536 
Berichte und Kommentare 
Anthropos 85.1990 
verb, and there is some evidence that these changes 
were already under way at Proto-Eastem-Oceanic, 
or even Proto-Oceanic level (Geraghty 1983: 267- 
269). Nevertheless, roughly half of Fijian verbs 
reflecting a PAN etymon retain faithfully the final 
consonant - Arms (1973: 543) reckons 45 %, Mil 
ner (1989: 62) 43 %. So, with a knowledge of the 
semantic realignments and phonotactic constraints 
involved, it is often possible to use Fijian, and 
probably other Eastern Oceanic, data to support the 
reconstruction of PAN final consonants. Curiously, 
Blust does this with Samoan, but not with Fijian or 
any other Oceanic language. For the record, Fijian 
and other Oceanic reflexes (some not mentioned by 
Blust) support the final consonant of the following 
roots: 
*Dem “think”: Arosi ’ado-mai 
*kep “seize, grasp, embrace”: Fijian rako-v “embrace” 
*nut “husk, fibre”: Bugotu penutu “coconut husk” 
*pas “tear or rip off’ (sic, but the more widespread meaning 
is “loosen, remove, untie”); Roviana rupa-h “loosen,” Arosi 
ruha-s “loose, let go, untie” 
*Tuk “knock, pound, beat”: Fijian vutu-k “pound,” natu-k 
“knead, pummel” 
A systematic search would doubtless produce 
more corroboratory evidence. For about ten re 
constructions for which Blust provides Oceanic 
reflexes, the final consonants are not supported by 
the evidence. I will not discuss these further, as 
it would involve detailed discussion of the Fijian 
data, not to mention considerably more research 
into languages other than Fijian, to determine 
whether they have also been subject to semantic 
realignment of final consonants and phonotactic 
restrictions - but I believe such research will even 
tually result in as secure a reconstruction of final 
consonants for Proto-Oceanic and Proto-Eastem- 
Oceanic as for any other subgroup of Austrone- 
sian. 
The final consonants of Fijian transitive verbs 
merit further attention here, because they con 
stitute a set of subsyllabic submorphemes in an 
Austronesian language that have been recognized 
as such since 1850 (see Milner 1989: 59-67 for 
a summary of the literature). Essentially, many 
verbs show a non-etymological final consonant, 
usually referred to as the “thematic consonant,” 
which is determined by the semantic class to which 
the verb belongs, so long as it is not homorganic 
with either of the consonants of the base. The 
thematic consonant only surfaces when the verb 
is transitive, and is always followed by a vowel. 
For example, verbs of motion take v: lako-v “go- 
for,” cici-v “run-for,” lade-v “jump-over,” siro-v 
“descend”; but the opposite of “descend,” cabe-t 
“ascend,” may not take v because there is a bilabial 
consonant in the base. Verbs of physical func 
tion take c [d]\ rai-c “see,” rogo-c “hear,” boi-c 
“smell,” buno-c “sweat-on,” regu-c “kiss,” lua-c 
“vomit-on”; but cegu-v “breathe” may not take c 
because the base contains c. Verbs of breaking take 
k: musu-k “break,” vow-k “smash,” diri-k “crack 
open,” dresu-k “rip, shred,” basu-k “tear open”; 
but gutu-v “snap, cut (vine, etc.)” may not take k, 
since the base contains a velar consonant. I do not 
propose to offer an analysis of this phenomenon, 
but I trust it is sufficiently clear that it represents 
another problem for the traditional definition of 
the morpheme, as has indeed been recognized by 
one who helped develop that traditional definition: 
“The behaviour of the Fijian thematic consonants 
was one of the real facts about languages that led 
me slowly but surely to abandon what I now refer 
to as the ‘atomic morpheme theory,’ the theory 
of grammatico-lexical structure I helped develop 
in the 1940’s and to which I clung for a long 
time. That theory proposes that every phonemic- 
ally relevant piece in any utterance must be a part 
of one or another morpheme (or of the phonemic 
representation of one or another morpheme), and 
that morphemes are minimal meaningful elements 
in much the same sense in which we all assumed 
phonemes were minimum meaningless but differ 
entiating elements” (Charles Hockett p. c., quoted 
in Milner 1989: 77). Blust’s monograph may not 
replace Flockett’s doubts with certainty, but it is 
a thoughtful and provocative discussion of where 
the limits of morphology lie, based on an extensive 
collection of data which no one but Blust could 
have mustered so skilfully. 
The book is both easy in the hand and a treat 
to the eye, well bound and designed, and with 
only a handful of innocuous typographical errors, 
though I fear many readers will be looking in vain 
for maps to locate the myriad languages. Blust is to 
be congratulated, once again, on a very important 
contribution to Austronesian studies, and to lin 
guistics generally. 
References Cited 
Arms, David G. 
1973 Whence the Fijian Transitive Endings? Oceanic Lin 
guistics 12: 503-558. 
Blust, Robert 
1988 Austronesian Root Theory. An Essay on the Limits of 
Morphology. Amsterdam: John Benjamins Publishing 
Company. (Studies in Language Companion Series, 19)