- Issues in Austronesian Historical Phonology
This book contains ten articles originally presented at the Ninth International Conference on Austronesian Linguistics and the Fifth International Conference on Oceanic Linguistics held at Australian National University in Canberra during January 2002. It has as its theme Austronesian historical phonology. There is an overview of the PAN phoneme system (Wolff), a Sprachbund involving final /a/ mutation (Tadmor), a mixed tree- and wave-theoretical account (McGinn), sound changes through its macro-groups (McGinn, Bukar-Sadong Land Dayak and Rejang; Mead, Celebic; van den Berg, Tukang Besi and Muna-Buton) and micro-groups (Mead, Saluan-Banggai), unusual sound changes (Schmidt, Rotuman; Lynch, Proto-Loyalties), and reflexes in specific languages (Blust, Selau; Wolff, Fijian).
John Wolff forms the alpha and omega of this work, presenting the first article on The sounds of Proto-Austronesian, and the tenth and final one on Fijian reflexes of PAN phonemes (see below). His is a continuing and cogent review of PAN reconstructions,1 and a legitimate attempt to limit the PAN phoneme inventory to that of a real language. The consonant inventory of Austronesian languages rarely exceeds the 23 found in Paiwan.2 Malay, Indonesian, and Fijian have 19;3 most Philippine languages have from 19 (Western Bukidnon Manobo) to 14 (Kayapa Kallahan [Reid 1971]); and Hawaiian boasts the second smallest inventory in the world with only 8.4 Why then would or should the reconstructed system have 27 or more?
Although the purpose of his paper is to deduce "what the PAN phonemes sounded like" (5), he first discusses the system and the reflexes. The first chart presents Wolff's revised inventory of 18 PAN consonants. I have combined it with his chart 1a to show the typographic symbols traditionally used, with which readers may be more familiar:
|VOICED CONSONANTS||*b = *b||*d = *D||*j = *Z||*g = *g- -j- -j||*Ɣ = *R||WOLFF TRADITIONAL|
|VOICELESS STOPS||*p = *p||*t = *C, *t||*c = *s||*k = *k||*q = *q||WOLFF TRADITIONAL|
|NASALS||*m = *m||*n = *n||*ñ = *ñ, *N, *L||*ŋ = *ŋ||WOLFF TRADITIONAL|
|LIQUIDS||*w = *w||*l = *l||*y = *y||WOLFF TRADITIONAL|
|SIBILANT||s = *S||WOLFF TRADITIONAL|
[End Page 505]
Also included is phonemic or contrastive stress, which may well account for the split of *t into *t and *C, and of *ñ into *N and *L. Gone are: PAN *c, *d, *g (noninitial), *L, *N, *r, *T, *z. Unmentioned but gone by default are: *ʔ, *H, *W, *X.
This goal and its economy are admirable. In doing so, he makes two default assumptions:
1. "Phonemic contrasts [that] cannot be documented did not exist" (3);
2. "Although the protolanguage surely had variation … , unless there is evidence to the contrary, only one of two or more variant forms in the protolanguage can be assumed to have come down to modern times."
The latter is an evidentiary requirement under which doublets have to be held suspect. He accepts *tiduƔ and *tuduƔ 'sleep,' because there can be no phonological, geographic, subgrouping, or contact explanations that would explain them away.
He also makes two more assumptions about the processes of change:
3 Changes "have to be phonologically motivated" (3); and
4 "Sound change proceeds on a word-by-word basis, and is not completed until all forms with the phoneme in a given environment have been replaced by the innovation" (4), that is, LEXICAL DIFFUSION.
He exemplifies the latter with the merger of PAN *d and *g (*j) into Javanese /r/ as well as /dh/. Borrowing does not provide a satisfactory hypothesis for the split, so the conclusion must be that the change was incomplete (5). Here doublets (such as Old Javanese dhengö and rengö) are allowed. We can also add that the same would apply to the merger of PAN *Ɣ (*R) and *g (*g-, *-j-, *-j), which is found split into Ilokano /r/ and /g/.
Wolff usually accounts for semantic discrepancies by offering alternative glosses. However, in chart ii under "heavy" he cites Paiwan v'qatj, which either is the root...