Reduplication as Affixation in Paiwan

In this thesis I investigate the morphological structure of reduplication in Northern Paiwan, a Formosan language of Southern Taiwan. The main objective is to use Optimality Theory (OT), as first formulated by Prince and Smolensky (1993), to show that reduplication can be considered the same as affixation. This is in support of Marantz (1982).I do this by first classifying reduplication into two classes: prefixing (Ca reduplication or CaRED) and suffixing (root reduplication or RtRED). Factors determining these classes are phonological structure and semantics. Phonologically, RtRED is dimoraic and copies all of its segments from the stem. CaRED, on the other hand, is monomoraic with the vowel invariably surfacing as a. I find that the more prototypical and commonly used reduplication, RtRED, also has the more prototypical semantic functions of reduplication. CaRED, on the contrary, has more specialized meanings with a narrower distribution. Next, I show that reduplication and affixation follow the same constraints. Thus, suffixal reduplicants follow the same constraints as suffixes and prefixal reduplicants follow the same constraints as prefixes. In addition, suffixes and prefixes can also follow constraints previously restricted to the domain of reduplication. One of the conclusions that come forth from the analysis is that templates are needed to describe Paiwan root reduplication. This is in opposition to recent efforts by McCarthy (1997), Gafos (1998) and Spaelti (1999) to eliminate template constraints. My analysis also needs to consider the word-final coda to be extraprosodic. To do so, following Harris and Gussmann (1998), word final codas are reanalyzed as onsets. In this way, word-final codas will not violate No-Coda. In addition, I look at Max-BR and how it has no important role in analyses that require templates. I also choose to work in Spaelti's (1999) lexical-surface (LS) framework, as opposed to McCarthy and Prince's (1995a) input-base-reduplicant (IBR) framework. I motivate this by showing that the LS interface will solve all of the problems created by the IBR interface. With this analysis arises one complete OT grammar which can be used to analyze both reduplicants and affixes. In addition, included in the Appendix is a successful application of this grammar to Thao, another Formosan language. This analysis shows how the grammar can handle word-internal codas and consonant clusters, both of which are abundant in Thao and missing in Paiwan.M.A