbolized diphthongs and/or glide-vowel sequences in different grades (see column 14). Heights and other qualities may also have varied by grade.

With one exception (see below), the symbol y represents a nonfront, nonlow, unrounded vowel: e.g., [ɨ ɘ ə ɯ ɤ ʌ], etc. Cf. the use of y to romanize Russian ы.

Rhyme 105 -wya may have been [ɥa]. w and y respectively indicate the labiality and palatality of the glide [ɥ].

K. Cycle: zero, -q-, or -r-.

Tangut rhymes can be grouped into at least three cycles. The term 'cycle' refers to a somewhat fixed sequence of vowels within each cycle (ideally u, i, a, y, e, o). I have adopted Gong's groupings:

1. The basic cycle (rhymes 1-60). Unmarked.

2. The tense cycle (rhymes 61-76) indicated with -q (which is not a consonant). Tension is often indicated by a subscript dot in Tangutological literature, but Arakawa's symbol -q is easier to type and read. There is no agreement on the ending of the tense cycle: e.g., Sofronov ended it at rhyme 75, and Arakawa ended it at rhyme 79.

3. The retroflex cycle (rhymes 77-103) indicated with -r (which indicates vowel retroflexion and is not a consonant). Nearly all r-syllables are in the retroflex cycle. I note exceptions in column P.

Kychanov and Sofronov also recognize a fourth cycle (rhymes 99-105).

Rhymes 104-105 are like cycle 1 rhymes: i.e., they are neither tense nor retroflex. Kychanov and Sofronov regarded them as part of the fourth cycle.

L. Nasal/w:

Final -n indicates nasalization of the previous vowel and is not a final nasal [n].

Final -w is a glide [w]. It is in complementary distribution with -n, so I have combined the two into one column to save space. Historically both are traces of earlier codas: -n is from earlier nasals and -w is from *-k as well as *-w.

M. Prime: An apostrophe serving as an easily typed substitute for the prime symbol representing an unknown distinctive phonetic quality. Gong interpreted this quality as vowel length, though there is no correlation between prime and vowel length in Tangut transcriptions of Sanskrit.

N. Grade: Although grades are not explicitly identified in the Tangut phonological tradition to the best of my knowledge, correlations between certain Tangut rhymes and the four grades of Chinese traditional phonology have been recognized for over a century. These correlations have been interpreted as evidence for a Chinese-like system of grades in Tangut. There is no agreement on the number of grades or their phonetic qualities. Here I generally follow Gong's classification with two exceptions:

- I regard rhymes 4 and 6 as Grade II, not Grades I and III.

- I divide Gong's Grade III into Grades III and IV

There are strong but not absolute correlations between consonant classes and grades. The typical pattern is as follows:

I list exceptions to this pattern in the notes (column P).

O. Variant: Number of main character entry in Li Fanwen (2008) for a variant. E.g., the main entry for L0486 is L0017.

P. Notes (which I was unable to fit into the PDF version):

1. All instances of Class VIII h-syllables listed as Class II in Homophones.

2. All initial consonants followed by rhymes with unexpected grades.

3. R-syllables outside the retroflex cycle.

4. Miscellaneous and not comprehensive:

- Possible duplicate character (L6000a).

- Uncertain initial (L3358).

- Typos in Li Fanwen 2008 (L3975, L4718)

- Loan from/transcriptions of Chinese from L5995a onward

- Ghost characters in Nevsky and Sofronov.

- Important and interesting Sanskrit transcription characters.

- Variant second character for the Tangut autonym Tibetanized as Mi-nyag (L0831).

- Chinese phonetic approximations for L2583 (possibly indicating a -w- not in Gong's reconstruction) and L2955

*2.2.6:16: The Chinese term 牙音 'tooth sounds' literally translated into Tangut as

0039 1586 2korn1 1ghiq2 'tooth sound'

refers to velars which are not pronounced with the teeth.

The Chinese and Tangut terms for classes VII and VIII contain different words for 'tooth': 齒 and

0169 1shwi3 'tooth'.

**2.2.6:50: Also known as 流風音 'flowing wind sounds', the name used by Nishida and Arakawa. I have not seen any Tangut term for Class IX containing a word meaning 'flow'. The term I am familiar with is

2302 5425 1586 1ly3 2zheq3 1ghiq2 'wind (l-) and zhe-sounds'

which is a Tangut equivalent of Chinese 來日音 'l- and zh-sounds'.

1ly3 'wind' may tell us that the Tangut thought laterals sounded like the wind.

2zheq3 is a transcription character without any meaning.