|Abstract:||This thesis examines the status of onsets and their effects on stress and prosody. I argue that moraic onsets exist, a claim that contradicts standard phonological models (Hyman 1985, Hayes 1989, Gordon 1999, Morï¿½n 1999) which assume that onsets are not moraic, given that in the overwhelming majority of languages onsets are inert for prosodic processes.
Using data from Pirahï¿½, Karo and Arabela stress, I show that weightful onsets actively participate in weight-sensitive stress assignment. Moreover, I point out that if onsets can be moraic, a host of other weight-based phenomena, should also be able to utilize them. This is exactly right, as verified by word minimality in Bella Coola, Samothraki Greek compensatory lengthening (CL), onset geminates in Pattani Malay, Trukese and Marshallese and a variety of other data, e.g. Trique CL, Bellonese reduplication. Crucially, this is not a prediction shared by previous prominence-based analyses of similar facts (Hayes 1995, Gordon 2005, Smith 2005). Prominence is inherently designed to account only for stress, not for other weight-based phenomena. If one were to entertain a prominence account, then most of the data above would remain unexplained.
However, not all onsets can be moraic. The proposed model is restrictive in admitting only two kinds of moraic onsets: those which are underlying, i.e. emerging as geminates, and those which are derived in the output and serve for stress purposes. While the former can be of any featural content (since they are lexically specified and thus unpredictable), for the latter ones, I claim that only voiceless onsets can be moraic, whereas voiced ones are never moraic. This relates to a well-known generalisation affecting a different prosodic phenomenon, namely tone. Voiceless onsets raise the pitch of the following vowel, voiced ones lower it. In many languages such pitch perturbation is interpreted as tone. My proposal is that in some other languages, this pitch perturbation is instead interpreted as stress and is formally represented by means of moras, which are only assigned to the stress-attracting voiceless onsets. Pirahï¿½, Karo and Arabela data empirically confirm this finding.
|Article:||This article has been withdrawn.|