The position of Mantauran in the Rukai family remains controversial because the structure of this dialect has been obscured by drastic phonological and syntactic changes. Phonologically, it has undergone a process of spirantization, PR * ƒÒ > Mt ƒæ; PR * ê and ƒÔ > Mt ƒ´; PR * ƒ× > Mt ƒØ. Syntactically, it differs from the other Rukai dialects and the Formosan languages as a whole: (1) it lacks the nominal case marking commonly found in most of these languages, (2) it has bound pronouns which tend to coalesce with the verb stem, (3) the third person pronoun has been reanalyzed as a "non-agent" agreement marker.

    The present paper investigates the nature of grammatical relations in Mantauran and analyzes the syntactic and semantic properties of subjects and objects in this language. I try to account for the development of the "verb-object" agreement in this dialect from a historical perspective. I show that in order to understand the synchronic variations mentioned above, we should take into account the development of the language and argue that (1) the third-person pronoun has taken on an important role in the coding of grammatical relations as a result of grammaticalization and reanalysis, (2) word order -- largely determined by pragmatic factors -- can be accounted for if we take into consideration the syntactic changes undergone, i.e., Mantauran is turning into a head-marking language, (3) grammatical relations are semantically-based. The pronominal agreement of Mantauran is further illustrated by a text given in the appendix.

Keywords: Formosan, Rukai, Mantauran, pronominal agreement