In this talk, I use traditional constituency diagnostics such as coordination and adverb placement to investigate the status of the verb phase in Austronesian. Conservative Malayo-Polynesian languages are shown to treat the verb and Agent of a Non-Actor Voice clause as a constituent which excludes the Patient (in agreement with earlier analyses of languages such as Tagalog and Malagasy). This stands in strong contrast to (both accusative and ergative) languages of mainland SE Asia, most of which show relatively good evidence for a traditional Verb Phrase that includes the verb and Patient while excluding the Agent.
Expanding on earlier work, I argue that the unusual grouping of transitive Agent with the verb is accounted for in Austronesian by the historical reanalysis of possessors as agents. Possessors naturally form a constituent with the noun phrases they modify and thus the Possessor/Agent forms a constituent with the verb after the N>V reanalysis. This leads us to an interesting prediction regarding Puyuma and Tsou under Ross’s (2009) Nuclear Austronesian hypothesis. If only the realis paradigm consisting of Patient Voice *-en, Locative Voice *-an, and Circumstantial Voice *Si- derives historically from the reanalysis of nominalizations, we do not expect the unusual constituency of Non-Actor Voice clauses in Malayo-Polynesian languages to show up in the irrealis paradigm, which never passed through this nominal stage.
Evidence from Puyuma and Tsou points towards the existence of just such an asymmetry. Teng (2008) strongly suggests an unmarked word order in Puyuma Non-Actor Voice clauses that is highly unusual for Formosan and Philippine languages: the pivot (or NOM argument) regularly precedes the transitive agent. The unmarked status of this order was supported by experimental elicitation in recent fieldwork. Crucially, however, the unmarked Pred GenP NomP order of NAV clauses in Nuclear Austronesian languages also emerged in Puyuma in clauses with nominal predicates. This can be interpreted to support the Nuclear An Hypothesis as it represents an additional syntactic parallelism between nominal predicates in extra-Nuclear An languages and the canonical declaratives of Nuclear An languages. Beyond these particular arguments specific to Austronesian, I aim to address the wider relevance of word order and clausal constituency to solve historical questions.