For almost three years, I've been reconstructing a lot of diphthongs in Tangut:

Basic vowel u i a ə e o
Grade I: (first half) nonhigh əu əi a ə e o
Grade II: lowered ʊ ɪ æ ʌ ɛ ɔ
Grade III: (first half) high; nonpalatal -ɨ- ɨu ɨi ɨa ɨə ɨe ɨo
Grade IV: (first half) high; palatal -i- iu i ia ie io

My reconstruction lacks the Grade II -i- of Gong's reconstruction that almost always corresponds to zero in Tibetan transcriptions (see below). Moreover, my reconstruction has a Grade III -ɨ- that generally corresponds to zero in Tibetan transcriptions (since the Tibetan script had no letter for -ɨ- or -ɰ-) whereas Gong's reconstruction has a -j-.

Gong's reconstruction (with vowels grouped in the same six columns to facilitate comparison)

Basic vowel u i a ə e o
Grade I: zero medial u e a ə ej o
Grade II: medial -i- (none) ie ia iej io
Grade III (= my Grade III and IV): medial -j- ju ji ja jij jo

However, my reconstruction still has a Grade IV -i- that generally corresponds to zero in Tibetan transcriptions. The following new reconstruction has far less -i-:

Basic vowel u i a ə e o
Grade I: (first half) lowered əu əi a ae ao
Grade II: lowered ʊ ɪ æ ʌ ɛ ɔ
Grade III: (first half) raised; nonpalatal u ɨ ɐ ə ɤ o
Grade IV: (first half) raised; palatal y i e ø

(The title of this post derives from a short-lived reconstruction of Grade III ə as rounded ʉ and Grade IV ə as unrounded ɨ. But Grade III ə was never transcribed with a Tibetan rounded vowel and the Grade III : IV distinction is not rounded vs. unrounded.)

Some problems:

1. The pattern of the four grades is extremely complicated.

2. There are no Tibetan transcriptions confirming any of the diphthongs: e.g., no Tibetan aHe or even aHi corresponding to ae, etc.

3. The diphthongs and are unusual. The former exists in Khmer and the latter might have once existed in Khmer as an intermediate stage between *aa and modern iə.

4. ɤ doesn't sound much like e and would be easily confused with ə.

Here are the most common Tibetan transcriptions. No Tangut vowel was consistently transcribed as yV.

Basic vowel u i a ə e o
Grade I u i a a/i/u/e/o i/e o
Grade II i e o/a
Grade III o
Grade IV


1. Grade I ə was transcribed as every Tibetan vowel, implying that it was none of them.

2. Grade II ə was transcribed as Tibetan i, which matches Gong's better than my ʌ. (But a correspondence between Tibetan i and my ʌ reminds me of how Middle Chinese was inexplicably borrowed as earlier Korean ʌ.)

3. Grade I e was transcribed as Tibetan i as well as e. Gong's ej matches both possibilities better than my e or ae.

4. Grade II o was transcribed as o or a which implies ɔ. There is one u transcription which I can't explain.

The challenge, in short, is to come up with a set of vowel distinctions that

- are not unnecessarily exotic and improbable (Occam's razor)

- were difficult for Tibetans to perceive: e.g., u i a e o were usually transcribed as u i a e o regardless of grade

- are internally consistent within a grade

- match what is known about Chinese grades (since Gong established that the Tangut grade system roughly matched the Chinese grade system)

I have yet to come up with a solution that fulfills all of the above criteria.

Here's an amalgam of my two proposals that remains unsatisfying:

Basic vowel u i a ɨ e o
Grade I: (first half) nonhigh əu əi a ə < əɨ e o
Grade II: lowered ʊ ɪ æ ɪ ɛ ɔ
Grade III: (first half) high; nonpalatal u ɨi ɨa ɨ ɨe ɨo
Grade IV: (first half) high; palatal y < iu i ia iɨ ie ø < io


1. The fourth basic vowel is now ɨ instead of ə which is regarded as a simplification of əɨ.

2. Grade II ɨ is dotless ɪ, the central counterpart of back ʊ and front ɪ. ɪ is a better match for its Tibetan transcription i than my earlier ʌ which was too low and back.

3. Grade III u is simply u. I thought it was strange that my earlier system had no simple u. However, u is an exception to the overall pattern of Grade III.

4. Grade III ɨ is parallel to Grade III u, but the former was transcribed three times as Tibetan a as well as i. My earlier reconstruction ɨə can account for both transcriptions.

5. Grade IV u and o were sometimes transcribed with Tibetan -y-. If the dialects of the Tibetan transcribers had no front rounded vowels, (y)u and (y)o would be appropriate transcriptions for Tangut y and ø. Perhaps the Tibetan transcriptions reflect a dialect of Tangut in which iu and io have monophthongized whereas the grade system was developed for a dialect which preserved -i-

6. There is not one instance of Grade IV a being transcribed as Tibetan ya or e. Hence both my reconstructions (ia and eɐ) are dubious. Maybe Grade IV ia fused into Grade II æ in the Tangut dialect that the Tibetans transcribed.

7. Grade IV ɨ is iɨ, which is a better match for its Tibetan transcription i than my earlier iə. However, can account for occasional Tibetan transcriptions with nonhigh vowels (a e o).

Maybe it's futile to come up with a single vowel system that matches both the Tibetan transcriptions and the grade system implied by Chinese transcriptive evidence on the one hand and the arrangement of rhymes into sets of three or four in Tangraphic Sea. The Tibetan transcriptions may represent a Tangut dialect in which the grade system was breaking down.

At the moment, I assume that the grade system was transparent in the Tangraphic Sea dialect. Hence I favor the first reconstruction above for that dialect with a small fix: lower mid central ɜ instead of lower mid back ʌ for Grade II ə. The Tibetan transcription i for that vowel may reflect a diphthong like ɜɨ with a raised second half (or lowered first half, depending on whether the original vowel was high or mid).

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2010 Amritavision