In part 5, I wrote,

Third, MK -Vri is common, and even if many cases are from PK *-Vti, it is diffficult to believe that PK had a lot of *-Vrʌy and *-Vrɯy.

I assume that certain sound sequences are more probable than others: e.g.,

ra > ri > rʌy, rɯy

open syllables with basic vowels > syllables with nonbasic vowels + glides

is more likely than

rʌy, rɯy > ri > ra

syllables with nonbasic vowels + glides > open syllables with basic vowels

If the actual distribution is very different from the expected distribution, this may imply that

an unexpectedly common sequence derives from an earlier expected common sequence

an unexpectedly uncommon sequence derives from an earlier expected uncommon sequence

For instance, in Mandarin, su is more common than sa because su < Old Chinese *sa, whereas Md sa is from OC *sap, *sat, and *sri'.

In Korean, n and r have similar behavior. In the modern language, both n and r become l when they are placed together: /nr/ > [ll]. Even in Old Korean, there are a few cases in which *n and a liquid (which I will provisionally write as *r) seem to be interchangeable: e.g., the various spellings for the name now rendered as 'Shilla':

-n- spellings: 徐那伐, 詞腦

liquid spellings: 斯盧, 斯羅, 徐羅伐

-n- + liquid: 新羅

(10.26.2:45: I would rather not reconstruct a nasalized liquid, though a nasalized lateral has been reported as an allophone of the modern Korean liquid [Jun 1999, 2000 and Seo 2004].)

Therefore I might expect the distributional patterns of n and r to be similar in the environments V_ʌy and V_ɯy in Middle Korean.

There is no known sound change

ʌy, ɯy > i after -Vn-

whereas the sound change

ʌy, ɯy > i after -Vr-

may have occurred. If so, then there should be no Vrʌy or Vrɯy in Middle Korean. We have already seen that there is at least one MK Vrʌy word, sorʌy 'sound'. If this is an isolated archaism that had not undergone the aforementioned 'rʌy-duction', I would expect more cases of Vnʌy and Vy than Vrʌy and Vrɯy. Conversely, if the 'rʌy-duction' never occurred after r, the number of cases of ʌy and ɯy after n and r should be similar.

Here is a list of all the premodern Korean nʌy / nɯy / rʌy / rɯy words known to me:

nʌy H 'smoke' (1459 wOl 9:7)

onʌy 'notch of an arrow' (1720? wEO yuhE sang:40)

munɯy 'continually' (1514 sok samgang hEngshilto yOl:25)

not related to modern munɯy [muni] 'pattern' which might be related to Sino-Korean 紋 mun 'pattern'

arʌy RH 'before; early' (1459 wOl 1:33)

kyərʌy 'tribe' (1720? wEO yuhE sang:13)

sorʌy LH (1459 wOl 1:33)

tarʌy 'saddle flap' (1834? yu-sshi mulmyOnggo 1 mo)

nirɯy- 'be dull' (1775? han-chOng mungam 173b)

murɯy 'hail' (1720? wEO yuhE sang:2)

This list may not be complete. So far, there are more words with rʌy / rɯy than nʌy / nɯy. Some of these r-forms may in fact have been Proto-Korean *tʌy / *tɯy words or even *tʌri / tɯri words. But the others may reflect PK *rʌy / *rɯy. It's still not clear whether any of the r-forms are archaisms.

Next: -ɯy -ʌy-gain.

10.26.00:43: It is, however, obvious that most of these forms have attestations dating after the loss of ʌ in the 16th century, so it is not certain whether their spellings with ʌ are historically accurate or not. (This is why no pitches are indicated for the later forms; pitches are not marked in any texts after the 16th century.)

Martin (1992: 26) thought that ɯy was still distinct from i "into the 19th century, at least", so I assume the ɯy spellings are accurate.

10.26.00:48: I forgot to mention that there are many more premodern Korean words with mʌy and even some with mɯy. I mentioned two of them back in part 1. Presumably this abundance of m-forms is partly due to merger of original *mVy forms with *mVri forms via *r-loss. DID PROTO-KOREANIC HAVE TWO LIQUIDS? (PART 6)

Regardless of the answer to that question, it is certain that some instances of later Korean r after a vowel and before i are from earlier *t: e.g., the Early Middle Korean form equivalent to later məri 'head' was phonetically transcribed in Chinese as 麻帝 *mati with *-t-, not *-l-. The question is whether *ti is the sole source of later ri.

In part 5, I rejected Proto-Koreanic *rʌy and/or *rɯy as potential sources for later ri for three reasons. The first two aren't as strong as I thought, because:

- It's possible that Sino-Korean readings were exempted from the sound change that reduced *rʌy to ri. Readers of SK might have consistently pronounced rʌy-graphs with -ʌy, even after a vowel:

in native words: *-Vrʌy > -Vri

in SK words: *-Vrʌy > -Vrʌy

Cf. how *-t- became -r- in intervocalic position in native Korean words but not in most SK words. Nonetheless, there are a few SK words with *-t-lenition, so one might expect one or two SK words that had undergone 'rʌy-duction'. No such words have been identified.

- It's possible that-ʌy and -ɯy did not simplify to -i after r because the rule only applied within morphemes and not across morphemic boundaries. This would allow those endings to maintain consistent shapes.

(And that would also explain why the rule didn't apply to Sino-Korean, since there would be a morphemic boundary between a SK rʌy and a preceding vowel.)

I'll deal with my third reason tomorrow.

For now, I want to mention a category of evidence that I totally overlooked: native Korean words ending in -Vrʌy and -Vrɯy. If *-Vrʌy and *-Vrɯy became Middle Korean -Vri after a vowel within morphemes, then there should be no morphemes ending in -Vrʌy and -Vrɯy. However, Yu Chhangton's Yi Dynasty Korean dictionary has this triplet for 'sound':

sori LH (1447 sOk 19:14; still sori in modern Korean)

sorʌy LH (1459 wOl 1:33)

sorɯy LH (1586 sohak OnhE 6:91)

The last form is found in later texts from the 16th to 18th centuries, so I will disregard it, though it's possible that it did exist in earlier centuries.

One might look at the first two and say, aha, that proves that sorʌy became sori. Alhough sori is actually attested a decade before sorʌy, the latter must be more archaic, since there is no known change of -i to -ʌy in Korean. It is, however, not clear whether sorʌy > sori exemplifies

- a regular pre-MK sound change: *-Vrʌy > -ri

reflected by sori but not by sorʌy, which would be an archaism

- or a sporadic post-MK change: -ʌy > -i (instead of the expected -E)

cf. timtshʌy which irregularly became kimchhi instead of the expected chimchhE

Next: Nasal numbers and rhotic ratios. DID PROTO-KOREANIC HAVE TWO LIQUIDS? (PART 5)

In part 1, I pointed out that

- some early Koreanic words seem to have lost a liquid before i in Middle Korean

- yet Middle Korean still had words with ri: e.g., tsari 'seat' (instead of tsay).

I came up with two ways to explain such words:

1. Their ri comes from an earlier *ti

2. Their ri comes from Proto-Koreanic liquid 1 + *i, whereas -i- and -y partly come from Proto-Koreanic liquid 2 + *i.

I would now change the second solution to

2. Their ri comes from an earlier *ti and a Proto-Koreanic *li, whereas -i- and -y partly come from Proto-Koreanic *ri.

Here's what would have happened to three hypothetical PK words in Middle Korean:

PK *ati > MK ari (*t lenited intervocalically)

PK *ali > MK ari (*l became r)

PK *ari > MK ay (*r was lost before i)

I almost added a third solution when I wrote part 1:

3. Their ri comes from an earlier *ti and a PK *rʌy and/or *rɯy, whereas -i- and -y partly come from PK *ri.

This solution avoided proposing a second liquid for PK. Although it seemed more economical at first, I rejected it for three reasons.

First, Sino-Korean contains readings like rE < MK rʌy reflecting a Middle Chinese *ləy (phonetically closer to [lʌy]?). If PK*-Vrʌy became MK -Vri, then there should be Sino-Korean words ending in -ri instead of -rE < MK rʌy: e.g.

如 來 'Tathagata' should be yOri rather than yOrE < MK zyərʌy

But I know of no such words. This indicates that PK -ʌy was not reduced to -i after a vowel-liquid sequence.

Unfortunately, the SK reading rɯy does not exist, so there are no SK words ending in -Vrɯy. That sequence can, however, be found in native noun-affix sequences (see below).

Second, when MK words ending in -Vr are followed by -ʌy and -ɯy, those vowel-glide sequences are not reduced to -i: e.g.,

parʌr + ʌy (1482 kUm-sam 3:3b; found in Martin 1992: 744) should be parʌri rather than parʌrʌy

ir + ʌy (1447 sOk sO:5b; found in Martin 1992: 923) should be iri rather than irʌy

ir + ɯy (1459 wOl 1:11; found in YCT 1964: 604) should be iri rather than irɯy

Third, MK -Vri is common, and even if many cases are from PK *-Vti, it is diffficult to believe that PK had a lot of *-Vrʌy and *-Vrɯy.

Next: Sound reasons to reject my rejection. ALTERNATIONS FROM AFFIXES? PART 3: TWO GOOD TO BE TRUE, OR BAD TO THE BONE?

If the Tangut rising tone originated from lost laryngeals as I proposed in part 2, I would expect the following pattern of correspondences:

TangutPre-TangutgDong-brgyad rGyalrongWritten BurmeseWritten TibetanOld Chinese
rising tone*-'zerocreaky tone -' < *-'zero*-'
*-h < *-(C)s-sbreathy tone -h < *-s-(C)s*-(C)-s

These two examples almost work.

GlossTangutPre-TangutgDong-brgyad rGyalrongWritten BurmeseWritten TibetanOld Chinese
'moon'TT3313 lhyịy 2.60?*slye'slala'sla? 夕*slak 'evening' (not the expected *sla')
'bone'TT2314 ryiry 2.68?*rïesshë-rU (not rUs!)ruih < *-srus(None? Starostin proposed 骸 which is not a good phonetic match [OC *rgə]; the expected cognate would be *rus)

(It is not certain whether 夕 OC *slak 'evening' really is cognate to the other words. See Schuessler [2007: 522]. If it is cognate, then it is another example of the OC *-k : non-OC zero correspondence in 'hundred' from part 2.)

However, 'bone' is not as solid as it seems. There is another more common word for bone, TT3936 ryïr 1.86, which has the level tone (and a different vowel). Which word is original? I would guess that the level tone word - the more common word - was original, and that the rising tone word was derived via affixes that also altered the vowel quality. (Or did pre-Tangut have root vowel alternation?)

pre-Tangut ?*rï > ryïr 1.86

pre-Tangut ?*rï-e-s > ryiry 2.68 (phonetically [ryery?])

(Tibetan transcriptions of 2.68 have -e, indicating a mid vowel.)

(Was the suffix -e a grammaticalized word?)

A non *-s root for 'bone' agrees with gDong-brgyad rGyalrong shë-rU.

'Two', 'seven', and 'know' are messy: some languages reflect an earlier *-s and others don't:

GlossTangutPre-TangutgDong-brgyad rGyalrongWritten BurmeseWritten TibetanOld Chinese
'two'TT0590 nyïï 1.32 (with level tone!)?*nyïïʁnUshnac (< *snit)gñis*nis
'seven'TT1986 shyạ 1.64 (with level tone!)?*C-shyakU-shnUsku'-hnac (< *snit)(bdun is non-cognate; Matisoff cited stis from the Bodic language Kanauri)*s-hnit
'know'TT2058 syï 2.28 (also cf. TT1272 KNOWLEDGE syiy 2.33)?*sïs (> ?*sï-e-s 'knowledge')kë-sUssi' (not sih, but other Lolo-Burmese languages imply Proto-LB *sih)shes? 思 'think' (acc. to Starostin; vowel is unexpected; no -s!)

I don't know whether forms like gDong-brgyad rGyalrong kë-nU-fse and Zbu rGyalrong kɐ-nə-fsî and nə-fse' (all 'recognize; be familiar') are related to the above words for 'know'.

Jacques (2003) lists six Tangut cognates for gDong-brgyad rGyalrong words ending in -s. I have listed three of them and added a seventh ('know') above. Here are the other four:

gD tU-Gmas 'wound' : TT2293 myaa 1.23 <*myaa 'scar' (with level tone!)

gD tU-las 'forehead' : TT4734 lyạ 1.64 < *C-lya 'id.' (with level tone!)

gD tU-rtshës 'lung' : TT2322 tsə̣ 1.68 < C-tsə 'id.' (with level tone!)

gD mUs 'sort' : TT1771 2.25 < məs 'id.' (finally, the expected rising tone)

Out of these seven, five have level tones instead of the expected rising tones. One could claim that rGyalrong added an *-s suffix to those five words, whereas pre-Tangut did not.

Without knowing very much about Qiangic historical phonology, I would reconstruct these words without *-s at the Proto-Qiangic level and assume that -s was sometimes added later.

The *-s mismatch problem also occurs between Proto-Karen (as reconstructed by Benedict 1979 [cited in Matisoff 2003: 477]) and ST languages outside of the Lolo-Burmese subgroup:

GlossProto-KarenTangutPre-TangutgDong-brygad rGyalrongWritten BurmeseWritten TibetanOld Chinese
'four'*lisTT3509 lyïïr 1.92?*r-lïïkU-Bde < ?*pə-tleyleh < *liyh < *-sbzhi < *blyi*s-hlit-s
'five'*ngasTT3119 ngwə 1.27?*ngwəkU-mngu < *mə-ng-ngah < *-slnga*nga
'nine'*'kusTT3568 gyïï 1.32 ?*ngkïïkU-ngUt (with -t!? Somang has kə-nggû)kuih < *-sdgu*ku' ~ kwə'

Note, however, that all of the above examples are numerals, and as Matisoff [2003: 478] pointed out, "there is a strong tendency for numerals to influence each other's form." In any case, the point is that one cannot expect absolutely perfect correspondences involving -s in ST languages. This makes it extremely difficult to test a partial sibilant origin of the Tangut rising tone using the comparative method. ALTERNATIONS FROM AFFIXES? PART 2: REASONS TO RISE

In part 1, I proposed that pre-Tangut had presyllables, more complex onsets, and final -r. Comparison with other languages indicates that it also had at least two other types of codas that were later lost: nasals and stops. A minimal inventory of pre-Tangut codas might be:

Glides: ??*-y, ??*-w

These are the only two possible codas in Tangut. In some cases, Tangut -w is from pre-Tangut ?*-k, ?*-ng, ?*-m, ?*-p. It's not clear whether all instances of these glides are secondary or not. See Jacques (2003) for gDong-brygad rGyalrong words with -G (< *-k), (< *-ng), -m, -B (< *-p) corresponding to Tangut -w.

Liquid: ?*-r

Nasals: ?*-ng, ?*-n, ?*-m

Stops: ?*-k, ?*-t, ?*-p

Dental codas seem to have left no trace in Tangut, unlike velar and labial codas. Perhaps dental codas were lost very early.

This list is not that different from Guillaume Jacques' (2004: 266) list of proto-rGyalrong codas:

Glide: *-y

Liquids: *-r, *-l

Nasals: *-ng, *-m (no *-n!)

Stops: *-q, *-k, *-t, *-p

Fricative: *-s

It is also similar to the inventory of codas in Old Chinese:

Glides: *-y (partly or wholly < *-l), *-w

Liquid: *-r (see above for *-l)

Nasals: *-ng, *-n, *-m

Stops: *-', -kw, *-k, *-t, *-p

Fricative: *-s

with one major difference: OC permitted combinations of codas with *-' and *-s. These two (post)codas became the sources of the 'rising' and 'departing' tones of Middle Chinese. Syllables with nonglottal stop codas developed the 'entering' tone and all other syllables developed the 'level' tone:

OC *-' > MC rising tone

OC *-(C)-s > MC departing tone

OC *-kw/k/t/p > MC entering tone

all other OC syllables > MC level tone

Tangut only had three tones: 'level', 'rising', and an extremely rare 'entering' tone. The names of the Tangut tones were translations of the Chinese terms and may or may not have been descriptively accurate. (Cf. how the 'level' tone is now rising or falling in some modern Chinese languages.)

In Chinese usage, the 'level' tone is the default tone, and 'level' tone syllables outnumber syllables in each of the other categories. My 1967 reprint of the Middle Chinese rhyme dictionary Guangyun has 526 pages, excluding the front and back matter. 40% of the book is devoted to 'level' tone syllables, and syllables of each of the other tones are nearly evenly distributed among the remaining 60%:

ToneNumber of pagesPercentage of book
Level < OC syllables without stops or *-s21240.3
Rising < OC *-'10620.2
Departing < OC *-(C)-s10820.5
Entering < OC *-kw/k/t/p10019.0
All tones526100.0

The Tangut 'level' tone similarly dominates the Precious Rhymes of the Tangraphic Sea:

ToneNumber of pages (each pair of pages counts as 1)Percentage of book
Level26 + 4 in 'mixed categories'60.0
Risingnearly 16 + 4 in 'mixed categories'39.7
Enteringless than 1 (only 11 syllables)0.3
All tones50100.0

The 11 'entering' tone syllables also appear in Homophones. Gong reconstructed four of them with the rising tone, presumably on the basis of their placement in Homophones. He did not reconstruct a tone for the rest, but I suspect that they too had alternate readings with the rising tone. Moreover, the entering tone syllables are listed at the end of the PRTS rising tone volume. This indicates that the rising and entering tones may have been phonetically similar.

I would go further and suspect that they both had similar origins. The Tangut entering tone syllables may have been a handful of words which still had final stops lost elsewhere in the language. Could the Tangut rising tone have arisen from a pre-Tangut final glottal stop *-' and/or a final fricative *-s?

Suppose that pre-Tangut had a coda distribution similar to that of Old Chinese:

40% without a stop or ?*-s

20% final glottal stop

20% final ?*-s

20% final stops other than glottal stop: ?*-k/t/p

Now suppose that syllables with final stops other than glottal stop joined the default category and that the rising tone originated from lost final laryngeals:

60% level tone (< 40% without a stop or ?*-s + 20% final stops other than glottal stop)

40% rising tone (< 20% ?*-' + 20% ?*-h < *-(C)-s)

This is more or less the 3 : 2 ratio found in PRTS.

Some sample derivations:

1a: Level tone from final open syllable:

TT2297 NOSE ?*naa > nyii 2.12

cf. gDong-brgyad rGyalrong tU-shna 'nose', WT sna 'nose'

1b: Level tone from final nasal:

TT1718 THREE ?*C(V)-som > sọ 1.70

cf. the second half of gDong-brgyad rGyalrong sqa-fsum 'thirteen', WT gsum 'three'

1c: Level tone from final stop:

TT3448 SIX ?*h-trVk > chhyiw 1.46

(?*h- is a hypothetical 'aspirator' prefix that I will explain in a future installment.)

cf. gDong-brgyad rGyalrong kU-ChëG < *trOk 'six', WT drug 'six'

2a: Rising tone from final laryngeal ?*-' or?*-h (< *-s)

TT5770 HUNDRED ?*rya'/h > 'yir 2.72

cf. gDong-brgyad rGyalrong Gurzha < *rya 'hundred', WT brgya 'hundred'

It's tempting to correlate a pre-Tangut final laryngeal with the final *-k of 百 OC prak, but are there any other cases of Tangut rising tone : OC *-k : zero in non-Sinitic Sino-Tibetan languages? (10.21.22:39: Maybe. See here.)

2b: Rising tone from final stop-s cluster:

TT2384 PANTHER ?*r-zek-s > ?*zerɰh > zerw 2.78

cf. gDong-brgyad rGyalrong kU-rtsëG < ?*rtsek 'panther', WT gzig 'panther'

If the rising tone is from a final laryngeal, then rising tone cognates of level tone words arose from the addition of suffixes: e.g.,

TT2668 KNIFE khyi 1.11 < ?*h-gi

TT0667 TO-CUT khyi 2.10 < ?*h-gi-s

Next: The Tibetan test. And does rGyalrong prove me wrong?

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