In "Rolling My Eyes", I listed Min forms for 目 MC *muk 'eye' ending in -a(a)k. Here are all Min forms from Hanyu fangyin zihui with -ak corresponding to rhyme dictionary Middle Chinese *-uk and even *-ok. Forms in parentheses are later borrowings into Chaozhou that do not fit the pattern.

Sinograph MC OC Colloquial Amoy Chaozhou
*muk *muk or *mikw bak bak (but Lin and Chen 1996 have mak)
*puk *puk pak (hok)
*khuk *khuk khak khek (with -e-!)
*luk *ruk lak lak
*tɕhuok *thok tak (tshok)
*mok *mok bak bak

There are also forms ending in -aŋ corrresponding to rhyme dictionary MC *-uŋ such as 'dream', also discussed in "Rolling My Eyes".

Sinograph MC OC Colloquial Amoy Chaozhou
*muŋ *məŋs baŋ maŋ
*buŋ *bəŋ paŋ paŋ
*phuŋ *phuŋ phaŋ (hoŋ)
*ɖuŋ *lruŋ thaŋ thaŋ
*douŋ *duŋ thaŋ tĩã (!)
*nouŋ *nuŋ laŋ laŋ
*buoŋ(h) *boŋ(s) paŋ (hoŋ), phaŋ (aspiration does not match Amoy)
*toŋh *toŋs taŋ (toŋ)
*doŋʔ *doŋʔ taŋ (toŋ)
*thoŋ *hloŋ thaŋ (thoŋ)
*doŋ *loŋ taŋ taŋ
*thoŋh *hloŋs thiã (!; the -i- is inexplicable, unless it's a trace of an OJ *-j- that otherwise vanished in emphatic syllables) (thoŋ)
*soŋh *soŋs saŋ saŋ
*khoŋʔ *khoŋʔ kaŋ (sic; should this be aspirated?) khaŋ
*ɣoŋ *goŋ
*ʔoŋ *ʔoŋ (oŋ)

I think the Min forms reflect vowel breaking in the ancestor of the earliest stratum of Min:

OC *-uŋ/k > *-əwŋ/k > -aŋ/k

OC *-oŋ/k > *-awŋ/k > -aŋ/k

6.15.00:39: Added 疼.

Although OC *-əŋ seems to have become Min -aŋ, as implied by 夢 and 馮., the Amoy colloquial reflex of OC emphatic *-əŋ is -ĩ (with a front high vowel that I wouldn't expect in a formerly emphatic syllable) except for 等 which is tan (sic!) instead of the expected tĩ. ROLLING MY EYES

John Bentley drew my attention to the spelling 巻目 for the place name 巻向 Makimuku in Man'youshuu 7.1087. This spelling is unusual because 目 'eye' is read as moku in the Go-on stratum of Sino-Japanese.

Since there is no Sino-Japanese reading muku (muku is a native reading for 向), 目 moku could have been an inexact match for Old Japanese -muku.

But I suspect that 目 moku actually represented an archaic pronunciation of -muku as *[moku]. Old Japanese u can come from pre-OJ *o as well as *u. So it's possible that Makimuku was once *makimoku. It's even possible that the OJ verb stem 向 muk- 'to face' was pre-OJ *mok- if the normal spelling 巻向 predates the raising of pre-OJ *o to OJ u. (Is there any Ryukyuan evidence for a mid vowel in 'to face'?)

However, there are several complications.

First, the borrowing of the Go-on reading moku for 目 must postdate raising in Japanese, because a *moku borrowed before raising should have become muku. Hence 巻目 cannot be a very old spelling of Makimuku, unless it is the sole surviving hint of a long-extinct reading muku < *moku.

Native words Sino-Japanese The name Makimuku Spelling of Makimuku
Preraising Pre-OJ *o (if 目 Go-on moku were borrowed during this period, it would become muku after raising) Pre-OJ *makimoku 巻向? (if 'to face' was pre-OJ *mok-)
Postraising OJ u 目 Go-on moku borrowed; later, the Kan-on reading *mboku (> modern boku) was borrowed OJ makimuku, with archaic variant makimoku? 巻向, with 巻目 representing the archaic variant makimoku?

Second, a mid vowel for 目 'eye' is unexpected. Ever since Karlgren, 'eye' has been reconstructed with high vowels in Middle Chinese: e.g., Karlgren's *mjuk and my *muk. The word may go back to an Old Chinese *muk or *mikw. (Probable external cognates like Written Tibetan mig, Written Burmese myak, and Tangut me R34 1.33 point toward a palatal vowel or medial.) Yet 'eye' has a mid vowel in Sino-Korean as well as both major strata of Sino-Japanese:

'eye' Sino-Japanese Sino-Korean Sino-Vietnamese
Early Middle Chinese *muk Go-on moku (not muku) (the earlier SK and SV readings are either identical to the later ones, or have been lost)
Late Middle Chinese *muk (*mbuk in the northwest) Kan-on boku < *mboku (not buku) mok (not 묵 muk) ̣mục (as expected)

Third, an initial m- for 'eye' is unexpected because *mu generally became *vu in Late Middle Chinese. Of course, if 'eye' had an unexpected *o, then it would not have been subject to the *m > *v shift.

I briefly considered the idea that SJ and SK reflected an MC *mok from an OC variant with a low-vowel prefix conditioning emphasis and vowel lowering:

Normal: OC *(Cɯ-)muk > southern LMC *muk > borrowed into Viet as mục

Variant: OC *Cʌ-muk > *muk > *mouk > EMC and eastern LMC *mok (> Go-on moku, SK mok), northwestern LMC *mbok (> Kan-on boku)

However, all OC/MC *muk-type syllables correspond to SJ and SK readings with mid vowels:

MC *muk sinographs Sino-Japanese Sino-Korean Sino-Vietnamese
牧睦穆 as well as 目 'eye' and 苜 with 目 as phonetic Go-on moku (not muku) (the earlier SK and SV readings are either identical to the later ones, or have been lost)
Kan-on boku < *mboku (not buku) mok (not 묵 muk) ̣mục (as expected)

It is doubtful that all of these words had low vowel prefixes in the OC dialects ancestral to the sources of SJ and SK.

Unexpected nonhigh vowels (in bold) appear in Sinoxenic borrowings for various MC *Pu(ŋ/k) syllables. (I don't have Numoto's articles on hand, so the SJ column may contain errors.)

Example sinographs EMC (ignoring tones) Go-on LMC (ignoring tones; *mb- in the NW dialect only) Kan-on Sino-Korean Sino-Vietnamese
紑否缶富 *pu fu < *pu *fu fu < *pu pu phu, phâu
風楓諷 *puŋ fu(u) < *pu(u) *fuŋ hou < *pou phung phung, phong
*puk fuku < *puku *fuk fuku < *puku p(h)ok phúc, phước, bức
紑覆副 *phu fu < *pu *fhu fu < *pu pu phu, phâu, pho
豐賵 *phuŋ fu < *pu *fhuŋ hou < *pou phung phong, phung
*phuk fuku < *puku *fhuk fuku < *puku pok phúc
浮負婦 *bu bu < *mbu *fɦu fu < *pu pu phu, phâu, phưu
馮鳳 *buŋ bu < *mbu (not buu!) *fɦuŋ hou < *pou pong phung, phương
伏服復 *buk buku < *mbuku *fɦuk fuku < *puku pok phục
牟矛謀 *mu mu *m(b)u (not *vu!) bou < *mbou mu, mo mâu, mưu
*muŋ mu *m(b)uŋ (not *vuŋ!) bou < *mbou mong mông
目苜牧睦穆 *muk moku *m(b)uk (not *vuk!) boku < *mboku mok mục

Sino-Vietnamese also has nonrounded vowels before *-n and *-t:

Example sinographs EMC Go-on LMC Kan-on Sino-Korean Sino-Vietnamese
分芬焚文 *Pun -un, -on < *-ən *Fun -un -un -ân; also -ăn after v-
不拂佛物 *Put -utsu < *-ut, -otsu < *-ət *Fut -utsu < *-ut -ur -ât

After nonlabial initials, SV has -uân/t.

LMC *fh- is highly improbable and can be rewritten as *f-.

All the LMC labiodental fricatives could be rewritten as affricates: *pf-, *pfh-, *pfɦ-, *ɱv-.

SV ă = [a], SV â = [ə], -o = [ɔ], -ong = [awŋ͡m], -oc = [awk͡p], -ông = [əwŋ͡m] ư =[ɨ], ươ = [ɨə].

The mismatches between the Sinoxenic varieties indicate that they could not have derived from a single standard MC dialect such as the (artificial?) generic dialects recorded in dictionaries and rhyme tables. I hypothesize that the mismatches reflect a process of labial dissimilation that affected different syllables in different MC dialects: e.g.,

*mu > *mu in the source of Go-on and one (earlier?) source of SK, *mbow or *mbəw in the source of Kan-on, *mow in one (later?) source of SK, *məw and *mɨw in the source(s) of SV

*muŋ > *mouŋ in the sources of Kan-on, SK, and SV; Go-on mu may reflect *muŋ without dissimilation

*muk > *mouk in the sources of SJ and SK but not SV

Perhaps the conditioning environment for labial dissimilation was a three-part sequence of labials: *PuW. What I have been writing as *-u, *-uŋ, *-uk could be interpreted as /uw/ /uŋʷ/ /ukʷ/ or /ɨw/ /ɨŋʷ/ /ɨkʷ/, and the final labiovelar could have been pronounced as velar-labial sequences as in modern Vietnamese: [uŋ͡m] and [uk͡p].

*Pu followed by a dental was exempt from labial dissimilation. The shift of *-un/t to *-(w)ən/t in the source of SV (but not SJ and SK) may initially look like labial dissimilation, but it occurred after all initials and is simply vowel breaking.

Go-on -on/tsu < *-ən/t instead of -un/tsu after labials is not due to labial dissimilation but due to the lack of labial assimilation in its source dialect: e.g.,

文 OC *mən > Go-on source EMC ?*mən but dictionary EMC *mun

物 OC *mət > Go-on source EMC ?*mət but dictionary EMC *mut

Vowel dissimilation after *m prevented it from being subject to labiodentalization. *mu would have become *vu, whereas an *m- in a *mɨw(ŋk) or *məw(ŋ/k) would have remained intact.

Some instances of dissimilation must postdate labiodentalization: e.g., the southern LMC *fəw implied by SV phâu must be from an earlier *fu < *pu. It would be too complicated to claim that labials sporadically labiodentalized after non-*u vowels.

SJ and SK -u in some cases could have been the closest equivalents to EMC and LMC *-ɨw, a rhyme that did not exist in Japanese or Korean. Go-on mu for 謀 could reflect a reading 謀 like the *mɨw implied by SV mưu.

SK -u may also reflect LMC *-əw. Hence SK pu for 芣 could reflect a reading like the *fɦəw implied by SV phầu.

Lest one think that these non-u vowels in Sinoxenic merely reflect imperfect borrowing, Chinese languages also have unexpected non-u vowels. Min dialects even have unrounded vowels for 目 'eye' and 夢 'dream' (Amoy from Hanyu fangyin zihui; other data from Lin Lunlun and Chen Xiaofeng 1996):

目 'eye': MC *muk < OC *muk or *mikw

Colloquial Amoy bak (the expected bɔk is literary; buk is not possible in Amoy)

Swatow, Chaozhou, Jieyang mak

Haifeng, Haikang, Dianbai mak and mok

Sanxiang maak

these resemble Written Burmese myak; the resemblance to Proto-Japonic *ma- and Viet mắt must be fortuitious

夢 'dream': MC *muŋh < OC *məŋs

Colloquial Amoy baŋ (the expected bɔŋ is literary; buŋ is not possible in Amoy)

Swatow, Chaozhou, Jieyang, Haifeng, Haikang maŋ

Dianbai maŋ and moŋ

Sanxiang maaŋ

It's not clear whether the Min a(a)-forms for 'dream' directly come from OC *məŋs without any intermediate rounded vowel stage.

Even standard Mandarin has nonhigh reflexes of *-u(ŋ/k) after labials due to (mostly) recent wave(s) of dissimilation: e.g.,

EMC *pu: 否缶 Md fou as well as 富 Md fu

EMC *Puŋ > Md Feng

but EMC *Puk > Md Fu

EMC *mu: 牟謀 Md mou and even 矛 Md mao (other non-Mandarin languages have ma[a]w for 矛; could they reflect an MC *maw < OC *Cʌ-mu or *mæw < OC *rʌ-mu absent from the dictionary tradition?)

EMC *Pun > Md Fen

this superficially is reminiscent of SV -ân/-ăn for this rhyme, but unlike SV, Mandarin has rounded vowels after nonlabial initials: e.g.,

EMC *kun > Md jun (but SV quân)

An apparently isolated case of a nonlow Mandarin reflex of *u before a nonlabial coda is

EMC *but: 佛坲 Md fo as well as 怫 Md fu

Although so much effort has been poured into reco)nstructing the MC dialects used in dictionaries and rhyme tables, variation among MC dialects is still largely unclear. Much more work is needed to confirm my northern *m(b)ouk (> SJ, SK) / southern *muk (> SV) 'eye' idea.

In any case, Go-on moku for 'eye' probably postdates pre-OJ vowel raising. (If moku were an isolated case, it could be regarded as a potential archaism which had escaped vowel raising, but Go-on has no muku at all, implying that a regular phonological rule was involved: i.e., *muk > *mouk in the source dialect.)

I can only think of one way to claim that Go-on moku for 'eye' predates pre-OJ vowel raising:

1. OC *muk or *mikw became *mɨwk in southern Chinese, either through breaking or gravity assimilation (*i became grave like its neighboring consonants).

The Min mak-type words for 'eye' could be from this southern *mɨwk via a stage like *mək.

2. Southern EMC *mɨwk was borrowed by Paekche speakers who taught the reading to pre-OJ speakers. Note that nobody consciously thought of *mɨwk as 'Go-on'; early Sino-Japanese was simply Chinese filtered through two accents (Paekche and pre-OJ).

*mɨwk would have been close to native muku, so if 巻目 is an old spelling, it stood for *makimuku, not *makimoku.

3. If pre-OJ had a vowel that lowered to OJ (as proposed by Frellesvig and Whitman), then pre-OJ *mɨwk became OJ *məwk, which was ultimately Japanized as moku.

The F&W lowering scenario bothers me because it goes against the general raising trend in pre-OJ:

Pre-OJ *əi > OJ ɨy

Pre-OJ *ai > OJ əy

Pre-OJ *e > OJ i

Pre-OJ *o > OJ u

The F&W lowering scenario also has interesting implications for Go-on and its source dialect: e.g., some Go-on o < OJ ə could be from pre-OJ *ɨ:

文 OC *mən > Go-on source EMC ?*mɨn > modern Go-on mon

物 OC *mət > Go-on source EMC ?*mɨt > modern Go-on motsu VOICED OBSTRUENTS IN TANGUT: A NEW SO-WU-TION?

Tangut had at least three types of obstruents apart from nasals. The first two were clearly voiceless unaspirated and voiceless aspirated:

Homophones chapter voiceless unaspirated voiceless aspirated
I p- ph-
III t- th-
IV c- ch-
V k- kh-
VI ts- tsh-
VII tś- tśh-

I have standardized the notation of the initials of chapters IV and VII for simplicity.

Only Nishida (1964) and Arakawa (1999) reconstruct voiceless obsturents for the initials of chapter IV.

The identification of the third type is highly controversial:

Homophones chapter Solution 1: plain voiced Solution 2: prenasalized Solution 3: voiced aspirates Solution 4: prenasalized or preaspirated
I b- mb- bh- mb- or hb- or ɦb-
III d- nd- dh- nd- or hd- or ɦd-
IV ɲɟ- (Nishida and LXK only)
V g- ŋg- gh- ŋg- or hg- or ɦg-
VI dz- ndz- dzh-
VII dź- ndź- dźh- ndź- or hdź- or ɦdź-

Solution 1: Wang Jingru (1930), Gong Hwang-cherng (since 1981), Li Fanwen (1986), Arakawa (1999)

Solution 2: Nishida (1964), Sofronov (1968)

Solution 3: Li Xinkui (1980) (I have not seen this, so I am basing this column on the table of Li Fanwen 1986: 126-127 which has ɲɟ- instead of the expected ɟh-)

Solution 4: Tai Chung Pui (2008)

Some researchers proposed more than one voiced obstruent series.

Nevsky (1926) proposed an unaspirated/aspirated distinction: b- vs. bh-.

Kychanov and Sofronov (1963) proposed a four-way distinction between b-, bh-, mb- (high register), and mb- (low register). I could interpret the last two as mp- and mb-.

However, the Tibetan transcription evidence, taken at face value, seems to indicate that a single initial was transcribed in multiple ways:

Li Fanwen number Gong This site Rhyme Tibetan transcriptions
supporting solution 1 supporting solution 2 supporting solution 3 supporting an amalgam of solution 2 and 3
0433 bju biu R3 1.3 b-?, dbu ?[bu] (with d- indicating level tone) Hbu ?[mbu]
3791 bji bi R11 2.10 Hbii(H) ?[mbii] dbhi(H) ?[bɦi] (with d- indicating level tone) Hbhi(H) ?[mbɦi]

If there were really multiple voiced obstruent series, they should have been distinguished instead of confused in the transcriptions.

The figures below are from Tai's section on fanqie analysis which only include transcriptions with identifiable initial spellers. Dubious transcriptions (e.g., m- for a chapter III tangraph) and transcriptions with incomplete initials are excluded. Preinitials presumably indicating level tone and/or retroflexion are in parentheses.

Homophones chapter Gong Number of Tibetan transcriptions
Voiceless Prenasalized voiceless Plain voiced Prenasalized voiced Voiced aspirated Prenasalized voiced aspirate
I b- (d-)b-: 9 Hb-: 41 (d-)bh-: 5 Hbh-: 6
III d- t-: 1 (b/g/r-)d-: 11 Hd-: 9 dh-: 6 Hdh-: 2
V g- (b/r-)g-: 2 Hg-: 17 why no (H)gh-, (H)dzh-, (H)jh-?
VI dz- (g-)ts-: 7 Hts-: 1 (b/m/g-)dz-: 22 Hdz-: 20
VII dź- (b/g-)j-: 39 Hj-: 13

There is no clear pattern. Perhaps looking at all transcriptions would help. Picking the initials implied by the most frequent transcriptions would result in an improbable, imbalanced system:

Homophones chapter Gong Plain voiced Prenasalized voiced
I b- mb-
III d- d-
V g- ŋg-
VI dz- dz-
VII dź- dź-

I will continue to use Gong's simple voiced symbols (with slight variations), even though I now suspect that the third Tangut obstruent series was like the third series of the Wu dialects

... whose precise phonetic nature differs somewhat from region to region. In most of the northern Wu dialects, the third series has a lenis voiceless onset followed by breathy voice or murmur when they occur initially in a phrase [b̥ɦ, d̥ɦ], etc.; when they occur after another syllable in a phrase, however, they are fully voiced. In the southern part of Zhejiang this series is voiced throughout, and is without any perceptible breathy voice or murmur: [b, d, g], etc. (Norman 1988: 199-200)

Perhaps the Tangut third series was [pɦ tɦ kɦ tsɦ tʃɦ], which was unlike anything in Chinese, Tibetan, or Sanskrit:

Homophones chapter This site (phonetic) Transcribed in Tibetan as Transcribed in Chinese as Used to transcribe Sanskrit
I pɦ- (H)b(h)- *mb- with and without diacritic mv-? (!) (via Chinese which had borrowed Skt m as *m, which then became *mb in the northwest?)
III tɦ- (H)d(h)-, t- *nd- with and without diacritic, *th- (no examples)
V kɦ- (H)g- *ŋg-, *ŋg-k- (fanqie) g-
VI tsɦ- (H)dz-, (H)ts- *ndʒ-(t)s- (fanqie for *ndz- not in Chinese), diacritic + *ts- j- (via Tibetan-style Skt dz-?)
VII tʃɦ- (H)j- *ndʒ-tʃ- (fanqie), *ndʒ- with and without diacritic (no examples)

Tibetan H- and -h- could have been attempts to record Tangut [ɦ].

I briefly considered the possibility that HC- represented consonants that were fricatives like H (which was [ɣ] in Old Tibetan according to Hill 2005): e.g.,

Hb- = [β]

Hd- = [ð]

However, I would expect H- instead of Hg-, z- instead of Hdz-, and zh- instead of Hj-.

The pronunciation of H in Tangut period Tibetan is unknown. It might have become a silent letter that was used as a diacritic in the transcriptions.

The diacritic in the Chinese transcriptions would not have been necessary if Tangut simply had prenasalized obstruents like Chinese. They imply that Tangut period northwestern Chinese prenasalized obstruents were not perfect fits for the Tangut third series. (The third series of northwestern Chinese had merged with the second: e.g., *b- > *pɦ- > *ph-. Tangut contains Chinese loans dating from before and after the merger.) TAI CHUNG PUI'S "A STUDY OF TIBETAN PHONOLOGICAL TRANSCRIPTION IN TANGUT BUDDHIST FRAGMENTS"

is a PhD dissertation from last month. I didn't know it existed until Guillaume Jacques told me about it last night. He predicted it would keep me awake, and it did. I was away from my computer until late tonight and I couldn't stop thinking about it. It's a dream come true for me. The last line of Tai's English abstract sums up the attraction for me:

This manual provides the most detailed information ever on the Tibetan transcription of Tangut characters.

Tai proposes a consonantal system for Tangut based on Gong's with some revisions. Since Tai did not provide a single table outlining his system, I compiled one below. Consonants in bold differ from Gong's. The groupings (e.g., 'complex 1') are mine.

Homophones chapter -voice +voice
unaspirated aspirated simple complex 1 complex 2 nasal
I p- ph- Cb- m-
II w-
III t- th- Cd- n-
V k- kh- Cg- ŋ-
VI ts- tsh- Cdz-
s- z-
VII tś- tśh- Cdź-
ś- ź-
VIII ʔ- x- ɣ-
IX lh- l- ld-

(According to Gong 2003, his x- and ɣ- are symbols for [h] and [ɦ] which I consider to be the aspirated and simple voiced counterparts of glottal stop.)

There are three main differences between Tai's and Gong's systems:

1. The 'complex 1' initials which correspond to simple voiced obstruents in Gong's system; their C- represents "a weak nasal or glottal sound"

2. The 'complex 2' initials which correspond to l-, z-, ź- in Gong's system

3. The reconstruction of voiced fricative initials z- and ź- instead of s- and ś- for some syllables in chapters VI and VII of Homophones

I'll comment on these differences in future posts. ONE + SMALL = MANY MYSTERIES

The first tangraph of

tia thɑ gia tia (TRECD 1386-2; see here)

looks like


TT1075 lew R44 1.43 'one'

TT1730 tsɨ R31 1.30 'small, little; too, also'

even though it has no known meaning and does not sound like either of its apparent parts.

Its actual analysis from the Tangraphic Sea is


TT1076 tia R20 1.20 (transcription tangraph) =

'jewel' = right of TT3047 ɣɑ R17 1.17 'bag, pocket, sack' +

all of TT1730 tsɨ R31 1.30 'small, little; too, also'

TT3047 in turn consists of


TT3047 ɣɑ R17 1.17 'bag, pocket, sack' =

'jewel' = left of TT3052 noo R54 2.45 'bag, sack' +

all of TT1075 lew R44 1.43 'one'

'bag, sack' is obviously semantic, but 'one' has no semantic or phonetic resemblance to ɣɑ 'bag, pocket, sack'. And why is 'jewel' (following Nishida's [1966: 244] interpretation) on the left of both 'bag' tangraphs?

tia thɑ gia tia is obviously Skt tathaagata, but three other words written with TT1076 tia R20 1.20 have unknown etymologies:

TRECD 1386-1: tia niə 'tantra' (there is a slight phonetic resemblance to the Sanskrit, but I would expect tɑ̃ tʃa)

niə is a plural marker but must be phonetic here

TRECD 1386-3 tia ki 'virgin; goddess' (there is no similar Sanskrit word; is this a transcription of a Tibetan word?)

ki is a transcription tangraph

TRECD 1386-4 tia mieʳ (a Tangut surname; is a Tibetan origin possible?)

mieʳ (a surname tangraph; Homophones gave tia as its clarifier, suggesting that mieʳ may only occur as the second half of the surname tia mieʳ; TRECD 4688 listed no polysyllabic words beginning with mieʳ) TANGUT FOR TATHAAGATA

Sanskrit tathaagata 'thus-gone' was borrowed into Tangut as

tia thɑ gia tia (TRECD 1386-2)

rather than tɑ thɑɑ gɑ tɑ. Compare this with the Chinese transcriptions listed in Soothill and Hodous' A Dictionary of Chinese Buddhist Terms (readings in Middle Chinese; does MC *-d- for Skt -t- reflect a Middle Indic intermediary with intervocalic voicing?):

多陀阿伽陀 MC *tɑ dɑ ʔɑ gɨa dɑ (there was no MC *gɑ)

多他阿伽陀 MC *tɑ thɑ ʔɑ gɨa dɑ

多他阿伽駄 MC *tɑ thɑ ʔɑ gɨa dɑ(ʔ)

多他阿伽度 MC *tɑ thɑ ʔɑ gɨa dak/doh

cf. the nom. sg. tathaagataḥ and tathaagato depending on the phonological environment

多阿竭 MC *tɑ ʔɑ gɨat

怛闥阿竭 MC *tɑt thɑt ʔɑ gɨat

怛薩阿竭 MC *tɑt sɑt ʔɑ gɨat

怛他蘗多 MC *tɑt thɑ (?)

蘗 was MC *phih, *pɛk, *pek, or *bɨeŋ with labial initials!

the closest appropriate graph like 蘗 is 桀 MC *gɨet, but 怛他桀多 is not attested

Why was Grade IV R20 -ia used for Skt -a instead of Grade I -ɑ? This is understandable in the case of the third syllable if the transcription was influenced by Chinese, since 伽 MC *gɨa had probably become *khja with *-j- in Tangut period northwestern Chinese. However, neither Sanskrit nor Chinese has anything -i-like in the first or fourth syllables, and Tangut did have the syllable ta R17 (though only with the level tone).

The rhyme of TT5625 thɑ 'force, compel' is unclear. Precious Rhymes of the Tangraphic Sea listed tha under the entering tone, even though Skt thaa does not end in a stop, the defining characteristic of the entering tone. (22:44: But note that two of the Chinese transcriptions have an entering tone syllable corresponding to thaa. Could TT5625 thɑ be a Tangutization of 闥 MC *thɑt?)

Why does TT1883 gia R20 2.17 'I' have a rising tone if rising tones were correlated with length, as I hypothesized in "Emolument and Ego"? The syllable gia R20 1.20 does exist.

6.11.23:56: Arakawa might reconstruct tia thɑ gia tia as taa tha gaa taa. Although his reconstruction or R20 as -aa eliminates the problematic -i-, his three long vowels correspond to Sanskrit short vowels. EMOLUMENT AND EGO

I once thought that Gong's Tangut ʔj- should have been simplified to j-. However, the initial speller of this fanqie could not begin with a simple j-:


TT0449 ʔɑ R17 2.14 (dharani transcription character) =

reconstructed by Li Fanwen 1986 as ja without a glottal stop

TT0444 ʔiʳ (Gong: ʔjiʳ) R84 1.79 < *rji 'emolument (official salary)' +

TT3818 ŋɑ R17 2.14 'I'

TT0449 is a fanqie tangraph composed of parts of its fanqie spellers: 'language' on the left and an unknown element on the right.

I don't know why it was necessary to create TT0449 for ʔa since other tangraphs seem to exist for that syllable (though with uncertain rhymes; Homophones 44A26-44A28, 45B14; all four are in the level tone section of the Mixed Categories of the Tangraphic Sea):

TT3917 ʔɑ ?R17 1.17 'monk'

TT5485 ʔɑ ?R17 1.17 'mud boots'

is the left side a distortion of the left of 阿, the sinograph for Skt a?

Nishida (1966: 243) identified this element as 'far'

TT5479 ʔɑ ?R17 1.17 'one, a, big'

cognate to a 'one' in Qiang languages

TT0481 ʔɑ ?R17 1.17 (transcription character for Sanskrit a)

I don't know why TT0481 is listed separately from the other three in Homophones and Mixed Categories; the separation implies that they are not homophonous. I wonder if the first three are ʔɑɑ R22 1.22. One might expect ʔɑɑ R22 1.22 to be the closest Tangut equivalent of Skt aa, but the transcription tangraph for Skt aa was

TT0544 ʔaa R23 2.20 (with the left side of TT5109 'long' on the right)

(I can't find this in TRECD!)

with Grade II aa (and rising tone!). Maybe TT0481 was also Grade II: ʔa ?R18 1.18 (though the sinograph for Skt a was Grade I 阿 *ʔɑ!)

I also don't know why it was necessary to have both level tone and rising tone transcription tangraphs for ʔɑ, since classical Sanskrit has no tones. (TT0481 ʔɑ ?R17 1.17 definitely stood for Skt a, so TT0449 ʔɑ R17 2.14 < *ʔaH could have stood for Skt aḥ with visarga.)

A similarly puzzling pair is

TT5488 ʔo R51 1.49 (transcription tangraph for Skt o)

TT5487 ʔo R51 2.42 (transcription tangraph for Skt [unspecified by TRECD], 'sound of moaning'; left side of TT5109 'long' on the right)

which Grinstead identified as transcription tangraphs for (Skt?) a and aa. (Could these represent early Skt loans into Chinese that had undergone an a > o shift?)

Both Skt aa tangraphs (TT0544, TT5487) have rising tones. Could this mean that rising tone syllables were longer?

Like TT0449, TT5488 is another fanqie tangraph:


TT5488 ʔo R51 1.49 =

left of TT5479 ʔɑ ?R17 1.17 'one, a, big' +

left of TT4804 ʔo R51 1.49 'host; seller; master'

TT5488 is the initial speller of TT5487:


TT5487 ʔo R51 2.42 =

left of TT5488 ʔo R51 1.49 +

left of TT5109 dʒuo R53 1.51 'long'

borrowed from Tangut period NW Chn dʒo 'long' < Middle Chinese *ɖɨaŋ?

One might think these tangraphs represented Skt o, but Skt o is always long [oo], so there would be no need for tangraphs for short and long o.

TRECD (1508-1) listed

as 'transcription character [singular!] used in dharani', but ʔo ʔo is clearly not written with a single tangraph and there are no Sanskrit words like oo or aaa. SECONDARY GLOTTAL STOP IN TANGUT

After writing "Hlai Initial Glides", I realized that I had forgotten a key characteristic of my own reconstruction. I thought that Gong's -j- was suspiciously too frequent and reinterpreted his jV as rhymes with high vowels:

Gong's Grade III (he has no Grade IV) ju ji ja je jo
My Grade III (nonpalatal [first] vowel) u ɨi ɨa ɨə ɨe uo
My Grade IV (palatal [first] vowel) iu i ia ie io

This means that my equivalent of his ʔj- is really either ʔ- (Grade III) or ʔi- (Grade IV), not ʔj-.

Below, I've rewritten and expanded the ʔj- section of my table from last night:

Li Fanwen number Tangut gDong-brgyad rGyalrong Proto-rGyalrong root initial Gloss
3452 ʔie R37 2.33 qa-ʑo *j- sheep
3807 ʔie R37 1.36 kɯ-ʑo lightweight
1245 tɯ-ʑo self
2798 ʔiʳ R84 2.72 ɣu-rʑa *rj- (with irregular development in 'eight') hundred
4602 ʔiaʳ R87 1.82 kɯ-rcat eight

I hypothese that pre-Tangut *j(i)- merged with *ʔi- after jV-sequences were reinterpreted as iV-diphthongs:

'sheep': *j(i)eH > *ieH > ʔie R37 2.33

'lightweight', 'self': *j(i)e > *ie > ʔie R37 2.36homophonous with

merging with ʔie R37 < *ʔie(H) < *Cɯ-ʔe(H) with a glottal stop-initial root

'hundred': *rjiH > *rjiʳH > *jiʳH > *iʳH > ʔiʳ R84 2.72

becoming nearly homophonous with ʔi R11 < *ʔi with a glottal stop-initial

'eight': *rja > *rjaʳ > *jaʳ > *iaʳ > ʔiaʳ R87 1.82

Perhaps glottal stops were added to all pre-Tangut vowel-initial syllables:

Pre-Tangut After j > i reinterpretation Tangut
*j(i)e *ie ʔie
*jiʳ *iʳ ʔiʳ
*i *i ʔi
*jaʳ *iaʳ ʔiaʳ

However, I know of no language which distinguishes between je and ie. I prefer to think that a *j-/*ʔi- distinction was lost:

Pre-Tangut After j > i reinterpretation Tangut
*j(i)e *ie ʔie
*ʔie *ʔie
*jiʳ *iʳ ʔiʳ
*ʔi *ʔi ʔi
*jaʳ *iaʳ ʔiaʳ

Since I don't know of any language that distinguishes between ie and ʔie, I doubt that the middle stage lasted very long if it even existed at all.

I am now inclined to think that my Tangut v- is like Vietnamese v- which originates from a lenited labial stop and a labial glide:

Tangut Pre-Tangut Vietnamese Middle Vietnamese
v- *β- < *CV-P- v- (b- [β] < *C-P-
*w- v- < *w-

My Tangut ʔw- may be from earlier *qw- and/or *qV-P- (via *qV-β-). I'm not surprised that I could only find eight examples, since such sequences could not be very common (though I didn't expect to find two sets of homophones):

LFW0149 ʔweʳ R77 2.66 'protect'

LFW4976 ʔweʳ R77 2.66 'guard, defend'

6.11.00:31: from pre-Tangut ?*q-r-weH, a borrowing of a NW cognate ?*wej (cf. Kan-on wei) of Middle Chinese 衛 *wiəjh 'defend' with native affixes added?

LFW0175 ʔwi R11 2.10 'threshing ground'

LFW0645 ʔwọu R61 2.51 'help'

6.11.00:19: from pre-Tangut ?*qɤ-S-PuH, a borrowing of a NW cognate of Middle Chinese 輔 *buəʔ or 補 *poʔ 'help' with native affixes added?

LFW1056 ʔwọu R61 2.51 'tie'

LFW1609 ʔwọu R61 2.51 'domestic animal'

LFW2725 ʔwɔ R74 1.71 'round'

LFW4053 ʔwọ R73 1.70 'ice'

gDong-brgyad rGyalrong tɤ-jpɣom 'ice' suggests that the pre-Tangut root had a labial stop initial *p-; there must also have been a pre-Tangut acute prefix that conditioned the tense vowel: ?*qɤ-S-pom

Like Tangut, Vietnamese also has a ʔw- (spelled u- or o-) distinct from v-. HLAI INITIAL GLIDES

... don't exist according to Yuan Zhongshu (1994). Hlai dialects have ʔw- and ʔj- but no w- and j-. hw- and hj- might be voiceless glides rather than consonant clusters:

Baoding Tongshi Jiamao
ʔw- + +
ʔj- + + +
hw- + +
hj- +

The Xifang and Baisha dialects have no w-/j-like initials at all. All five dialects have the w-like fricative v-.

The absence of j- reminded me of Gong's Tangut reconstruction which doesn't seem to have any initial j-. (Cases of j- appear to be typos for ʔj- since their fanqie have initial spellers with ʔj-.)

Until now, I was inclined to reject Gong's ʔw- and ʔj- and rewrite them as simple w- and j-, even though they were classified as 'glottal' initials in Homophones. However, I now wonder if Tangut was like Baoding:

Baoding Gong's Tangut My Tangut until now My current Tangut Homophones initial class
v- w- v- v- labiodentals
ʔw- ʔw- w- ʔw- glottals
ʔj- ʔj- j- ʔj-

These initials have the following correspondences with gDong-brgyad rGyalrong in Jacques (2006):

Tangut gDong-brgyad rGyalrong Li Fanwen numbers
v- p- 5113, 3621, 294, 2625, 2926
pɣ- < *pk- 5134
mp- 320
sp- 385
rp- 5203, 5170
jp- < *lp- 4091
xp- < *kə-p- 2712
ph- 5109
jph- < *lph- 4585
mb- 923, 2140
w- 5049
sɣ- < *sə-w- 4966
ʁ- < *qw- 1697
rʁ- < *rq- (?*rqw-) 227 (labiodental initial uncertain; probably v-)
ʔw- jp- < *lp- 1805
jpɣ- < *lp- + velarized vowel 4053
ʔj- ʑ- < *j- 3452, 3807, 1245
rʑ- < *rj- 2798
rc- < ?*rj- (irregular) 4602

As I would expect, Tangut v- corresponds to gDong-brgyad rGyalrong labial stops. I believe that v- originated from *β, which in turn came from lenited labial stops in intervocalic position: *CVPV > *CVβV > vV. I cannot explain the instances in which Tangut v- corresponds to w-.

I don't know why gDong-brgyad rGyalrong jp- corresponds to Tangut ʔw- in two cases.

I presume that pre-Tangut *j- and *rj- merged into *j- which then became ʔj-. I don't understand why glottal stops developed in front of initial glides (if ʔw- < *w-, though the comparative evidence so far does not support this).

6.10.0:19: The Chinese surname 王 was borrowed into Hlai as ʔwaŋ with an initial glottal stop (Yuan 1994: 136). I wonder if the Hainanese pronunciation of 王 has an initial glottal stop. There is no glottal stop in Old Chinese *waŋ or generic Middle Chinese *wɨaŋ. HLAI PALATAL CODAS

The Baoding dialect of Hlai is the only Kra-Dai language that I know of with palatal codas. Such codas are typical of Mon-Khmer languages, which aren't spoken on Hainan. They are either independent innovations or retentions from some earlier stage (Proto-Hlai? Proto-Kra-Dai?). Their distribution is uneven:

Baoding rhymes ending in palatals (based on Yuan Zhongshu 1994: 4-10)

* * * * -uɲ -uuɲ
* * * * -uc -uuc
-eɲ * (-eeɲ?) ** *
-ec * ** -ooc
-aɲ -aaɲ
-ac -aac

* = Baoding has rhymes ending in velars and dentals but not palatals: e.g., it has -i(i)k but not -i(i)c; maximally palatal vowels cannot be followed by palatals.

(The rhyme -eeɲ appears on p. 232, but not in the rhyme charts, so I suspect it is a typo.)

** = Baoding has a rhyme ending in a velar but not a dental or a palatal: e.g., it has -oŋ but not -on or -oɲ.

Distribution of velar, palatal, and dental codas in Baoding (palatals in bold red)

-a- -aa- -e- -ee- -i- -ii- -ɯ- -ɯɯ -u- -uu- -o- -oo-
+ + + + + + + + + + +
-k + + + + + + + + + + +
+ + + ? + +
-c + + + + + +
-n + + + + + + + + + + +
-t + + + + + + + + + + +

There are also gaps in the velar and dental rows: no -uŋ, -uk, -on, or -ot. These four rhymes are in other Hlai dialects belonging to different subgroups (Bendi, Meifu, Jiamao [see the subgrouping in Yuan 1994: 13]), so either (a) Baoding lost these rhymes or (b) multiple branches of Hlai developed them (independently or through mutual influence). (a) seems to be the simpler option.

Hlai (presumably Baoding) palatals correspond to dentals in other Hlai dialects and the rest of Kra-Dai:

Comparisons beyond Hlai

'blood' (Yuan 1994: 24, 40 and Ostapirat 2000: 3, 220)

Hlai (Baoding?) ɬaac

Zhuang luut or lɯɯt (is one a typo?)

Dai lət or lɤt (is one a typo?)

Siamese lɯat

Kam phaat

Sanchong Sui phjaat

Proto-Kra *plat

'tail' (Ostapirat 2000: 6)

Hlai (Baoding?) tshuc

Sanchong Sui hət

Proto-Kra *C-tʃot

'to fly' (Yuan 1994: 40; I added Siamese)

Baoding ɓeɲ

Zhuang ɓin

Dai ɓin

Siamese bin

Kam pən

Sui win

Hlai-internal comparisons

'to buy' (Yuan 1994: 174)

Baoding tshac

Xifang, Tongshi, Baisha tshat

Jiamao taat

'good' (Yuan 1994: 173, 179)

Baoding ɬeɲ

Bendi (Baisha?) ɬen

Xifang, Qi (Tongshi?) ɬen (shared root in disyllabic word for 'beautiful')

(is this a loan from something like Cantonese 靚 leng 'beautiful'? does Hainanese [which is not a Yue language like Cantonese] have a cognate, or did it borrow the word from Cantonese?)

Is it really possible that Baoding is the only Kra-Dai language that has preserved final palatals? It's hard to understand why final dentals became palatals after non-palatal vowels (e.g., 'blood', 'tail', 'to buy'), unless those vowels were originally palatal: e.g., *-æt > Baoding -ac (palatality transferred to coda), non-Baoding -at (palatality completely lost).

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2008 Amritavision