Five years ago I rediscovered 村田治郎 Murata Jirō's 1957 book on the inscriptions of the Cloud Platform at 居庸關 Juyongguan¹ in the University of Hawaii library. I had last borrowed it around 1996. Of course my attention was drawn to the Tangut inscription. But, I confess, not for long. Soon after that I dove into the world of Tangut's distant relative Pyu. And I've been there for four years.

Then yesterday Andrew West reawakened my interest in the Juyongguan inscriptions.

Today I was looking at the Tangut inscription at Juyongguan, and the Tangut transcription


5300 3639 2770 4620

1ty4 2rer4 2lo1 1ka4

of Sanskrit trailokya 'three worlds' jumped out at me. I've used Trailokya as part of my long pen name for maybe twenty-five years now.

A few words on the transcription characters:

𘎤 5300 1ty4: The only consonant clusters possible in native Tangut words had -w- as their second element. So one strategy for transcribing Sanskrit consonant clusters was to break them up into CyC-sequences. Tangut y was a neutral vowel, and in Grade IV (indicated by my -4) it was something like [ɨ] or [ɯ].

𗣀 3639 2rer4: Tangut had no [aj]. Guillaume Jacques (2014: 206) does not even reconstruct *-aj at the pre-Tangut level. I am guessing pre-pre-Tangut *-aj became pre-Tangut *-ej (which Guillaume does reconstruct) and then Tangut -e.

Here's a possible example:

𘞪 5356 1teq4 < *Sɯ-taj² 'single' could be cognate to Jingpho tāi and Boro otay, part of a cognate set that Matisoff (2003: 262) glosses as 'single/one/whole/only'³.

(I finished the rest of the entry on 5.4.15:39, added a footnote on 5.6.19:06, and then failed to save the finished page. What follows is a new second half from 5.6.19:39.)

𗥹 2770 2lo1: For a long time, I used to think that Tangut tones might actually be phonations: tone 1 was the default phonation and tone 2 was the marked (creaky or breathy?) phonation. But the phonation hypothesis predicts that Sanskrit would be transcribed solely using Tangut characters for syllables with tone 1. There would be no reason to transcribe Sanskrit with Tangut characters for syllables with tone 2: i.e., a phonation that did not exist in Sanskrit. However, most Sanskrit Co-syllables⁴ were transcribed with Tangut characters for syllables with tone 2 (Arakawa 1999: 111).

Tone 1
Tone 2
Both tones

co, jo

to, do

pho, bo, mo

yo, ro

śo, ho

Why was tone 2 favored for Sanskrit Co-syllables?

Conversely, why was ko transcribed with a Tangut character for a syllable with tone 1?

And was there a reason to transcribe the remaining Sanskrit syllables with Tangut characters for syllables with both tones? For instance, was there something about the -lo- of trailokya that necessitated tone 2, whereas the lo in some other word was somehow different to Tangut ears and required the tone 1 character 𗓽 4710 1lo1?

𗡝 4620 1ka4: This character transcribed both Sanskrit ka and kya. Why not transcribe Sanskrit kya as ky ya (cf. 1ty4 2rer4 for trai above) or as a fanqie character for kya combining  part of a kV-character with part of a ya-character? Perhaps 1ka4 was something like [kja]. But if Grade IV (written here as -4) was characterized by [j], why could 1ka4 also represent Sanskrit ka? Was there no simple [ka] in Tangut? Were Grade I and II ka something other than [ka]: e.g., [qɑˁ] and [qɑʶ] like Middle Chinese *1ka1 and *1ka2? Why was there no Grade III ka?

Chinese and Tangut grades seem to be similar. So if the Middle Chinese transcription of Sanskrit ka was 迦 *1ka3, I would expect the Tangut transcription to be 1ka3 - a syllable that does not exist in Tangut!

To complicate matters, Grinstead (1972: 144) says 4620 could represent Sanskrit ke. 1ka4 must have sounded like Sanskrit ke as well as ka and kya. Maybe it had a front vowel: [kjæ]? 

¹This name was built into Windows 10's pinyin IME. It's interesting to see what's in and out of the IME.

Sometimes more annoying than interesting. For instance, the common character 家 jia 'house' isn't listed as a choice for jia. I've been typing 家族 jiazu 'family' and deleting the second character to type 家 jia.

At least 波 bo 'wave' is included as a choice for bo now. I recall having to type the wrong reading po to make it display in some older version of the Windows Mandarin IME. I just noticed that the bopomofo IME accepts both bo and po for 波 bo 'wave'.

²(Pre-)pre-Tangut *S- conditioned Tangut -q (my symbol for vowel tension) and pre-)pre-Tangut *-ɯ- (perhaps a front or back high vowel like *-i- or *-u- in pre-pre-Tangut) conditioned Grade IV.

³Matisoff (2003: 262) does not gloss the Jingpho and Boro forms.

⁴Many Sanskrit Co-syllables are absent from Arakawa's data: e.g., kho, gho, cho, jho, ṭo, etc.

⁵Arakawa (1999: 111) accidentally omitted the rhyme and first tone of 𗓽 4710 1lo1, the other Tangut transcription character for Sanskrit lo in his table.

Tangut Yinchuan font copyright © Prof. 景永时 Jing Yongshi
Tangut character image fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2019 Amritavision