... is an initial *ɦ-, if I am right.

Back around 1993, when I first discovered that Vietnamese was full of Chinese loans, I had fun teaching myself the correspondences between Sino-Vietnamese and the readings I already knew (Mandarin, Cantonese, Sino-Japanese, Sino-Korean) plus Karlgren and Toudou's Middle Chinese reconstructions. I was not prepared for an h- in the SV readings for 'friend', 'have', and 'right' which all share the same phonetic (又~ナ):

Gloss friend have right
Sino-Vietnamese hữu [hɨw]
Mandarin you
Cantonese jɐw
Sino-Japanese yuu < iu (new layer), u (old layer)
Sino-Korean u yu u
Karlgren *ji̭ə̭u:
My generic Early Middle Chinese *wuʔ
My generic Late Middle Chinese *jwɨw

(Is SJ u for 友 attested anywhere, or is it just theoretical?)

(SK 有 yu is more like later SJ iu and is probably from a later layer of borrowing than SK 友右 u.)

(Karlgren's colon represents the rising tone. I have left tones out of my LMC as well as Mandarin and Cantonese.)

(I left my dictionary of Toudou's reconstructions back in Hawaii.)

MC *j- and *w- usually correspond to SV d- (< *j-) and v-, not h-.

Prothesis of h- is otherwise unknown in Vietnamese historical phonology. Hence I assume that SV h- in those three morphemes must reflect a consonant in their Chinese source dialect. In the last two posts, I proposed that Cantonese j- partly comes from *hj-. Similarly, could Cantonese j- and SV h- come from a southern LMC *ɦ-?

Stage 1: SLMC *ɦɨw

(> borrowed into Vietnamese as hưu [hɨw] with appropriate tones)

Stage 2: Pre-Cantonese *ɦɨəw

(breaking: > *ɨə)

Stage 3: Pre-Cantonese *ɦiəw

(fronting: > *i)

Stage 4: Pre-Cantonese *ɦjəw

(diphthong to glide-vowel sequence: *-iə- > *-jə-)

Stage 5: Cantonese jɐw

(loss of *ɦ- before j-; lowering of *ə)

It's possible that *ɦ- devoiced to *h- before being lost before *j-:

Sinograph Gloss Before devoicing Devoicing h-loss and other changes
friend *ɦjəw *hjəw jɐw
rest *hjəw

1.18.7:45: *ɦ- normally became h- in Cantonese except before front vowels and *w-

Sinograph Gloss Southern Late Middle Chinese Cantonese Sino-Vietnamese
what *ɦɑ hà [ha]
harmony *ɦwɑ hòa [hwa]
tie *ɦej hɐj ̣ [hej]
grace *ɦwej wɐj huệ [hwej]
marquis *ɦəw hɐw hầu [həw]
now *ɦien jin hiện [hiən]
dark *ɦwien jyn huyền [hwiən]

(Not all Sino-Vietnamese forms happen to be this conservative.)

I'll discuss the exceptions next time. NINE HILL ENEMIES REST

Back in 1990, I became addicted to the comics of 邱福龍 Khoo Fuk Lung. I was surprised to learn that his surname was pronounced jɐw in Cantonese. In other southern Chinese languages, 邱 has an initial kh-: e.g., Chaozhou khu (source; its k = IPA [kʰ]). (Is Khoo of Chaozhou ancestry?) Until last night, I couldn't figure out why 邱 had an initial j- in Cantonese. Here's a scenario incorporating my ideas from "The Perfect Rest Stop?":

Gloss nine hill enemy rest
Southern Late Middle Chinese *kɨw *khɨw *gɨw *hɨw
Compare with Sino-Vietnamese (ignore tones) cửu [kɨw] khưu [xɨw] < *kh- cừu [kɨw] < *g- hưu [hɨw]
Pre-Cantonese stage 1: > *ɨə *kɨəw *khɨəw *gɨəw *hɨəw
Pre-Cantonese stage 2: > *i *kiəw *khiəw *giəw *hiəw
Pre-Cantonese stage 3: *-iə- > *-jə- *kjəw *khjəw *gjəw *hjəw
Pre-Cantonese stage 4: lenition of *kh- > *h- *hjəw
Cantonese (ordering of most changes uncertain):
> *çj- > j- followed by *Cj- > C-
> kh- in lower level tone syllables
> ɐ
kɐw jɐw khɐw jɐw


丘 'hill' is homophonous with the surname 邱 which may simply be 'Hill'.

仇 'enemy' is now read khɐw only as a surname; it is otherwise read sɐu or tshɐw (from the synonym 讎 SLMC *[d]ʑɨw 'enemy').

Only original *kh- lenited to *h-. The secondary kh- of khɐw from *g- must postdate that lenition.


Other examples of *(k)hj- > j- from A Chinese Talking Syllabary of the Cantonese Dialect:

Upper level tone:

SLMC *khɨw > *khjəw > jɐw: 蚯坵 (homophonous with 丘邱)

SLMC *hɨw > *hjəw > jɐw: 咻髹庥茠貅 (homophonous with 休)

Upper rising tone:

SLMC *khɨw > *khjəw > jɐw:

SLMC *hɨw > *hjəw > jɐw:

The absence of *(k)hj- > *j- in upper departing tone syllables is due to a chance gap.

*(k)hj- > j- is not in any lower tone syllables since *kh- and *h- could only be in SLMC upper tone syllables.

*h- was not lost if it was not followed by *-j-: e.g.,

SLMC *hiew > hiw: 嘵 'sound of fear or argument' (not the perfective marker written with the same graph but probably derived from 休 SLMC *hɨw; see "The Perfect Rest Stop?") THE PERFECT REST STOP?

While Googling for Matisoff's (2004) article on 'brightening' in Tangut, I found a reference to it in Kwok Bit-chee and Kataoka Shin's (2006) "The Origin and Development of the Perfective Aspect Marker Hiu in Early Cantonese". I agree with their derivation of Ct 嘵 hiw 'perfective marker' from 休 'rest; stop'. This etymology seems implausible from a modern Cantonese perspective because 嘵 hiw only shares a tone and a final -u with 休 jɐw. But I wonder if 嘵 hiu is an archaism:

1. In southern Late Middle Chinese, 休 was *hɨw. This was borrowed into Vietnamese and remains intact today as Sino-Vietnamese hưu [hɨw].

2. The southern LMC rhyme *-ɨw broke to *-ɨəw, as reflected in some later Sino-Vietnamese borrowings: e.g.,

'a surname': 仇 SLMC *gɨw > SV cừu [kɨw]

'seek': 求 SLMC *gɨw > *gɨəw > newer layer (?) SV cầu [kəw]

仇 and 求 were homophonous in SLMC and are still homophonous in modern Cantonese: both are khɐw.

Since as far as I know, Vietnamese has no syllable cườu, SV cừu [kɨw] and SV cầu [kəw] could also both be attempts to imitate SLMC *gɨəw.

However, the perfective aspect marker derived from 休 did not change from *hɨw to *hɨəw.

3. The central vowel fronted to *i, and the diphthong *iə was reinterpreted as a glide-vowel sequence *jə:

*hɨəw > *hiəw > *hjəw

The aspect marker *hɨw became *hiw. Its -i- was the main vowel and was not reinterpreted as a glide.

4. *h- assimilated to j- by becoming palatal *ç- before being ultimately lost, while at some point lowered to ɐ. The ordering of these two changes is uncertain:

*hjəw > *çjəw > modern Cantonese jɐw

The aspect marker *hiw did not lose its h- because it never developed a *-j-glide.

This scenario entails early (post-LMC) grammaticalization for 嘵 hiw predating the breaking of in step 2. However, 嘵 hiw is first attested in 1840, long after the borrowing of the last layer of Sino-Vietnamese circa the 10th century. Was 嘵 hiw unwritten for centuries, or is the above scenario simply wrong?

1.16.00:39: The choice of the graph 嘵 for the aspect marker must date from a time when that word was homophonous with 驍 hiw and other graphs containing the phonetic 堯. 驍 etc. were SLMC *hiew (cf. Sino-Vietnamese hiêu [hiəw]) which was distinct from SLMC 休 *hɨw (cf. hưu [hɨw]). If the above scenario is correct, the graph 嘵 was chosen after SLMC *ie monophthongized to i (phonetically [ii]).

嘵 originally represented an unrelated morpheme 'sound of fear or argument'. There is a phrase 嘵嘵不休 'argue nonstop' with both 嘵 and 休 plus 不 'not'. WHY BRIGHT MICE?

So far, I have been saying that

TT3016 xwɨi R10 1.10 and TT3018 ?xwiaa R24 1.23

are cognate words for 'mouse; rat' without any further elaboration. They differ in their degree of 'brightness'. xwɨi is fully 'brightened' with a high main vowel (i) whereas xwiaa is only partly 'brightened' with a long low main vowel (aa). Other sets of words differing in brightness are

'not' (cf. Old Chinese 無 *ma 'not have/exist')

TT1491 mi R11 1.11

TT3510 mie R37 1.36

'bitter' (cf. Old Chinese 苦 *khaʔ 'id.')

TT0608 and TT0705 khɪ R9 1.9

TT1791 kha R17 2.14 'bitter lettuce'

'patch' (cf. Old Chinese 補 *paʔ 'id.')

TT3997 piạ R67 1.64

TT2460 vɨẹ R64 2.54; v- is from *sV-p-

TT5627 bie R37 1.36; b- is from *N-p-

What conditioned the different degrees of brightness? I suspect that presyllables were involved:

Solution A

Presyllables conditioned brightening: e.g.,

*Cɯ-Ca > Ci whereas *Ca > Ca

Solution B

Presyllables blocked brightening: e.g.,

*Cʌ-Ca > Ca whereas *Ca > Ci

The fact that two function words underwent brightening, mi 'not' and ti 'do not' (cf. t(h)ɛ 'id.' in some Qiangic languages) implies that solution B might be correct, since I think those words were probably simple monosyllables in pre-Tangut:

*ma > mi

*ta > ti

Matisoff (2004) found that the most frequent Tangut reflex of earlier (his Proto-Tibeto-Burman) *-a is -i. This also supports solution B if one assumes that simple *CV syllables outnumbered sequisyllables and *C-CV syllables in pre-Tangut.

Non-i reflexes of *-a may reflect different types of earlier presyllables or prefixes: e.g.,

*type 1 presyllable/prefix-Ca > Cie (bright -i-, partly brightened main vowel)

*type 2 presyllable/prefix-Ca > Cia (bright -i-, nonbright main vowel)

*type 3 presyllable/prefix-Ca > Ca (wholly nonbright)

Thus I could reconstruct the pre-Tangut words for 'mouse; rat' as

*xwa > xwɨi R10 1.10 (bare root)

*type 2 presyllable/prefix-xwa-C > xwiaa R24 1.23

The bare root word is more common than the derivative which is absent from Tangraphic Sea, Precious Rhymes of the Tangraphic Sea, and the older edition of Homophones.

The root xwa resembles Matisoff's (2003: 663) Proto-Tibeto-Burman *wak 'rat', though the initials don't match. Tangut xw- cannot be derived from *s-w- which would have become Tangut v- followed by a tense vowel. DETERMINING THE DRACONIC GRAPHOMETER (the what?*)

The tangraph corresponding to the Chinese initial 非 in Cut Rhymes is

TT0179 -əi R8 1.8 'dragon'

I have deliberately omitted the initial of its reconstruction. 非 has initial f- in modern northwestern Chinese dialects and was *f- even in the pre-Tangut period, when it was transcribed in Tibetan as ph- and Khotanese Brahmi as hv- or hv-. (Neither the Tibetans nor Khotanese had letters for f in their alphabets.) Thus I guessed that 'dragon' might have had initial f- in Tangut. However, neither the Tibetan nor Chinese transcriptions of 'dragon' point to f-:

Tibetan: wyi (instead of phyi)

Tangut period NW Chinese: 嵬 ?*w- (instead of 非 *f-)

(Why the question mark?**)

In Leilin, TT2473, a homophone of TT0179, was used to transcribe the sinographs 威 ?*(ʔ)w- and 褘 ?*xw-.

At first glance, the evidence seems to support reconstructing w- like Sofronov and Gong. However, there are at least three problems with w-:

1. w- is obviously not labiodental. (1.14.0:14: This is the weakest objection. If the Tangut period NW Chinese pronunciation of the Chinese labiodental initial category 微 were bilabial *w-, then the Tangut might regard their own w- as 'labiodental'.)

2. Why would the Tangut choose a voiced glide w- instead of a voiceless fricative-glide sequence xw- as a native equivalent of the Chinese voiceless fricative 非 *f-?

3. If the initial were truly w-, it would be consistently transcribed in Tibetan as w-. However, another tangraphs with the same initial fanqie speller was transcribed with yww- and b-, implying a consonant with more friction: e.g., v-, which would be more similar to 非 *f- than w-.

For now, I will reconstruct the initial of 'dragon' as v-.

*Draconic refers to TT0179 'dragon' and graphometer is a calque of Chinese 字母 'character-mother' = 'character used as a label for an initial consonant'. Meter is from Greek μήτηρ, not μέτρον.

**1.14.0:48: The question mark indicates uncertainty. It's not certain whether TPNWC *w- was phonetically *[w] or *[v]. Modern NW dialects have either w- or v-.

I prefer to use the symbol *w- because if TPNWC had *v- instead of *w-, I would expect Tangut v- to directly correspond to TPNWC *v-:

Chinese initial name

Tangut period NW Chinese



Tangut (expected)



But Tangut v- corresponds to 非 *f- as well as 微. This may imply that TPNWC had no *v-, and that Tangut v- was the closest available approximation of TPNWC *f- and *w-:

Chinese initial name

Tangut period NW Chinese



Tangut (actual?)




The Tangut book Cut Rhymes takes its name from a famous Chinese rhyme dictionary but is actually a set of phonetic tables containing syllables arranged by initial and final. Unfortunately, only fragments of this text have survived. I have yet to actually see these tables apart from bits reproduced by Nevsky and Nishida.

Nevsky (1960 I: 135) listed the 36 initials of Tangut according to Cut Rhymes. The four labiodental initials of Cut Rhymes are all listed as labiodental in Homophones (in spite of the bilabial w- and velar x- reconstructed by Sofronov and Gong; see below!) and correspond to Chinese initials:

TT0179 vəi R8 1.8 'dragon': Chn 非

TT3018 xwiaa R24 1.23 'mouse; rat' : Chn 敷

(xw- carried over from Gong's reconstruction)

TT3016 xwɨi R10 1.10 'mouse; rat' : Chn 奉

(xw- carried over from Gong's reconstruction)

TT2409 vɨi R10 1.10 'send; dispatch; benefit from' : Chn 微

The values of the Chinese initials have changed over time:

Chinese initial name
Middle Chinese initial *p- *ph- *b- *m-
Pre-Tangut NW Chinese ?*pf- ?*pfh- ?*pfɦ- ?*ɱbv-
Tangut period NW Chinese *?*f- *?*v- or ?*w-

The Tangut author of Cut Rhymes was unaware of Middle Chinese or PTNWC, so I assume he only knew of two TPNWC labiodental initials, ?*f- and ?*v-. (I doubt that the unusual transitional initials ?*pfh- and ?*pfɦ- could have survived intact into the Tangut period.) Therefore I would not accept Nishida's (1964: 25) first reconstruction of a pf- : pfh- distinction in Tangut corresponding to the 非 : 敷奉 distinction in the Chinese initial names. In the Tangut period, all three had the same Chinese initial, so perhaps their corresponding Tangut initial names also had the same Tangut initial (f- or wh-?) in the Cut Rhymes dialect:

Chinese initial name
Tangut period NW Chinese *?*f- *?*v- or ?*w-
Tangut initial name
Gloss of Tangut initial name dragon mouse 1 mouse 2 send
Nishida 1964's first reconstruction pf- pfh- ɱ-
Nishida 1964's revised reconstruction v- f- w-
Sofronov 1968 (ignoring medial -i̭-) w- xw- v-
Gong 1997 (ignoring medial -j-) w- xw- w-
My latest reconstruction up to now v- xw- v-
My revised reconstruction A ?f- v-
My revised reconstruction B ?wh- [ʍ] w-

Next: Did 'dragon' and 'mouse' really share the same initial? THE EXPENSE OF LADY FLYING GRACE

Guillaume Jacques pointed out that

TT3016 ?xwɨi R10 1.10 'mouse; rat'

the initial fanqie speller for the cognate word

TT3018 ?xwiaa R24 1.23 'mouse; rat'

transcribes the following sinographs in the Tangut translation of Leilin:

*f- 'expense'

*f- 'imperial concubine'

*f- 'to fly'

*xw- 'grace'

(The Tangut period NW Chinese initials are fairly certain, but the finals aren't. The latter are irrelevant for this discussion, so I have refrained from speculation.)

If TT3016 had initial f- in the Tangut dialect underlying the Leilin translation, it would not be used to transcribe TPNWC *xw-. But if Leilin Tangut lacked f-, then xw- would be suitable to transcribe TPNWC *f- as well as *xw-.

In the Pearl, TT3016 was transcribed as TPWNC 携 *xw- instead of a TPNWC *f- sinograph.

Therefore I would reconstruct xw- as the initial of TT3016 in Leilin and Pearl Tangut. But did TT3016 had the same initial in all other varieties of Tangut? In Homophones, some x(w)- tangraphs such as TT3016 are in the labiodental chapter rather than the glottal chapter with other x(w)- ?[h(w)] tangraphs. This implies that Homophones Tangut had a labiodental initial (?f-) corresponding to x(w)- in non-Homophones Tangut.

Tangraphs with initial fanqie spellers belonging to the chain containing the initial fanqie speller for TT3016 Homophones chapter Homophones Tangut Non-Homophones Tangut
TT3016 'mouse; rat' II (labiodentals) ?fwɨi R10 1.10 xwɨi R10 1.10
TT3018 'mouse; rat' ?fiaa R24 1.23 xwiaa R24 1.23
TT1022 'blow; puff' ?fɨu R2 1.2 xɨu R2 1.2
TT5299 'Buddha'; loan from TPNWC 佛 *f- ?fɨə R30 1.29 xɨə R30 1.29
TT5300 'treasure'
TT5301 (meaning unknown)
TT3371 'clear; distinct' VIII (glottals) xəu R1 1.1
TT5559 'fragrant'; loan from TPNWC 香 *x- xɨõ R58 1.56

Note that five of the six of the possible f-syllables so far belong to Grade III, just as Late Middle Chinese *f-syllables belong to Chinese Grade III.

(1.12.0:48: The sole exception is Grade IV TT3018 xwiaa R24 1.23.)

(1.12.0:09: But note that xɨõ R58 1.56 is also Grade III but cannot have initial *f- since it is based on TPNWC *x-. Perhaps all Homophones f(w)ɨ- correspond to non-Homophones Grade III x(w)ɨ-, but not vice versa.)

Reconstructing x(w)- for the readings of all of the above tangraphs begs the question: why would some of them be placed in chapter II instead of chapter VIII of Homophones?

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