|Title:||Similarity Among Variants: Output-Variant Correspondence|
|Abstract:||This thesis addresses the issues of opacity involved in variant formations, couched within Optimality Theory (Prince and Smolensky 1993; McCarthy and Prince 1993b). In natural languages there are a plethora of cases where one input form can be mapped onto more than one distinct output (e.g., [fÃ¦mIli] and [fÃ¦mli] for 'family'). Concentrating on phonological phenomena involved in such context, this thesis argues for the following points.
(1) To descriptively point out that derivational opacity is often found in the context of variant formations.
(2) To argue for the existence of OV-Correspondence (Output-Variant Correspondence) which militates for the identity between a base form and its variant form.
(3) To critically assess multi-stratal OT model.
(4) To propose a model that generates more than one output form from one input without any reranking or multi-strata.
(5) To show that OV-Correspondence approach has a wider empirical coverage than the approach appealing to local constraint conjunction system in terms of opacity in variant forms.
These points are developed in a following way. The first section is a general introduction, which clarifies my overall goals and provides some basic backgrounds.
Section 2 is devoted to achieve (1) and (2) above. I point out that in the context of variant formations, opacity is often found. Concrete data will be taken up from various languages, including Japanese, Sea Dayak, Mwera and others. Also it will be pointed out that marked syllable structures, which are otherwise absolutely banned from surfacing, are unexpectedly licensed in variant formations. The discussion revolves around the data from English, Japanese and Ganda. One pattern that pertains both of the phenomena above is the emergence of the marked: marked structures only emerge in the context of free variation.
Meanwhile I provide a principled account of the problems above. I argue that in natural languages there is fundamental requirement that outputs that share the same input be phonologically identical. Employing Optimality Theory as the analytical device, these demands are articulated in the form of Correspondence (McCarthy and Prince 1995), which is dubbed here OV (Output-Variant)-Correspondence. This proposal is an extension of Benua's Transderivational Correspondence Theory (1997ab).
In section 3, I argue that OV-faithfulness approach can be independently applied to account for other opaque phenomena: compensatory lengthening and vowel lengthening concomitant with vowel coalescence. This section further supports my OV-Correspondence Theory.
Section 4 is devoted to the discussion of theoretical issues. In the first half of this section, as the third goal of this thesis, I critically assess a version of Optimality Theory that incorporates multi strata (Serial OT; see e.g., Kiparsky 1998). Although this model can be used to account for the problems I present, I argue that the monostratal model that employs OV-Correspondence is better both conceptually and empirically. One of the main arguments is that reranking is too powerful, and allowing this system in OT predicts the existence of languages that are not attested.
Meanwhile I present a new model of how variants might arise without recourse to any reranking or Partial Order Theory (e.g., Anttila 2000a). I show that variant formations often involve anti-faithfulness effects (Alderete 1999), and moreover, variants often exhibit an additional sociological meaning. Based upon these observations, I argue that variant formations are morphologically driven. Variants are generated when they underlyingly have a sociolinguistic morpheme, which is required to be phonologically realized by a morpheme realization constraint (Kurisu 2001). This unifies the OV-Faith approach more closely with general OO-Faith approach (Benua 1997b).
Finally in section 5, I show that although the approach with recourse to local conjunction (Smolensky 1993, 1995, 1997 among others) can account for some of the problems presented in section 2, the OV-Faith approach has wider empirical coverage. Section 6 briefly concludes the paper.