The purpose of this project is to survey the distribution and behaviour of a phenomenon which has hitherto received very little attention, but which is central in the grammar of several Austronesian languages in Taiwan (i.e. the Formosan languages): the realization of various types of adverbial as matrix verbs, with the main verb realized as a dependent (as if we were to say "He oftens to work"). Exactly how this is to be analysed can have important repercussions on our understanding of language structure, especially taking into account another interesting phenomenon found in the said languages, namely that certain noun phrase modifiers (such as quantifiers) are realized as primary predicates (i.e. "[The students I have seen] are many" instead of "I have seen many students"). If these two phenomena can be shown to be related, it may be evidence of a hitherto undescribed and theoretically unexpected way of constructing clauses, the relation subject-predicate (considered universal in some grammatical models) being replaced in some languages by its logical opposite (core predicate - situation), a difference which would cut much deeper than word order variation. To test this and other possible analyses (e.g. the mapping between meaning and word class), more detailed information about both constructions is required. The purpose of this project is to collect such information on two Formosan languages, Seediq and one other, to be chosen on location.