11.4.2.23:59: UCHINAAGUCHI PART 12: ROE FROM ROPE

Last night, I presented two scenarios to explain why Okinawan ichitai 'five people' looks like ichi 'five' + tai 'two people'. Thanks to Sven Osterkamp for convincing me that solution B is correct. He reminded me of obsolete Japanese forms like mitari 'three people', etc. with -tari. I couldn't find them in Iwanami kogo jiten last night, so I wasn't sure how old they were. But today I found them (a) in IKJ under the entry -tari, (b) online, and (c) listed as cognates of the Okinawan forms in Sakihara (2006). Triple d'oh.

One huge problem with solution A is that there's no reason for anyone to misanalyze *putari 'two people' as *pu-tari. There are no compounds with reflexes of *pu- for 'two'. Also, there are no forms like mi-ri for 'three people' preserving the supposedly original suffix -ri, though last night I was thinking that the misanalysis took place at the Proto-Japonic level so no attested languages ever had mi-ri, etc.

(4.3.00:39: And lastly, it's simpler to assume that only 'one person' and 'two people' changed rather than 'three people' and up.)

I asked if "consonance [was] really so bad" in solution B. The only similar case I could think of last night was the reduction of -shiku-class adjectives in Japanese. In classical Japanese, -ku and -shiku adjectives have completely identical paradigms except in the conclusive form:

The original conclusive of the -shiku class was probably *-sisi which was later simplified to -si (now -shi) to avoid two consecutive si. But I rejected that precedent for haplology last night because it involved identical syllables, whereas the reduction of 'one' involved syllables with different vowels:

*pitə-tari > *pitəri

(but Nihon shoki has pindari < *-tari with reduction of the root rather than the suffix!)

Perhaps *puta-tari was the first to reduce, with *pitə-tari reducing by analogy to an 'adjacent' word. Note that 'nonadjacent' *itu-tari 'five people' did not reduce to itari in Japanese or itai in Okinawan: the actual forms are J itsutari (now replaced by Sino-Japanese gonin) and O ichitai (coexisting with Sino-Okinawan gunin). Perhaps the u of tu in *itu-tari wasn't sufficiently similar to the *a or *-tari to motivate haplology.

Now let's get all those people out of the way and get to the title. You now have all the rules you need to explain the derivation of Uchinaa 'Okinawa' except one. What is the additional rule? And what is the Okinawan cognate of J kawa 'skin'?

11.4.1.23:56: UCHINAAGUCHI PART 11: 5 X 2 ≠ 5

From last night: Bonus questions for review:

Because t became ch after i: itt > ichch (which I respell as the less awkward-looking itch).

- What normally happens to ai in Okinawan?

It became ee.

- Why is the -ai of [the Okinawan words for] '... people' exempt from that fate?

Because it's from -ari: cf. the -ari of Japanese futari 'two people'.

There have been three layers of ai in the history of Okinawan:

Modern Okinawan ari is from earlier *are, not Proto-Japonic *ari.

Now for the seemingly odd math: e.g., Okinawan ichitai 'five people' looking like ichi 'five' + tai 'two people'. I'd like to reconstruct a consistent method for counting people in pre-Okinawan. Here are two solutions. Neither is convincing.

Solution A (Why this is wrong)

The original Proto-Japonic suffix for 'person' was -ri as in Japanese.

'One person' was *pitə-ri and 'two people' was *puta-ri.

The latter was misanalyzed as *pu-tari (why?) and this new suffix *-tari was added after 'three' through 'five'.

The first syllables of *pitə-ri and *puta-ri were lost, ultimately resulting in Okinawan chui and tai - the latter resembling the suffix -tai < *-tari. (Amami retains -r- after -i: churi 'one person', taari 'two people', mitchari 'three people'.)

4.2.1:55: The -i- of 'one person' must have been lost after it caused the following -t- to palatalize to ch-:

-it- > -ich- > ch-.

Solution B (Why this is correct)

The original Proto-Japonic suffix for 'person' was *-tari.

In later Proto-Japonic, *-tari was reduced to *-ri after numbers ending in -tV to avoid a -tVta sequence (is consonance really so bad?).

Ryukyuan languages retained *-tari after 'three', but Japanese lost it and replaced it with Sino-Japanese 人 -nin.

As in solution A, 'two people' came to resemble the suffix -tai < *-tari. In both scenarios, the suffix *-tari was added to numerals before 'two' lost its first syllable and became a homophone.

In any case, the question is, why do 'one/two people' have *-ri whereas 'three' and up have *-tari?

11.3.31.23:59: UCHINAAGUCHI PART 10: 5 X 2 = 5?

People are counted in a partly puzzling manner in Okinawan. 'Three people' through 'five people'* contain what appears to be 'two people':

Is that a correct analysis? If not, what's really going on? My thoughts tomorrow.

*I have excluded 'six people' and up since they are transparently based on Chinese roots: e.g.,

O rukunin : J rokunin 'six people' < Chn 六 'six' + 人 'person'

Bonus questions for review:

- What normally happens to ai in Okinawan? Why is the -ai of '... people' exempt from that fate?

11.3.30.23:20: BEYOND UCHINAAGUCHI: MURE ETYMOLOGIES

In my last two entries, I mentioned the cognate pair

Okinawan mui 'mountain' : Japanese mori 'forest'

Which meaning, if any, is older? These words resemble

Old Japanese mor- 'pile up'

Old Japanese mure 'mountain'

and forms on the Korean peninsula:

Middle Korean moyh, moro 'mountain'

Paekche *mure (written as 武禮 in NIhon shoki) 'mountain' < *mora-i 'high thing'?

Paekche *mora- (毛良 in a toponym) 'high'

See Bentley (2000: 426, 430) for further discussion of the Paekche forms.

I suspect that the Japonic forms are borrowed from peninsular (i.e., Koreanic) language forms for 'mountain':

1. Proto-Japonic *mori 'mountain' was borrowed from some Koreanic language.

I regard the meaning 'forest' (< 'what grows on mountains' - though not all mountains!) as a Japanese innovation.

2. Paekche *o rose to *u (Vovin 2010: 143). The spelling 毛良 *mora for 'high' may predate this shift, whereas Old Japanese mure 'mountain' was borrowed from Paekche *mure 'mountain' after the shift.

Vovin provides another example of this shift: Paekche 久麻 *kuma 'bear', borrowed into Japanese as kuma. Paekche's 'niece' Middle Korean kom preserves the original *o. (Middle Korean is descended from a 'sister' of Paekche.)

Early Sino-Japanese (i.e., 'Go-on' stratum) -u corresponding to Middle Chinese *-o (e.g., 苦 ESJ ku : MC *khoʔ 'bitter') may have been borrowed through Paekche after raising. (Western Old Japanese also raised *o to *u, so it's also possible that 苦 was borrowed from Paekche as *ko which was raised to ku within WOJ.)

3. In pre-Middle Korean, a liquid was lost between *o and *i (cf. Okinawan -ui < *-ori) resulting in moyh rather than *morih. MK -h may be from a *-k suffix. Vovin (2010: 29) proposed that Proto-Korean verb stems ending in *-k became MK -h stems, and I modify his proposal to account for MK noun stems in -h.

11.3.29.23:48: UCHINAAGUCHI PART 9: A MUDDY MESS

From last night: Is the gw in the following Okinawan word a retention or an innovation?

O mingwi : J nigori 'muddiness'

cf.

O -gwii : J -goe 'voice'

O mui 'mountain' : J mori 'forest'

O migui : J meguri 'circulation'

It's an innovation. kw and gw were introduced to Japonic through Chinese loanwords and Okinawan even developed them in native words: e.g., O -gwii < *-goe < *-N-kəwe.

I think the gwi of O mingwi 'muddiness' is an irregular contraction of an earlier *-ori.

O migui 'circulation' indicates that -gui is possible in Okinawan.

I would have expected *migui instead of mingwi by analogy with O mui : J mori.

Original O *ri became i, whereas O *re became a new ri:

re > ri > i

I don't understand why there's an -n- in mingwi. Given that g < *ŋg, could the nasal be an archaism?

But surprisingly, Amami has nïgurï 'to become muddy' with initial n- and ï < *e, implying earlier *ne-! (Cf. Amami mïzï < Proto-Japonic *mendu 'water' from last night's post.)

The root of these 'mud' words in Old Japanese is niŋgə-. The second syllable ŋgə can be reconstructed back into Proto-Japonic, but should the first syllable be reconstructed as

*me- (implied by Okinawan and Nakijin; 4.1.00:02: *mi- would have caused *-g- to palatalize to -j-)

or as

*ne- (implied by Amami)?

Japanese ni- could be from either Proto-Japonic syllable.

11.3.28.23:58: UCHINAAGUCHI PART 8: MEW MANES

From last night: How can the m and n mismatches in these pairs be explained?

O mii- : J nii- 'new'

O nooji : J myouji 'name'

I'm not sure what the explanation for 'new' is. I've long thought that Proto-Japonic *mi sporadically became ni in Japanese. One could also posit a Proto-Japonic *Xi with different reflexes in O and J, but I don't think there are enough cases of O mi : J ni to justify a Proto-Japonic phoneme *X. Reconstructing an *X before *i also implies that *X existed before other vowels, and I have no idea what *X's reflexes would have been before *a, *u, etc.

I briefly wondered if

Proto-Japonic *mi > O mi, Old Japanese and J ni

Proto-Japonic *me > O mi, Old Japanese and J mi

but Old Japanese mi corresponds to Amami mi < Proto-Japonic *mi as well as Amami mï < Proto-Japonic *me (e.g., 'water'):

(3.29.00:29: Could *m- have dissimilated from its fellow labial *-p- in the Japanese word for 'new'? But such an explanation won't account for another case of O m- : J n- that I'll deal with next time.)

The explanation for 'name' seems more straightforward. 'Name' is a borrowing from Chinese 名字 with m-, so Okinawan n- cannot be original. Moreover, there is also an Okinawan myooji 'name'. I cannot find any Sino-Okinawan words with ny, so I think *y was lost after *n:

*my- > *ny- > n-

O myooji may have been borrowed from Japanese after this change was completed.

The second half of this change

*ny- > n-

Next: A muddy mess.

3.29.1:45: I almost forgot to end with a question: We have seen that Okinawan preserves a gw lost in Japanese. Is the gw in the following Okinawan word a retention or an innovation?

O mingwi : J nigori 'muddiness'

cf.

O -gwii : J -goe 'voice'

O mui 'mountain' : J mori 'forest'

(O yama means 'forest' whereas J yama means 'mountain'!)

O migui : J meguri 'circulation'

11.3.27.23:21: UCHINAAGUCHI PART 7: A MII-NII-MYSTERY

From last night: How would these words be reconstructed in Proto-Japonic?

OJ kɨ 'tree' < PJ *kəi

OJ əkɨ- 'to rise' < PJ *əkəi

OJ kinzu 'wound' < PJ *kenzu

What else has Okinawan preserved that Japanese has lost? Find more conservative features in the following data. Two should be easy, one is medium, and one is hard.

O kan : J kan 'sense' (感)

O Kwannun : J Kannon 'Avalokiteśvara' (観音)

O gakumun : J gakumon 'learning' (学問)

O gwansu : J ganso 'ancestor' (元祖)

O yin : J en 'relation' (縁)

O fiijii : J heizei 'usual' (平生)

Feature 2: Okinawan retains gw, but Japanese has merged gw with g.

Feature 3: Okinawan retains y- before what used to be e: O yin < *yen which became J en with y-loss.

The English word yen (originally homophonous with 'relation') was borrowed before Japanese y-loss.

Feature 4: Okinawan retains f- which has weakened to h- in Japanese. 'Usual' also has an Okinawan variant hiijii The sound change f- > h- is more frequent than the reverse. Nonglottals may be reduced to glottal stop and the glottal fricative h which in turn may be lost. I might expect h- > f- before a labial vowel (e.g, Cantonese 虎 *hu > fu 'tiger') but not before a palatal vowel (sequence) like ii or ei.

Now for the mystery in the title. There are lots of cases of O mi corresponding to J mi or me: e.g.,

O mii : J mi 'fruit'

O mii : J me 'eye'

O miji : J mizu 'water'

Similarly, O ni normally corresponds to J ni or ne: e.g.,

O nii : J ni 'load'

O nii : J ne 'price'

O niji : J neji 'screw'

(O has mandatory long vowels in monosyllabic words.)

O mii- : J nii- 'new'

O nooji : J myouji 'name'

(Don't worry about O -oo- instead of expected -uu-.)

Discussion of possible answers tomorrow.