Corpus-based approach meets LFG: the puzzling case of voice alternations of kena-verbs in Indonesian

This paper discusses the meaning-preserving hypothesis of voice alternation in Indonesian from an LFG perspective. The hypothesis predicts that the meaning encoded by a transitive verb is available for both active and passive forms, differing only in the alignment of grammatical relations and semantic roles. Using quantitative corpus linguistic analysis, we argue that voice alternation needs to be relativised to (i) a certain sense of a verb and (ii) (statistical) usage constraints of the verb�s semantics in certain voices. We also demonstrate the viability of the LFG framework and related analytical issues in capturing such empirical facts.