Date: Wed, 26 Mar 1997 14:48:39 +0100
To: " AUSTRONESIAN LANGUAGES AND LINGUISTICS" 
Subject: Re: Bazaar Malay; Mal. nyawa

I at first answered directly to C.C. and J.W., but looks like this is becoming an interesting discussion, so here's my squalk to the cage shake:

25-Mar-97 to C.C.: 
Malay nyawa indeed reflects an apparently original Proto-Austronesian form, which has been reconstructed so far as:

*n'ava	(Dempwolff 1938)
*ñiSawa	(Tsuchida 1975)
*(Ra-)SiNawa	(Zorc 1982)
*(Ri-)ñ@Saua	(Mahdi 1988)

where the apostrophe in n' is a "palatalized articulation" mark, @ is IPa turned-around e (so-called "schwa"), segments in parantheses are optional ("prefix"). Reflexes are attested for languages of Taiwan, the Philippines Indonesia, and Melanesia (and further out in Oceania), i.e., it is not restricted to Indianized languages. Incidentally, in Malay it is not a special feature of Bazaar Malay.

25-Mar-97 to J.W.: 
Here are some more reflexes (from Mahdi 1988):

Thao	Snau	(S= sh) perhaps a loan
Itbayaten	hinawa
Aklanon	ginhawa	(!! compare your Cebuano form)
Pangasinan	lina'wa	(disambiguates initial *[gR] to *R)
Kapampangan	inawa	(ditto)
West Bukidnon Manobo	g@hinawa	(metathesis i/@)
		all "breath, etc."
Wogeo	ma-ñawa	"to rest"

J.W. replied to me: 
Thanks for the message. I wonder if the Bukidnon form isn't the most conservative -- i.e. a proto-Philippine form being *Rehinyawa. I don't know that we could go back to PAN with that because the prefix (?Re-) is only reflected as far south as Baree and no further north than Itbayatan.

This is indeed a very plausible solution. Only the relative position of *@ (your *e) to *i and of *S (> PPh *h) to *ñ (your *ny) seems to be subject to metathesis and one could probably debate endlessly over whether it is

(1)	*(R@)Siñaua (your proposal, cf also Zorc 1982)
(2)	*(Ri)S@ñaua
(3)	*(R@)ñiSaua
(4)	*(Ri)ñ@Saua (my 1988 proposal)

(3) seems objectional because of the sequence *ñi, but Bob Blust quite neatly obviates this by positing *n in place of *ñ, because *ni/_V in tri- or more -syllabic forms (also bisyllabic forms of a gramaticalized nature, e.g. pronouns) would predictably result in ny in those languages in which we find it in reflexes of the protoform. The sole counterevidence is the Kanakanavu reflex, which would require an *ñ (Tsuchida). Your objection to Tsuchida's alignment is well placed, because it is based on very few examples, and there is at least one counter-example. Nevertheless, I prefer to be a bit cautious here. *ñ is a relatively rarely occurring proto-phoneme, so one would not expect many examples. These indeed happen to be all the data we have on *[ñ] in Kanakanavu, and all adjacent languages in the wider neighbourhood have united *ñ with *N or *n, so the counterevidence in Kanakanavu could be a borrowing.

Thao suggests (2), but in view of the analogical irregularity (loss of final vowel) it should pobably be connected with the Tsou reflex through metathesis n/s. In the Philippines, Cebuano and Aklanon point to (4), Pangasinan and Kapampangan are impartial to (4) or (2), while WB Manobo and Itbayaten are conform with (1).

So my conclusion at this point is, that we are undecided between (1) *(R@)Siñaua, (3a) *(R@)niSaua, and (4) *(Ri)ñ@Saua, where (3a) depends on the rejection of Tsuchida's *ñ > Kanakanavu ny assignment. If the optional *(R[]) had indeed been a later innovation, not shared by languages of Taiwan, then Kanakanavu points to (1), while Tsou and Thao (which are probably not independent) suggest (4). Not very much, because Kanakanavu ng raises questions, while the Tsou and Thao reflexes have an undisputable irregularity.