English stress preservation and Stratal Optimality Theory

Since Chomsky & Halle (1968), English stress preservation has been important in generative discussions of morphophonological interaction. This thesis carries out empirical investigations into English stress preservation, and uses their results to argue for a particular version of Optimality Theory: Stratal Optimality Theory ('Stratal OT') (Kiparsky, 1998a, 2000, 2003a; Bermudez-Otero, 1999, 2003, in preparation). In particular, the version of Stratal OT proposed in Bermudez-Otero (in preparation) and Bermudez-Otero & McMahon (2006) is supported.The empirical investigations focus upon the type of preservation where preserved stress is subordinated in the preserving word ('weak preservation'). Evidence for the existence of weak preservation is presented. However, it is also shown that weak preservation is not consistently successful, but that it is, rather, probabilistically dependent upon word frequency. This result is expected in light of work like Hay (2003), where it is proposed that word frequency affects the strength of relationships between words: stress preservation is an indicator of such a relationship.Stratal OT can handle the existence of English stress preservation: by incorporating the cyclic interaction between morphological and phonological modules proposed in Lexical Phonology and Morphology ('LPM'), Stratal OT has the intrinsic serialism which is necessary to predict a phenomenon like English stress preservation. It is shown that the same cannot be said for those of models of OT which attempt to handle preservation while avoiding such serialism, notably, Benua (1997).Bermudez-Otero's (in preparation) proposal of 'fake cyclicity' for the first stratum in Stratal OT can capture weak preservation's probabilistic dependence upon word frequency. Fake cyclicity rejects the cycle which has previously been proposed to handle weak stress preservation, in LPM and elsewhere; instead, fake cyclicity proposes that weak preservation is a result of blocking among stored lexical entries. Blocking is independently established as a psycholinguistic phenomenon that is probabilistically dependent upon word frequency; in contrast, the cycle is not a probabilistic mechanism, and so can only handle instances of stress preservation failure by stipulation.Ph.D