I've been overlooking the most obvious example of -ya: R21 corresponding to a rhyme with a coda: the disyllabic name

Arakawa mI: nya:

my miɨ nɨaa


is the Tangut equivalent of Written Tibetan mi-ɲag. The WT name must have been borrowed before R21 lost its final consonant (*-k? *-ɣ? *-g? *-q?), since there is no reason to regard WT -g as a suffix.

3.8.0:10: I've already mentioned the other example of Arakawa -ya : WT -ag that I know of:

'black': Arakawa nya: R21 1.21 : WT nag-po ARAKAWA'S -YA: : PROTO-RGYALRONG *-AQ

I forgot to mention last night that Arakawa's R21 -ya: (without a final apostrophe = glottal stop) corresponded to -aʁ rhymes in gDong-brgyad rGyalrong (Jacques 2006):

'black': Arakawa nya: R21 1.21 : DBG kɯ-ɲaʁ < Proto-rGyalrong ?*ɲaq

'deep': Arakawa nya: R21 1.21 : DBG kɯ-rnaʁ < Proto-rGyalrong ?*rnaq

However, Arakawa's -ya:' R24 with a glottal stop also corresponds to DBG rhymes with codas:

Arakawa mya:' R24 1.23 'scar' : DBG tɯ ɣmas 'wound'

but note Written Tibetran rma 'wound' (not rmas)

Arakawa mya:' R24 1.23 'fruit' : DBG kɯ mat 'id.'

Could one conclude that

pre-Tangut *-aq > ya:

pre-Tangut *-as, *-at > ya:'

DBG -aʁ also corresponds to other Arakawa open-vowel rhymes:

-a R17

-a: R19

-a: R20

-ạ R66 (but there is no -ạ', so perhaps *-' > Ø after tense vowels)

with one exception:

Arakawa la:' R22 1.22 'thick' : DBG kɯ jaʁ 'id.'

This counterexample prevents me from claiming that all earlier *-aq became Tangut -(y)a(:). (I will not address the problem of the conditioning factors for -y- and vowel length.) I could deal with it in two ways:

Archaism: 'Thick', being a common word, retained its final glottal stop after other *-aq words had lost it in a chain shift:

*-as, *-at > *-aʔ > *-a

Final cluster: 'Thick' had a suffix *-s (why? what would its function be?) in pre-Tangut which caused it to develop like an *-as word:

'thick': *laqs > *las > la:' R22 1.22

'wound': myas > *mya:' R24 1.23

cf. how Old Chinese *-aks and *-as both have the same Middle Chinese reflexes

I prefer the first explanation since it does not require an ad hoc suffix with no known function. AN INTRIGUING I-ÑI-TIAL CORRESPONDENCE

TT4479 'black' R21 1.21

which I think would be nya: in Arakawa's reconstruction has an initial ny- [nj] that is even closer to the ɲ- of

Qiang (-x ~-q is secondary and not from very early *-k [Huang and LaPolla 1996: 375])

Mawo ɲiq

Taoping ɲi ɲi

Ronghong ɲix ~ɲiq


Proto-rGyalrong ?*ɲaq

Japhug kɯ-ɲaʁ

Zbu kə-ɲɐ̂χ

(no Somang or Cogtse cognates)

than my nɨ-. A shift of *ɲ- to is unusual and without any precedent as far as I know. So perhaps Arakawa's ny- (and Gong's equivalent nj-) are correct.

A Tangut word nearly homophonous with 'black'

TT3186 R21 2.18 (not 1.21!) 'dung; excrement'

was transcribed in Tibetan as nya. Tibetan -y- does not directly confirm the palatal glides of Arakawa or Gong's reconstructions, since it could also represent the -ɨ- of my Tangut reconstruction which has no exact Tibetan equivalent.

The Tibetan root for 'black' is nag and its Burmese cognate is nak. Both have initial n-, and Huang and LaPolla (1996: 375) assume Qiang ɲ- is from earlier *n-. What conditioned the palatalization of *n- in Qiang and rGyalrong (and perhaps also Tangut)?

In the Pearl, the Tangut word for 'black' has the very bizarre Chinese transcription 嘿 implying x-, not ɲ- or n-. Could this imply that the tangraph

TT4479 'black'

had a second reading that was a loanword from Chinese 黑 'black' (with a 口 added to indicate some unknown difference in pronunciation from the Chinese original)? NYA:-T A COUNTEREXAMPLE?

In my last post, I was perplexed by the unexpected appearance of n- before Grade III rhyme 21 -ɨaa. Perhaps this reconstruction is wrong. Arakawa (1999) has a very different reconstruction of R21 and other rhymes in his rhyme group III (partly corresponding to Gong's and my rhyme group IV):

Rhyme Arakawa grade Arakawa Gong grade Gong My grade This site
R21 IV (sic!) -ya: III -jaa III -ɨaa
R22 I -a' I -aa I -aa
R23 II -ya' II -iaa II -ææ
R24 III -a:' III (again!) -jaa IV -iaa

The reason for the unusual order of grades is unknown. Other rhyme groups contain rhymes arranged by grades in ascending order: e.g.,

Rhyme Arakawa grade Arakawa Gong grade Gong My grade This site
R17 I -a I -aa I -a
R18 II -ya II -iaa II
R19 IIIa -a: III -jaa III -ia
R20 IIIb IV -ia

In my reconstruction, n- normally precedes -i- but not -ɨ-. Nonetheless, nɨaa R21 exists but the expected niaa R24 doesn't.

In Arakawa's reconstruction, those syllables might correspond to nya: R21 and na:' R24. Since n- can appear before other -V: without -y- in Arakawa's reconstruction, the absence of na:' R24 is accidental. And since n- can appear before -y- in Arakawa's reconstruction, there is nothing odd about nya: R21 - or is there?

As a rule, there are no -yV: in Arakawa's reconstruction with two exceptions: -ya: R21 and -ya:n R103. (The latter is equivalent to Gong's -joor and my -ɨooʳ/-iooʳ.) Is there any other language which avoids -jV: with the sole exception of -jaa, presumably because aa is the most basic (and frequent?) of long vowels and therefore most likely to be the sole survivor of postglide shortening: -jVV > -jV? NƗAA-T A COUNTEREXAMPLE?

In my first post on The Black Swan, I presented a test of the compensatory lengthening hypothesis: i.e., the derivation of Tangut long vowels from earlier vowel-final consonant sequences. The Tangut title of The Black Swan has two long vowels:

məɨ bã ŋææ nɨaa

They correspond to vowel-final consonant sequences in related languages:

Tangut ŋææ (second syllable and root of 'goose')

Old Chinese 鵝 *ŋaj < ?*ŋal or Old Chinese 雁 *ŋrans

Written Tibetan ŋaŋ 'goose, swan'

Written Burmese ŋanh < *-ns 'goose'.

I would not rather not link the WT form with the OC and WT forms by positing an irregular assimilation *-ŋ-s > *-n-s (with subsequent -n ~ -l confusion in OC for a double dose of ad hoc contrivance)

Tangut nɨaa 'black'

(Old Chinese 黑 *hmək < ?*sʌ-mək 'black' lacks *n, and there is no Chinese-internal evidence for a cluster *mn)

Written Tibetan nag-po 'black'

Written Burmese nak 'black'

Does the above evidence 'prove' that Tangut -VV orignated from *-VC? Perhaps some -VV did, but perhaps not all. Are there any external codaless cognates for Tangut -VV words? For simplicity, I will only look at the Grade III rhyme 21 -ɨaa of 'black' and the similar Grade I rhyme -aa and Grade IV rhyme 22 -iaa in Guillaume Jacques' list of proposed Tangut-gDong-brgyad rGyalrong cognates. (The Grade II rhyme 23 -ææ is not represented.)

Tangut grade Tangut rhyme gDong-brgyad rGyalrong rhyme Number of cognates
I -aa -aʁ < Proto-rGyalrong *-Vq 1
III -ɨaa 2
IV -iaa -at 1
-as 1

On the rather shaky basis of five words, one could conclude that Tangut -(ɨ/i)aa might always come from *-VC. One could even go further on a limb and speculate that acute codas conditioned the Grade IV rhyme -iaa (but what would condition the Grade I/III distinction?).

Can one stop here and assume that all *-VC became long vowels in Tangut? No, for gDong-brgyad rGyalrong -aʁ generally corresponds to Tangut short vowels:

Matches between -aʁ and Tangut short vowels Tangut long vowels
17 3

I suspect some Tangut tense vowels are contractions of earlier tense long vowels. (Tangut has no short tense vowels.) If all Tangut tense vowels in this data set are treated as long, long vowels barely outnumber short vowels:

Matches between -aʁ and Tangut short vowels Tangut long vowels including tense vowels
9 11 = 3 + 8 cases of R63 -ạ < ?*-ạạ


*-VC > Tangut short vowel -V

*s- + -VC > *-ṾṾ > Tangut short tense vowel -Ṿ

(The preinitial *s- may be one if not the only source of tension.)

*-VC + unknown conditioning factor > Tangut long vowel -VV

If so, what was the unknown conditioning factor? Was the factor associated with the source of short vowels rather than long vowels?

I have been assuming that all the proposed cognate pairs are correct. Are they?

Lastly, nɨaa 'black' and its homophones (disregarding tones) are counterexamples to another hypothesis of mine. I proposed that dentals like n- could never occur in Grade III: e.g.,

niu R3 (Grade IV) exists but not nɨu R2 (Grade III)

ni R11 (Grade IV) exists but not nɨi R10 (Grade III)

nia R20 (Grade IV) exists but not nɨa R19 (Grade III)

niɨ R31 (Grade IV) exists but probably not R30 (Grade III)

the initial of LFW0713 R30 1.29 is disputable

nie R37 (Grade IV) exists but not nɨe R36 (Grade III)

so why does nɨaa R21 (Grade III) exist but not niaa R24 (Grade IV)? KRISHNA'S GOOSE

Here's the solution to a Sanskrit contest of mine on another site:

कृष्णहंसेन सह पतिष्यति।

kṛṣṇahaṃsena saha patiṣyati.

'(He) will fly with the black swan' (referring to Taleb's book The Black Swan)

Breaking it down word by word:

kṛṣṇahaṃsena 'black swan'

kṣṇa 'black'; could also refer to the god Krishna; cognate to Chernobyl and Crna Gora ('Black Mountain' = Montenegro)

haṃsena 'goose; swan' (instrumental singular of haṃsa); cognate to English goose which retains the g- lost in Sanskrit; both are descended from a Proto-Indo-European *gHens (which has a surprising Old Chiense lookalike)

saha 'with'; unlike English with, it follows a noun and requires that noun to be in the instrumental case (hence haṃsena saha rather than haṃsa saha)

patiṣyati 'will fly' (third person singular)

pat- 'fly'

-iṣya- (future tense)

-ti (third person singular); cognate to archaic English -th, -t in many European languages

the implied subject is sa 'he' (the feminine form saa 'she' is cognate to English she)

This message indicated that I was going to read (= 'fly with') The Black Swan.

Later that day, I wrote

कृष्णहंसेन सह पतति।

kṛṣṇahaṃsena saha patati.

with a slightly different verb form. Note the time. Can you guess what patati might mean? Select the invisible text below to see the answer.

patati is 'flies' (or 'flieth'). The -a- indicates present tense. I was hinting that I was currently reading The Black Swan - and I still am! I'm on p. 90.

The root pat can also mean 'fall' and is cognate to feather and pterodactyl 'wing-finger'.

There is a homophonous root pat meaning 'reign' that is cognate to power, potent, despot, and even hospital.

One could therefore translate the above lines as 'He will fall with the black swan' or 'he will rule with the black swan', etc. - though I have no intention of doing either! BLACK HEAVEN GOOSE

is the literal translation of

məɨ bã ŋææ nɨaa
the mystery phrase from this post.

Its structure is noun-adjective:

məɨ bã ŋææ (my calque of Chinese 天鵝 'swan' < 'heaven-goose')

məɨ 'heaven'

cognate to Written Burmese muih 'sky'

bã ŋææ 'goose'

both characters have 'bird' on the left

no known cognates of the first syllable

see here for cognates of the second syllable

nɨaa 'black'

cognate to Written Tibetan nag 'black', Written Burmese nak 'black'

I was trying to translate the title of Nassim Nicholas Taleb's The Black Swan into Tangut (which has no 'the'). I wish to thank David Boxenhorn for recommending this book. I'm only up to page 75 so far, and I wish I could write comments about every page! Although the author is not a linguist, so much of what he says has linguistic applications. For instance, on p. 58 he noted that

Popper introduced the mechanism of conjectures and refutations, which works as follows: you formulate a (bold) conjecture and you start looking for the observation that would prove you wrong. This is the alternative to our search for confirmatory instances. If you think the task is easy, you will be disappointed - few humans have a natural ability to do this. I confess that I am not one of them; it does not come naturally.

I confess that I too am not one of them. Nonetheless, I did do the right thing last week before I started reading The Black Swan. I tested the Tangut vowel length hypothesis by looking for counterexamples in Sanskrit transcriptions. (But the interpretation of those transcriptions - if they are indeed transcriptions directly from Sanskrit and not loanwords from Chinese - is not as clear-cut as that post implies.) I could have looked for examples of Tangut and Sanskrit long vowels matching, but I didn't.

Yesterday, I mentioned the compensatory lengthening hypothesis which would derive Tangut long vowels (already in doubt due to the failure of the Sanskrit test) from earlier *-VC sequences that lost their codas. If this hypothesis were false, I would expect at least some external cognates to end in vowels at the earliest reconstructible stage:

If hypothesis is false If hypothesis is true
Tangut -VV : non-Tangut *-V(C) Tangut -VV : non-Tangut *-VC (but never *-V!)

Finding external cognates ending in consonants does not confirm the hypothesis, since such cognates could exist regardless of whether the hypothesis is true or not.

I could claim that pre-Tangut retained final stops lost in the rest of Sino-Tibetan: e.g.,

Proto-Sino-Tibetan *-Vg > Tangut -VV but non-Tangut -V

but this is an ad hoc copout. Of course, all this begs the question of how one determines whether a word in Tibetan, etc. is truly cognate to a Tangut word.

Forthcoming: Do Tangut long vowels correspond to early final vowels outside Tangut?

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2009 Amritavision