Argument Alternation and Argument Structure in Symmetrical Voice Languages: A Case Study of Transfer Verbs in Amis, Puyuma, and Seediq

Abstract

This dissertation investigates argument structure/alternation in symmetrical voice languages based on the study of transfer verbs in three Formosan languages: Amis, Puyuma, and Seediq. The morphosyntax of transfer verbs is carefully examined according to the three-way classification of transfer verbs. With respect to morphological composition, all three languages exhibit a distinction between give/send-type verbs and throw-type verbs. The finding is consistent with the semantic basis of the classification: give/send-type verbs lexicalize caused possession/motion, while throw-type verbs are two-argument verbs with no involvement of causative semantics. The derivational status of Philippinet-type voice marking is established upon scrutiny of the argument structure of Formosan transfer verbs. Most of the transfer verbs undergo argument alternation between the recipient/goal and the transported theme by means of locative/circumstantial voice (LV/CV) marking. Some "transfer verbs," however, do not always allow argument alternation, as a particular voice form of these verbs may involve a thematic role (location/instrument/beneficiary/patient) other than those in a transfer event. Lexical variation in argument alternation restriction is found within "subclasses" of transfer verbs—an observation not predicted by the ditransitivity hierarchy. To account for the absence/presence of transfer interpretation denoted by different voice "forms" of the same root, I argue that symmetrical voice marking interacts with roots and provides a "constructional" meaning. By means of the semantic map, I show that Formosan LV marker is responsible for designating a set of conceptually contiguous thematic roles (goal/recipient/location/patient), whereas the CV marker targets a different set (theme/instrument/beneficiary/stimulus). Finally, I examine whether current generative theories can account for the argument structure of voice-coded verbs in Formosan languages. I first point out the empirical problems of the applicative analyses of Formosan LV/CV verbs. Formal applicative analyses typically assume a pre-existing subcategorization frame of the verb/root. This assumption, however, does not hold in symmetrical voice languages, where roots prove to be category-less (and therefore argument-less). Embracing the exo-skeletal approach, I propose a feature-based analysis: LV and CV verbs contain distinct functional projections (FPs), specified with the event feature [ground] and [cause], respectively. The FP verbalizes the root and introduces the internal argument as a result of feature valuation.