|Title:||Mostly Predictable: Cyclicity and the Distribution of Schwa in Itelmen|
|Authors:||Jonathan David Bobaljik|
|Comment:||16pp (8 sheets, postscript 2-up layout)|
|Abstract:||The Itelmen language shows many instances of regular schwa/zero
alternations. Even though the language permits quite extensive
consonant clusters, I argue in this paper that the alternating
schwas (and perhaps all instances of schwa) are epenthetic;
schwa is inserted to break up a disfavoured consonant cluster.
The rule which inserts schwa must apply cyclically in the
verbal system, but non-cyclically in the nominal system.
Apparent examples of cyclic rule application are problematic
for non-derivational, constraint-based approaches to phonology,
such as many versions of Optimality Theory (OT). Thus, much
recent work has been devoted to reanlayzing purported examples
of cyclicity from an O.T. perspective. While it may be possible
to devise an account of the Itelmen data in terms of parallel
constraint evaluation, current O.T. approaches are insufficient;
in particular, the best candidate for an explanation of the
Noun/Verb differences (Base Identity) makes exactly the wrong
predictions for Itelmen.