are constantly on my mind:

守護 shouhu 'defend' is a synonym compound:

shou 'defend'

hu 'defend'

-zhe '-er; one who ...'

They are also known as

Vogterne in Danish (Vogte may be cognate to part of the English name for them)

Őrzők in Hungarian

Strażnicy in Polish

Vartijat in Finnish

(3.15.1:18: Is vartija 'guard' borrowed from Germanic? Cf. ward.)

and - no surprise -

Custodes in Latin

Who are they? What will I see tomorrow? (I don't want to click on the second link. I want to be surprised.) ARE UVULAR STOPS MORE PRONE TO VOICING?

For many years, I've been puzzled by the voiced pronunciation of ق q in Persian: e.g.,

Arabic qurʔaan > borrowed into Persian > now Persian ɢorʔaan

Earlier Persian q has become [ɣ] in noninitial position: e.g.,

Arabic tˁaariq > borrowed into Persian > now Persian tareɣ

However, nonuvular voiceless stops (e.g., k and t) in Arabic borrowings have not become g and d in modern Persian.

Last night, I noticed a somewhat similar phenomenon for the first time in another language. The Zbu cognate of gDong-brgyad rGyalrong ta-q 'needle' (last seen here) is tɐ-ʁâv with voiced ʁ. Since I cannot find any Zbu words with initial ʁ- corresponding to gDong-brgyad q-, Zbu -ʁ- must be the product of intervocalic voicing (unlike the unconditioned voicing of q in Persian and even in some Arabic dialects*: e.g., Libyan Arabic). One might expect other stops and even affricates to be subject to intervocalic voicing, but -q- can also be found in Zbu: e.g.,

kɐ-qɐ̂χ (also qɐ̂χ, qə́χ) 'remove skin'; cf. gDong-brgyad kɤ-q

(3.13.0:39: Could -q- have been restored by analogy with the prefixless Zbu forms?)

*-p- is the only other stop that sometimes lenites in Zbu: e.g.,

və-v 'low'; cf. gDong-brgyad ɯ-pa

(3:13.0:56: Could Tangut bi 'low' be from *Ni-pa?)

kə-ɣvə̂x 'inflate oneself'; cf. gDong-brgyad kɯ-pɯɣ

A *-p- word without lenition is

kə-pɐʔ 'Chinese' (which doesn't sound like any Chinese word for 'Chinese' - what's its etymology?); cf. gDong-brgyad ku-pa

(3.13.0:45: Guillaume Jacques [2004: 273] does not list any Zbu word pɐʔ 'Chinese' that could motivate the analogical restoration of -p- in kə-pɐʔ.)

What does p have in common with q (besides being a voiceless nonacute stop)?

*3.13.0:55: I think the shift of q to g in Arabic dialects could be part of a chain shift absent in standard Arabic:

q > g > j [dʒ]

Standard Arabic has no g.

No such chain shift existed in Persian, so I can't propose a parallel explanation for the voicing of Persian q.

Nonstandard Arabic g can correspond to standard j: e.g., in Egyptian Arabic. I've long assumed that Egyptian g is conservative, since I have never heard of a language that hardened to g, and I would rather not posit a 'round trip' change:

*g > *dʒ > g

What is the evidence for regarding Egyptian g as an innovation? FILLING THE ΓAP WITH FINE IRON

Recent posts may have given the mistaken impression that Tangut -a rhymes derive from *-aq. But I've only been looking at one part of a bigger picture. I don't have time to examine all other sources of Tangut -a, so I'll just fill a single gap with a word I've brought up before as a third example of lenition:


TT2597 ɣa < *Cʌ-ka R17 1.17 'needle' =

'metal' < top of TT2599 ʃɨõ R58 1.56 'iron' +

all of TT5522 tshiẽ R43 1.42 'fine'

Other non-Chinese Sino-Tibetan words for 'needle' imply pre-Tangut *Cʌ-qap:

gDong-brgyad rGyalrong ta qaβ < *qap

Written Tibetan khab < *kap

(But Zhongu Tibetan has khi, not qhi. Did *q- front to kh- before *i? Can qh- precede i? Sun [2003] lists no instances of qhi in Zhongu.)
Written Burmese ʔap (is WB ʔ- partly from *q-?; cf. similar proposals for Old Chinese)

It is tempting to regard Old Chinese 針 *kim as cognate, but there is no OC-internal evidence for a uvular initial or a nonhigh vowel. The *-m is unusual since the phonetic 十 *gip ends in *-p like the non-Chinese forms. I have wondered if OC oral/nasal coda alternations reflect lost suffixes: e.g., *-m < *-p-N(V).

Like Tangut, Qiang languages (Taoping and Mawo χe, Ronghong χa) have lost the original coda. SPEAKING OF THREAD PEOPLE

Mahaadatṛ asked me for a Tangut cognate of Written Tibetan thag 'weave' and I found

TT4659 la R17 1.17 'weave'

cf. Japhug rGyalrong kɤ-taʁ < *taq and Old Chinese 織 *tək 'id.' as well as WT thag < *tak

Although la seems to have nothing in common with thag except for its vowel, I think its lateral initial is a lenited dental:

*Cʌ-taq > *Cʌ-d- > *Cʌ-l- > la

cf. Proto-Korean *VtV > Middle Korean VrV

Proto-Lolo-Burmese *rak 'weave' could similarly be from an earlier *CV-tak that had lost its presyllable.

The word that inspired my secondary lateral hypothesis was

TT1075 lew R44 1.43 'one'

which could be from *Cʌ-tek or *Cʌ-tik. I compared it to Written Tibetan gcig < *k-tik 'one' and Old Chinese 隻 *tek 'single'.

The graph for la 'weave' makes little sense:


TT4659 la R17 1.17 'weave' =

'speech' < left of TT4639 kiạ R67 1.64 'extol' +

'thread' and 'person' < center and right of TT2374 pəụ R61 2.51 'weft'

The presence of 'thread' (cf. Chn 糸 'id.') is expected, but why does weave include 'person' (because only people can weave?) and, of all things, 'speech' (cf. Chn 讠, cursive of 言 'id.')? Why would 'extol weft' (which is not in Tangut object-verb order) make one think of 'weave'? kiạ 'extol' sounds nothing like la 'weave' and cannot be phonetic (a shared a is insufficient).

'Extol' looks like 'speech' + 'person' + (common mystery right-hand element). Its analysis is


TT4639 kiạ R67 1.64 'extol' =

'speech' < right of TT1252 dziụ R62 1.59 'witticism' +

'person' + ? < right of TT3484 dziw R46 1.45 'combine'

The analysis of

TT2374 pəụ R61 2.51 'weft'

is unknown. It looks like 'not' + 'thread' + 'person'. What does 'not a thread person'.have to do with weft? DOG + HAND = ?

What does the Tangut character


which looks like 'dog' + 'hand' + (unknown common right-hand element) represent? Answer here.

Its Tangraphic Sea analysis is


va = giu 'pig' (in the twelve-animal cycle) + va (phonetic)

The use of the 'dog' element is reminiscent of Chinese 犭 'dog' which appears on the left side of 猪 'pig' (whose right side is phonetic, just like its Tangut counterpart).

The analysis of giu 'pig' is somewhat circular:


giu 'pig' = va + giu 'three' (phonetic)

Oddly, the phonetic symbol va is not the phonetic sum of its parts:


va (phonetic symbol) = riụ (a surname) + bɨu (another surname)

One might think 'hand' on the left of va and riụ is a phonetic element, but lạ 'hand' sounds like neither syllable. PIG PIE

Both short and long Tangut -a correspond to gDong-brgyad rGyalrong -aʁ < *-aq. An example of a short -a correspondence is 'pig':

Tangut va R17 1.17 : gDong-brgyad rGyalrong paʁ < *paq

Compare with


Taoping pa

Ronghong pie

Mawo pi

(Cf. the vocalism of the Qiang words for 'thick': Mawo has the 'brightest' vowel while Taoping retains the original *a. Brightening is completely absent from the Tangut and gDong-brgyad rGyalrong forms. What are other rGyalrongic forms for 'pig?)

Written Tibetan phag < *pak

Written Burmese wak

Tangut v- and WB w- may not have a common origin. I suspect that some Tangut v- arose from intervocalically lenited *-p-:

*Cʌ-p- > *Cʌ-b- > *Cʌ-β- > *β- > v-

Some Vietnamese v- also originated from a similar sequence of changes: e.g.,

'mend': < *Cə-paʔ (borrowed from Chinese 補; the presyllable could be a Vietnamese native prefix or borrowed from Chinese)

Matisoff (2003: 662) reconstructed Proto-Tibeto-Burman *pʷak ~ *wak 'pig'. Perhaps *pʷ- simplified to *p- in Proto-Qiangic and pre-Tibetan. The Qiangic language Tangut independently recreated a *w-like initial v- (see above) whereas Burmese inherited an old *w-. But I've wondered if the earliest root initial was simply *p- which (irregularly?) lenited to *-w- in medial position. LAA LẠ LẠ: THICK, THICK HANDS

In Friday's post, I originally misglossed Arakawa la:' R22 1.22 as 'hand' instead of 'thick'. I confused it with Arakawa laq R66 1.63 'hand'. The words for 'thick' and 'hand' are nearly homophonous in both Tangut (regardless of reconstruction) and gDong-brgyad rGyalrong:

Gloss Tangraph Arakawa 1999 Gong 1997 and this site gDong-brgyad rGyalrong
thick la:' laa kɯ jaʁ < *ljaq
laq lạ
hand tɯ jaʁ < *ljaq

-q in Arakawa's reconstruction indicates tenseness of a preceding vowel and is not a uvular coda.

lạ 'thick' may be from *llaaq with a tense onset from an earlier prefix-root initial sequence *C-l-. I reconstruct an long vowel root *laaq on the basis of the long vowel of lax-vowelled laa 'thick'.

lạ 'hand' could be from *llaq or *llaaq.

There is no Tangut-internal evidence for a *-j- in 'thick' or 'hand'. Did pre-Tangut *lj- or *ʎ- merge with *l-?

I wonder if Qiang words for 'hand' once had *lj-:

Taoping i

Ronghong ji

(Mawo has dʒəpɑ 'hand', but I wonder if lɑ- in lɑxɑpi 'shoulder' was once 'arm'; cf. Ronghong jipi 'shoulder'. The i of Taoping and Ronghong could be from *-a < ?*-aq.)

Qiang words for 'thick' have l-:

Taoping la

Ronghong lie

Mawo li

Matisoff (2003) does not reconstruct an *l-word for 'thick' in Proto-Tibeto-Burman. Starostin's online Proto-Sino-Tibetan database also lacks an *l-word for 'thick'. Could ?*laq 'thick' be a Qiangic innovation unlike its near-homophone ?*ljaq 'hand' which was inherited from Proto-Sino-Tibetan?

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