This thesis investigates the motivational factors behind the choice of voice in the Oceanic language Äiwoo. This is a language with a so-called symmetrical voice system, which is characterized by the alternation of two or more clause types. Unlike asymmetrical voice systems, like the active-passive system found in many European languages, the different clause types in a symmetrical voice system are equally marked for voice morphosyntactically and are typically considered equally transitive. While this type of system is found in a number of western Austronesian languages, it is generally considered lost in Proto Oceanic, and most modern Oceanic languages have transitivity-based systems rather than symmetrical voice. Äiwoo, being an Oceanic language, is thus unique among this group of languages in having a symmetrical voice system. The clause system of Äiwoo consists of three voices: actor voice, undergoer voice and circumstantial voice. The focus of this thesis is the motivation behind the choice of the two main voices, namely actor voice and undergoer voice. In order to investigate this, I have employed three different explanatory models. These are topic continuity developed by Givón (1983, 1994), discourse transitivity developed by Hopper & Thompson (1980) and the three different activation states as discussed by Chafe (1987). I have coded 160 Äiwoo clauses according to the factors described in these models. Through significance testing, I found that aspect and the activation state of undergoers are significant predictors for voice, so that perfective aspect predicts the choice of undergoer voice and new undergoers predict the choice of actor voice. Based on these results, I argue that undergoer voice corresponds to a higher degree of semantic transitivity than actor voice, a tendency found in other Austronesian languages with symmetrical voice systems. The implications of this give rise to a discussion on the differences between a symmetrical voice system and a transitivity-based system, where I argue that Äiwoo is in the process of developing a transitivity-based system where undergoer voice is reanalysed as the standard transitive clause type.