11.5.3.2:30: UCHINAAGUCHI PART 39: FROM EELS TO ʔIYU-S

I still don't know how to reconstruct the Proto-Japonic word for 'eel'. I just needed a word that sounded like Shuri/Naha Okinawan ʔiyu 'fish' for the title of this post. I pronounce eel as [ʔiɰ] which is close to ʔiyu.

{i, e}y{u, o}

but in fact it's uo from Old Japanese uwo. Only the last segment is a regular match. The correspondences

O i : J u

O y : J w

are unusual. If I were a beginning historical linguist, I'd jump up and say, "I know the answer! The Proto-Japonic word was *üɥo!" I'd be using a compromise strategy:

*ü is a vowel like both O i and J u

*ɥ is a glide like both O y and J w

And I would almost certainly be wrong. I shouldn't reconstruct segments on the basis of rare correspondences. Moreover, I should avoid reconstructing rare segments. Only 5% of languages in UPSID have ü and only 1% have ɥ. This doesn't mean no languages should be reconstructed with such sounds. French is a well-known language with ü and ɥ (obviously not spelled that way in French). But it's highly unlikely that a language would have rare sounds in only one word. If there were lots of O i : J u correspondences, I'd gladly reconstruct Proto-Japonic *ü as their source. But that's not the case.

Maybe looking at more data might help. Thorpe (1983: 286) lists various Ryukyuan cognates of O ʔiyu : J uo: e.g.,

Naze ʔyu

Koniya and Nakijin ʔyuu (latter from Curry 2004: 269)

Hentona yuu

Ikema ïu

Oku iru

Sarahama ïzu

Ishigaki izu

Taketomi izyu

Kobama icu

Hateruma iyu

Hentona and Oku are spoken on Okinawa Island along with Shuri/Naha and Nakijin.

Thorpe reconstructed Proto-Ryukyuan *iyU 'fish'. The final vowel was probably *o given Old Japanese uwo, but we can't be certain about that. After all, the first vowels don't match.

Just to complicate matters, there is a Middle Japanese iwo that looks like a blend of the Ryukyuan and Old Japanese forms. Even in modern Japanese, there are forms with i-.Thorpe mentioned Izu Ooshima iwo (to the south of Tokyo, far from the Ryukyus) and Kagoshima io (in southern Kyushu, north of the Ryukyus). So i- cannot be a Ryukyuan innovation. Could the Kagoshima form go back to *iyo? Was uwo a central Old Japanese innovation contrasting with older i-forms surviving on the periphery? I'd need a lot more dialect data before I could even begin to answer that question.

11.5.2.1:41: UCHINAAGUCHI PART 38: 'SEA'-ING EELS

A reader pointed out that Japanese unagi 'eel' contains una- which looks like the combining form of 'sea' seen in parts 36 and 37. Is there any connection between unagi and, say, unaji 'sea road'? The latter is a transparent compound of una- 'sea' + genitive + chi 'road'. (The voicing of the initial consonant of the second element indicates possession: 'road of the sea'.) But can unagi be similarly cut up into una 'sea' + genitive + ki? I can't think of any ki that has anything to do with eels. If you encounter a word AB that looks like A but has a B without any obvious meaning, then AB may be an indivisible root.

In part 36, I derived una- from *omi-na-. In Shuri/Naha Okinawan, *o- can become a nasal before *m:

*o-ma 'that-space' > ʔnma 'there'

*omi- > ʔnmi- ~ ʔumi- 'sea'

However, I don't know if *o can also become a nasal before *n. Could S/NO ʔnnaji 'eel' be from *onagi rather than *unagi? I do know for sure that *o doesn't have to reduce to a nasal before *n:: e.g.,

*oni > ʔuni 'demon'

In Nakijin, 'there' and 'demon' are

ʔumaa ~ ʔmaa

ʔunii

and 'eel' is

ʔunaazi

with ʔu-.like the *o-words 'there' and 'demon' rather than just glottal stop like in ʔmaa < *uma 'horse' which has no ʔu-variant in Nakijin. Perhaps (N = either m or n):

According to the above scheme, Nakijin ʔunaazi (which has no variant *ʔnaazi) would be from Pre-Nakijin *onagi.

(See Curry 2004: 228-231 for further discussion of Nakijin initial ʔN-.)

But can *o- be projected back to Proto-Japonic? To complicate matters further, the earliest attestation of 'eel' is munaŋgi in Man'youshuu. There are other Old and Early Middle Japanese mu- corresponding to later u-. Here's what I think might be happening. Emphasis on "think".

1. Thomas Pellard (2008:146) established a method to distinguish between Proto-Ryukyuan *um- and *om- corresponding to J u-.

2. I project this distinction back into Proto-Japonic: e.g., *uma 'horse', *omi 'sea'.

3. First wave of nasalization: In pre-Old Japanese, *uN- may have become mN-, written as muN-: e.g., *uma > mma (written 牟麻 muma) as well as uma 'horse'.

4. *o- merged with *u- in Old Japanese: e.g., *omi > OJ umi 'sea'. (Modern Japanese initial o- is from Proto-Japonic *ə- or *wə- or *wo-, not *o- which became u-.)

However, at this point, no (?) secondary *u- from *o- became a nasal before another nasal.

5. Second wave of nasalization: In Early Middle Japanese, even secondary u- from *o- could become a nasal before another nasal:

*omai > OJ *uməy > EMJ mme (written mume) 'plum'

*um ... - > OJ umare- > EMJ mmare- (written mumare) 'be born'

(But as Vovin pointed out, the base verb um- was never mum- in EMJ!)

(For a very different view on 'horse' and 'plum', see section 3.3 of Vovin's Old Japanese grammar.)

According to the above scheme, Old Japanese munaŋgi 'eel' is a product of the first wave of nasalization, so it must go back to Proto-Japonic *unaŋgi and cannot be related to *omi or *ona- < *omi-na- 'sea'. But this contradicts my reconstruction of *onagi based on Nakijin. So I'm still uncertain about what the first vowel of 'eel' was.

At least I know a couple of things for sure:

There is an S/HO word ʔunagi 'that length' that has nothing to do with the sea. It's from ʔu- < *o- 'that' (no J cognate; corresponds in meaning to J so- < *sə-) and nagi 'length' (cf. J naga- 'long').

J unaji and S/HO ʔunaji 'nape' also have nothing to do with the sea. Their una- was once an independent word meaning 'nape' and is also in J unazuk- 'to nod'. Iwanami kogo jiten derives unaji from una + shi 'rear', but has no entry for shi 'rear', though it has a shi 'bottom'.

Nakijin has hazii 'nape' which doesn't have any obvious Japanese cognate. I would expect it to correspond to a nonexistent *aji or *kaji.