In "A Gri-n Origin?", I examined the Tangut KNIFE word family which contained several phonological alternations:

initials: k- ~ kh- ~ g-

finals: -yi ~ -yịy ~ -yọ ~ -yụ

tones: level ~ rising

In "Phonological Alternations in Tangut" (1988), Gong Hwang-cherng gave 289 examples of many more alternations within Tangut word families.

In this series of posts, I will present my guesses concerning the origins of these alternations - and of the overall Tangut phonological system.

I assume that pre-Tangut was an atonal language with monosyllabic roots and a richer set of initials (including clusters and presyllables) and final consonants. In short, it was more like rGyalrong or Mawo Qiang. This is not to say that rGyalrong or Mawo Qiang is pre-Tangut preserved intact after many centuries. But as relatives of Tangut, it is possible that they retain some features lost in Tangut itself. Of course, the opposite also may be possible.

Why did Tangut become so unlike its relatives? Chinese influence may be part of the answer. Although all three Qiangic languages that I have mentioned were influenced by Chinese, Tangut was more Sinified than the other two. I believe that Tangut followed the same path that Chinese had a millennium earlier: i.e., it shifted from a edgy sesquisyllabic language to a nuclear monosyllabic language:

Language type





Many: clusters and presyllables

Few; no tones



No clusters without glides; no presyllables

Many with tones


The gDong-brygad rGyalrong (DG) and Tangut words for 'six' embody the contrast between the 'edgy' and 'nuclear' types:

Edgy: DG kU-ChëG

CV presyllable + CVC structure

possible vowels are limited:

only three in presyllables (U, ë, a)

only eight in main syllables (including a rare ü)

no tones

fricatives allowed as codas

Nuclear: TT3448 chhyiw 1.46

no presyllable

CVG structure

(the first glide -y- is automatic before -i when it is not followed by another vowel and therefore may not be phonemic here)

i is one of 38 (!) possible vowels in Gong's reconstruction

i has one of two tones

only glides allowed as codas

All languages need a certain number of phonological distinctions (PD) to function. Languages differ in terms of how their PD are distributed. An edgy language has many of its PDs in onsets and codas, whereas a nuclear language has many of its PDs in vowels and tones.

Perhaps one could call edgy languages consonocentric and nuclear languages vococentric. These terms do, however, unfortunately imply that vowels don't matter in consonocentric languages and consonants don't matter in nuclear languages. But all phonemes matter. Nonetheless, the functional load of a vowel in rGyalrong (8 vowels) is much less than the functional load of a vowel in Tangut (38 vowels).

The rich vocalic system of Tangut probably developed from a simpler system which acquired additional dimensions due to the influence of surrounding consonants. There are four basic types of Tangut vowels:

plain (14 vowels: most of rhymes 1-60: the major cycle)

glottalized or tense (7 vowels: rhymes 61-76: the first minor cycle)

I am not certain of their phonetic quality; all I do know is that these vowels were neither plain nor retroflex.

retroflex (13 vowels: rhymes 77-103 and 105: the second minor cycle)

nasalized (4 vowels: rhymes 15, 16, 25-27, 104)

The plain vowels were presumably inherited more or less unchanged from pre-Tangut.

The glottalized vowels (indicated with subscript dots) may reflect lost preinitials and/or presyllables: e.g.,

*C1V1-C2V2 > C22

I suspect that the lost initial elements were stops that assimilated and merged with the onset of the main syllable, creating a short-lived class of onsets that were like the reinforced consonants of Korean which also originated from earlier clusters. Korean reinforced kk-, cch-, tt-, pp- (written here with doubled letters) "are pronounced with great muscular tension ... The laryngeal tension continues on into the vowel" (Martin 1992: 27). In the future, Korean reinforced consonants may merge with their plain counterparts, and the once-automatic tension of the vowels will become distinctive: e.g.,

'time': early Korean ?*pʌsʌta(C)i > Middle Korean pstay > modern Korean ttẸ > future Korean tẹ (?)

(the E ~ e height distinction is currently weak, and will probably disappear.)

A similar sequence of events may have occurred in Tangut:

*C1V1-C2V2 > C1C2V2 > C2C22 > C22

The fanqie for glottalized rhyme cycle syllables tend to have glottalized-vowel spellers for the initials as well as the finals. This may imply that the initials of those syllables were phonetically distinct: e.g., k in kạ might have been more like a reinforced Korean kk-.

The retroflex vowels originated from plain vowels in syllables containing *r:

*r-CV, *CrV, *CVr > CV r

Nearly all pre-Tangut syllables with initial *r- acquired retroflex vowels (the autorhotic rule):

*rV > rV r

(But how do I explain r- followed by non-retroflex vowels: e.g., TT2594 rã 2.22?)

Gong does not reconstruct any Tangut vowels which are both glottalized and retroflex: i.e., r. This may mean that pre-Tangut syllables with complex onsets and *r ended up with glottalized or retroflex vowels. (Did *Ṿ r merge with either *Ṿ or *V r?)

Nasalized vowels (ã, ĩ, ẽ, ũ) are in borrowings of Tangut period northwestern Chinese words ending in nasal vowels. Gong (十二世紀末漢語西北方音韻母系統的構擬, 1995) also reconstructed those four vowels for TPNWC.

There are no glottalized or retroflex nasal vowels in Gong's reconstruction, probably because there were no such vowels in TPNWC. (Arakawa [1999: 40-41], on the other hand, reconstructed enq and onq which I suspect are simplified notation for glottalized nasal vowels [ẹ̃] and [ọ̃].)

Pre-Tangut had nasal codas which did not leave a trace as nasalization in 11th century Tangut: e.g.,

TT1718 THREE: ?*C(V)-som > ?*C(V)-sõ > sọ 1.70 (phonetically like Korean 쏘 sso with a reinforced initial?)

In the above scenario, I ordered vowel nasalization before vowel glottalization. The opposite order is also possible, but it would entail another change: the merger of glottalized nasal and nonnasal vowels:

TT1718 THREE: ?*C(V)-som > ?*sọm > ?*sọ̃ > sọ 1.70

The lost preinitial of THREE may have been *p- or *k-: cf. gDong-brgyad rGyalrong fsum < *psum 'three' in sqa fsum 'thirteen'* and Written Tibetan gsum 'three'.

If the changes proposed in this post actually occurred, then pre-Tangut

- only had (14?) plain vowels: the other three types of vowels developed later

- had complex onsets reflected in later glottalized vowels

- had more r-s on the edges (as preinitials, medials, and codas) reflected in later retroflex vowels

Next: Did pre-Tangut have more codas?

*The native gDong-brgyad rGyalrong word for 'three' has been replaced by xsum < *ksum, a loanword from Tibetan (cf. Written Tibetan gsum 'three'; Jacques 2003: 24). A GRI-N ORIGIN?

The tangraph transcribing Sanskrit jñii -

TT2551 KNIFE gyi 1.11

- not only had a velar initial, but was part of a word family whose members all had velar initials. In "Phonological Alternations in Tangut" (1988), Gong identified these cognates (pp. 787 and 821):

TT2668 KNIFE khyi 1.11 (Tangraphic Sea glossed TT2551 as TT2551.)

TT0667 砍 TO-CUT khyi 2.10 (homophonous with the above word except for the rising tone)

I've found three more:

TT2582 SHAVE gyịy 1.61 (this whole tangraph forms the right of TT1534 below)

(for an example of 1.11/2.10 ~ 1.61 alternation, cf. Gong, p. 807, #115:



TT1534 割 CUT / 芟 MOW kyọ 2.64 (< ?*C-kyi-o)

TT1524 REAP kyụ 1.59 (a back-formation?; cf. -yo verbs derived from -yu verbs)

I think their common root was *kyi. (I could rewrite this as *ki, since Gong automatically reconstructed y before i.) The unaffixed root has not survived as an independent word. All attested forms contain one or more affixes:

Voicing prefix (a nasal?) + *kyi = gyi 1.11

Aspirating prefix + *kyi = khyi 1.11/2.10

Prefix conditioning vowel tensing + *kyi + *-o = kyọ 1.59

Tonal alternations may have originated as alternations between a suffixed and nonsuffixed form.

gyịy 1.61 may had originated as a root plus three affixes:

Voicing prefix + (tensing prefix + *kyi) + ?*-e = gyịy 1.61 (phonetically [gyẹ]?)

(10.20.2:24: This is completely backward!)

I propose an *-e suffix (wrong! see above!) because the Tibetan transcriptions for rhyme 1.61 mostly contain -e.

Rhyme 1.61/2.54 has been reconstructed by others with mid vowels:

Nishida (1964): -Ẹ

Hashimoto (1965): -ye

Sofronov (1968): -yẹ

Shi et al. (1983): -iə̃, -ïẽ

Li Fanwen (1986): -yẸ

Arakawa (1999): -enq = [ẹ̃]?

Gong proposed that TT2551 KNIFE gyi 1.11 could be cognate to Written Tibetan gri 'knife' (p. 787). If the root of KNIFE was *kyi, this proposal could not work for two reasons.

First, the initials don't match. I would expect Tibetan kh- to correspond to pre-Tangut *k-.

Second, WT -r- should correspond to vowel retroflexion in Tangut. But none of the Tangut words in the KNIFE word family have retroflex vowels.

As far as I know, WT r- cannot be followed by k(y)i or g(y)i in the same syllable:


One could try to derive WT gri 'knife' from *r-gi with metathesis (and an *r-prefix absent from the Tangut word family), but I don't see why metathesis would only occur before *i. There's nothing special about i that would encourage r-velar sequences to reverse. Moreover, the root-initial g- would still not match the pre-Tangut *k-.

I have not been able to find any rGyalrong words for 'knife' that are cognate to KNIFE gyi 1.11.

Could Japhug rGyalrong kë-krëG or Somang rGyalrong kɐ-krɐ^k (both meaning 'to cut grass') be cognate to KNIFE gyi 1.11? I don't think so, because Japhug -ëG does not corresponds to Tangut -yi 1.11 (Jacques 2003: 13).

Mawo Qiang rəgi 'small knife' (Sun Hongkai 1981: 26) looks like the pre-WT *r-gi I proposed above and could be cognate to KNIFE gyi 1.11 if gi is the root.

Next: When and when not to reconstruct prefixes in pre-Tangut. CLEVER IRON

Although I couldn't find any evidence for Tangut palatals derived from velars, I did find a velar-initial tangraph used to transcribe a palatal-initial Sanskrit syllable!

TT2551 KNIFE gyi 1.11

for jñii in Skt raajñii 'queen' (Nevsky 1960 II: 657)

This couldn't have been the result of Tibetan influence, since I think the Tibetanization of jñii would be dzñHi. Nor could it reflect Chinese influence, as the only Chinese transcription I know for raajñii 'queen' is 羅闍 Middle Chinese *la jyæ (as if it were for raaja 'king'!). is pronounced as [gy] in modern Hindi, but I don't know if that [gy]-pronunciation already existed in the Tangut period, or if such a pronunciation was ever transmitted to East Asia.

Two independent sources leave no doubt that KNIFE was a velar-initial word. It was listed in chapter V (velars) of Homophones, not chapter VII (palatals). Its fanqie spelling in Tangraphic Sea has a velar-initial speller (also listed in Homophones chapter V):


TT4218 PASS-OVER gyu 1.3 + TT0881 WOOD syi 1.11

The left side of KNIFE is phonetic in these tangraphs for velar-initial syllables -

TT2557 FALL gyi 1.11 (homophonous with KNIFE)

TT2558 REGRET gyi 1.11

- which were both transcribed in Tibetan as Hgyi [nggyi] (Nevsky 1960 II.528-529).


looks like it should also be read as gyi 1.11, but it isn't, as it represented a Chinese loanword lyow 1.56. It was derived from


the top of TT2605 WISDOM zhyïr 1.86 +

the bottom of TT0075 SNAKE phio 2.43

(which has 乚 instead of ヒ on the bottom right; Nevsky [1960 II.529] lists a variant of DRAGON with 乚 like SNAKE, and this variant appears in the copy of Homophones partly reprinted in Sofronov [1968 II.244])

The title refers to the Tangraphic Sea analysis of KNIFE:


TT2551 KNIFE gyi 1.11 =

top of TT2599 IRON shyow 1.56 +

center* of TT1276 CLEVER ryiry 1.74

(borrowed from or cognate to 利 OC *rits 'sharp' [blade]?)

Next: Internal and external evidence for a velar initial in KNIFE.

*Nishida radical 188 which might mean DEER. See the "Oh Deer" series that ran from March 30 to April 7. VELAR SOURCES FOR TANGUT PALATALS?

Back in March, I was puzzled by the palatal initial of the Tangut word for 'Khitan':

TT0988 0952

chhyï 1.29 tã 1.24

(And I'm still puzzled by why KHITAN is written with WOOD atop each tangraph.)

I proposed that

the earliest Tangut word for KHITAN had a prefix ... which fused with khy- to become chhy-.

If pre-Tangut *C-khy- really did become Tangut chhy-, I would ideally expect the following correspondence between gDong-brgyad rGyalrong (DGR, which still has presyllables) and Tangut (which has lost its presyllables):

DGR presyllable + velar initial : Tangut palatal initial

I would not expect this correspondence to be absolute because there surely would be cases in which one or the other language added a presyllable absent from their common ancestor (Proto-Qiangic) or dropped a Proto-Qiangic presyllable. Perhaps it would be sufficient to find cases of

DGR velar initial : Tangut palatal initial

Do such instances exist? I couldn't find any in Jacques (2003):

DGR initial (capital Chh-, J- are retroflex; number of cases in parentheses; all with presyllables unless noted otherwise)Tangut palatal initial
(no cases)ch-
Chh- (1)chh-
ñcG- (1), nj- (1), nJ- (2), Gnj- (1), GnJ- (1)j-
sh- (2 + 1 without presyllable), shn- (1), shr- (1), zr- (1)sh-
sh- (1 without presyllable)zh-

This does not necessarily mean that no instances exist. If I had access to a vocabulary of DGR, I might be able to find DGR (presyllable +) velar-initial words that resemble Tangut palatal-initial words. However, such resemblances may be purely coincidental. Comparative Tangut studies are still in their infancy, so today's 'cognates' may turn out to include a lot of unrelated lookalikes.

Here are a few Tangut j-words that may have internal or external cognates with velars:

TT1450 TRUTH jyị̈ 2.74, TT4909 TRUTH jyiy 1.35 : cf. TT2087 TRUTH Giey 1.34

TT3841 TEN jyị̈ : cf. Old Chinese 十 *gip 'ten' (but a loan from Middle Chinese *jip is more likely)

The first set has a rhyme alternation (2.74 ~ 1.35 ~ 1.34) that I can't find in Gong's articles on rhyme alternations.

Next: Did a Tangut velar really transcribe a foreign palatal? DID PROTO-KOREANIC HAVE TWO LIQUIDS? (PART 4)

The story of how sinography was adapted to write Japanese and Vietnamese is fairly clear, though unsolved problems remain. What is less clear is some of the reasoning behind the Korean adaptation of sinography. When sinographs were used to write Korean, they occasionally had readings which do not seem to be related to their original Chinese reading or their Middle Korean translation equivalent: e.g.,

遣 'send':

Middle Chinese *khien'/h < Old Chinese *khen'(s)

MK Sino-Korean kyən R (= rising) or H

MK tr. equiv. ponay- LH

used to write earlier Korean -ko H 'and' (verbal suffix)

why not write this with sinographs pronounced -ko H in Korean: e.g., 故?

why not write this with sinographs meaning 'and': e.g., 而?

冬 'winter':

Middle Chinese *toung < Old Chinese *tung

MK Sino-Korean tong L

MK tr. equiv. kyəzʌr LH

used to write the second half of earlier Korean 不冬 antɯr 'not' (an- < àní 'not'?; 不 is Chn for 'not')

no Chn reading sounds like -tɯr, so why not use a sinograph for a noun whose Korean translation sounds like it?: e.g., 野 tɯr 'field'?

(Using sinographs for verbs with similar-sounding Korean translations - e.g., 舉 tɯr- L 'raise' or 入 tɯr- H 'enter' - could lead to confusion: e.g., 不舉 looks like 'not raise' and 不入 looks like 'not enter'. 不野 'not field' makes no sense in Chinese, so the reader would know that 野 would have to be read as Korean tɯr.)

耳 'ear':

Middle Chinese *ñzhi < *ñi Old Chinese *nəng'

MK Sino-Korean zi R

MK tr. equiv. kuy H

used to write the first half of earlier Korean 耳亦 ptʌryə 'only'

no Chn reading sounds like ptʌr, so why not write the first syllable with a sinograph whose Korean translation sounds like ptʌr?: e.g., 摘 ptʌ- H 'pluck'?

亦 'also':

Middle Chinese *yiek < Old Chinese *lak or *yak (Schuessler 2007)

MK Sino-Korean yək H

MK tr. equiv. sto H

used to write the second half of earlier Korean 耳亦 ptʌryə 'only' and the end of 猶亦ohiryə HLH 'rather' (Chn 猶 'still, yet' is semantic; has the Korean semantic tag ohiryO 'rather' today). 

why not write with a sinograph pronounced in Korean: e.g., 餘?

why not write ptʌryə 'only' with a sinograph meaning 'only': e.g., 唯?

良 'good':

Middle Chinese *lïang < Old Chinese *rang

MK Sino-Korean ryang R

MK tr. equiv. tyoh- R

used to write 16 different syllables, according to the appendix of Yu Chhangdon's Yi Dynasty Language Dictionary:

zero-initial: a, an, ay, ə, əy, i

n-initial: nang

r-initial: ra, rang, rə, ryə, ri

y-initial: ya, yang, yas, yə

why not write the vowel-initial syllables with sinographs pronounced with initial vowels in Korean: e.g., 阿 a?

One could try to explain away such cases by assuming that they originated as semantographs for early Koreanic words that were lost in MK: e.g.,

*ko- 'send'

*tɯr 'winter'

*ptʌr < ?*pʌtʌr 'ear'

*a- 'good'

However, the above words are as real as Tangut B: i.e., both only exist as implied artifacts of a script whose principles remain obscure. There is no other evidence for them apart from these sinographs.

I wonder if the two-liquid hypothesis can partly explain why 良 was used to represent a large variety of syllables. Did the vowel- and y-initial readings originate from earlier readings with a liquid initial that was later lost?

良 Early Middle Chinese *lïang, Late Middle Chinese *lyang

used to write earlier Korean *l(y)a-, *r(y)a-type syllables

which later became r(y)a-, (y)a-type syllables

Liquid loss in some varieties of earlier Korean may also explain variation in spellings of names:

(All sinograph pronunciations are given in Late Old Chinese.)

加羅忽 LOC *kæ la hwət (Koryosa 58)

加 阿忽 LOC kæ 'a hwət (Sagi 37)

for an original Koguryo ?*karahot > ?*kaʁahot > ?*kaahot?

娥 利英 LOC *nga lih 'ïang

娥伊英 LOC *nga 'i 'ïang

娥英 LOC *nga 'ïang (all Yusa 1)

閼 英 LOC 'at 'ïang (Sagi 1)

for an original Shilla ?*ariyang > ?*aʁiyang > ?*a(i)yang?

Some name variations involving zero ~ liquid alternations could be explained by assuming that the 'liquidless' character was read with an archaic Old Chinese liquid-initial reading: e.g., the name of the state now known as Kara:

加羅 LOC *kæ la (Saki 44) : 加耶 LOC *kæ yæ (Saki 1)

Was 耶 still read with an initial liquid OC *la in the dialect known to the person who created the transcription?

But perhaps some if not all of these cases involve Koreanic-internal variation: e.g., some pronounced ?*kar(y)a as ?*kaya, and wrote the latter as 加耶 LOC *kæ yæ.

Moreover, OC cannot account for all zero ~ liquid alternations. There is no evidence for a liquid in 阿 LOC *'a < OC *'ay or 伊 LOC *'i < OC *'i. 娥 LOC *nga may have had a final liquid in OC (?*ngal) but it was long gone by the time the 娥-transcriptions were devised (i.e., after Sino-Korean *ng- shifted to zero, enabling both 娥 LOC *nga and 閼 LOC *'at to represent Shilla *a-).

10.17.00:27: I am not sure what to make of this name change:

Early Shilla 刀良 LOC *taw lïang > Late Shilla 道安 *toan (its Sino-Korean reading) (Sagi 34)

Possibility 1:

The name was originally *toran(g), and 良 reflects its original *-r-.

The newer spelling with 安 reflects a version of the name that had lost its *-r-.

Possibility 2:

The name never had a liquid. 良 'good' represented a Shilla word *an- 'good' that was later written phonetically as 安. The *a-reading of 良 survived into later centuries (see above). DID PROTO-KOREANIC HAVE TWO LIQUIDS? (PART 3)

Alexander Vovin's 2006 article "Why Manchu and Jurchen Look So Non-Tungusic?" gives a surprising answer to the title question (pp. 255-256):

The usual explanation for the lexical and structural peculiarities of Manchu and Jurchen [distinguishing them from other Tungusic languages] is based on the assumption that they are due to the areal influence of Khitan ... this usual explanation, although undoubtedly correct, is not sufficient ... We have to consider first the interplay of history and linguistics Although the [Khitan] Liao empire was established in 916 AD, and the Khitan tribes certainly made themselves known on the northern frontier of China long before that, it was not before 926 AD that Liao destroyed the Parhae (Bohai) state, and incorporated the territories of South Manchuria and present-day North-Eastern Korea that were the homeland of Jurchen tribes, and who were before 926 AD subjects of the Parhae. It is unlikely that strong Khitan influence could start prior to this date. Nevertheless, when we look at the earliest extant Jurchen inscriptions from the twelfth century, we already see there all the peculiar typological features of Jurchen and Manchu. It seems unlikely that the Khitan influence alone could be credited for all these drastic changes, and that they all took place within 200 years. I think we should not overlook another possibility: namely, the influence from the East ...

The Parhae state was founded in 686 AD by Tay Choyeng, a former Koguryo general. Thus, we would expect that the Parhae language, or more precisely the language of the Parhae elite was exerting a considerable influence on the Jurchen language prior to the Khitan conquest of Parhae in 926 AD for more than 200 years. Although the linguistic and ethnic identity of the Parhae elite remains controversial (Janhunen 1996: 138-39, 152), and nothing remains today of the Parhae language in the form of texts, or even glosses, there are certain historical and linguistic facts that should not be overlooked. First, it is quite clear that there was a direct connection between the Koguryo and Parhae elites ... Second, Parhae incorporated most of the former Koguryo territory. Therefore, we should presume that the ethnic and linguistic properties of the Koguryo and Parhae elites were essentially the same ...

Then, what was the language that both the Koguryo and the Parhae elite spoke? Since we do not have any texts written in those languages, the answer should be based on loanwords found in the languages of the people, namely Jurchen, whom these elites ruled for almost six hundred years [starting c. 323 AD]. It turns out that apart from early loans from some Mongolic language (probably Khitan, as was pointed out before), Jurchen and Manchu also have a layer of words that is traceable to a Korean-like source ... I think that in all probability these words were just borrowed by Jurchen from a variety or varieties of Old Korean spoken by the Parhae and Koguryo elites.

These words are of great potential interest for the problem of Proto-Koreanic liquids because Jurchen and Manchu distinguish between l and r. If these Koreanic loanwords contain both l and r, that may mean that the Koguryo/Parhae variety of Koreanic had two liquids (at least phonetically, if not phonemically).

All agree that Middle Korean in the 15th century had one liquid phoneme, though there is some disagreement about its allophones. I believe that this phoneme was pronounced [r] in all positions: cf. Soviet Korean*, a.k.a. Корё мар koryO mar 'Korea language', whose name has a final -r instead of the final -l in standard Korean mal 'language'.

The MK word for 'root' was pùrhúy. Vovin reconstructs its Proto-Koreanic source as *purïkuy. However, a medial -r- is difficult to reconcile with the -l- of Manchu fulehe < *puleke, which is Vovin's first proposed Koreanic loanword. If Koguryo or Parhae had a medial -r-, the Manchu word should be furehe. Two other proposed Koreanic loanwords in Manchu have -r-:

Manchu biyoran 'cliff of red earth': cf. MK pìryə̀y, modern standard Korean pyOrang 'id.'

The MK form has lost a final nasal and is more innovative than modern Korean dialects.

Manchu f- corresponded to MK p- in 'root'. Manchu b- suggests Koguryo/Parhae *b-. Was an original voiced initial the source of the low pitch of the first syllable of MK pìryə̀y 'cliff': i.e., pre-MK *bi- > MK pì? (But other evidence suggests there was no *b in Koguryo.)

Manchu chechere- 'to press tightly': cf. MK chìchï̀r-, modern standard Korean chichirU- 'press down'

Unfortunately, Vovin was not able to find any Koreanic loanwords with liquids in Jurchen.

Here are my attempts to explain away the -l- in fulehe 'root' without reconstructing a second liquid *l in Koguryo/Parhae:

1. In Koguryo/Parhae, PK *purïkuy had become *pulkə, with syllable-final *-r becoming *-l before a stop, as in modern standard Korean: e.g., pulg-Un 'red'**. But if this were the case, then the Manchu form should have been fulke or fulhe (cf. the -lk- ~ -lh- variation in Manchu julkun ~ julhun 'cavity at the base of the neck' [Rozycki 1981: 11]).

2. In Koguryo/Parhae, PK *-r- became *-l- between certain vowel sequences. I don't know of any similar phenomenon in any other language.

3. Koguryo/Parhae had a single liquid which did not sound like *l or *r in pre-Jurchen (the ancestor of Jurchen and Manchu). Hence this liquid was borrowed inconsistently as *l and as *r.

4. fulehe 'root' was borrowed from one dialect of Koguryo/Parhae which had shifted PK *r to *l in intervocalic position (unlike any other variety of Koreanic known to me), whereas biyoran 'cliff of red earth' and chechere- 'to press tightly' were borrowed from another dialect which had preserved PK *r.

The first two scenarios are very unlikely. I prefer to leave open the possibility that 'root' was Koguryo/Parhae *puləkə with *-l-, though I cannot rule out the third and fourth scenarios.

Claiming a second liquid phoneme for Koguryo/Parhae on the basis of a single example is obviously dangerous. I have found a second possible example incorporating Koreanic morphology that cannot be explained internally within Manchu:

Manchu algi- 'to be famous, to become known': cf. MK àrGóy- 'inform' < ǎr- 'know' + causative -Goy-.

(The native Manchu verb stem for 'know' is sa-, and its causative is sa-bu-.)

Could algi- be from a Koguryo/Parhae *algi-? Were they told that their Koguryo/Parhae rulers were *algi- (i.e., famous)? If only someone could ... let me know.

*According to King (1987 and p.c. in Martin 1992: 28), standard Korean -l and -ll- correspond to Soviet Korean -r and -l-.

**The modern standard Korean descendant of PK *purïkuy 'root' is puri withoutt any trace of PK *-k- which lenited to -h- and then disappeared entirely. pulg-Un 'red' is an unrelated word chosen as a phonetically similar example of -l- + stop. DID PROTO-KOREANIC HAVE TWO LIQUIDS? (PART 2)

In "Some Preliminaries to Reconstructing Liquids in Earlier Korean" (1996: 1070), S. Robert Ramsey wrote,

... the l-stems in Classes 1 and 2 are virtually the only [Middle Korean verb] stems that are distinguished solely by pitch.

That sentence has a footnote:

Another syllable type that shows such a contrast consists of stems ending in the semivowel -y: many of these, however, are morphemically complex (cf. Ramsey 1991: 223).

If the pitch distinction in classes 1 and 2 can be traced back to an earlier segmental distinction (p. 1071; I've rewritten Ramsey's examples in the notation used on this site: e.g., r instead of his l) -

Class 1: MK kʌr- L 'exchange' < pre-MK *kʌr1-

Class 2: MK kʌr- H 'grind' < pre-MK *kʌr2-

(L = low pitch; H = high)

then could the pitch distinction in morphologically simple -y stems also be traced back to a similar distinction? In part 1, I proposed that Middle Korean Vy could have partly come from an earlier liquid-i sequence. This lost liquid may have influenced the pitch of the preceding vowel: e.g.,

MK CVy- with one pitch < pre-MK *CVr1i-

MK CVy- with the other pitch < pre-MK *CVy-

I am guessing that the lost liquid was Ramsey's first liquid because "the Class 1 l-stems behave ... like [stems ending in] obstruents" (p. 1070), and obstruents lenited intervocally between pre-MK and MK. If this is correct, then these two MK verbs were segmentally different in pre-MK:

MK mʌy- L 'tie / sew (on), attach' < pre-MK *mʌr1i-

MK mʌy- H 'remove weeds' < pre-MK *mʌy-

(The liquid in*CVr2i- verbs would have remained intact in MK. Hence MK -ri- verbs would have come from pre-MK -r2i- verbs.)

Next: The Tungusic test.

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2007 Amritavision