Minangkabau is an Austronesian, Indonesian-type language spoken in West Sumatra by approximately seven million speakers. Despite its large number of speakers and the spread of Minangkabau people throughout the Indonesian Archipelago, Minangkabau remains under-described when compared to other Indonesian-type languages like Javanese. This study seeks to improve current understanding about Minangkabau by describing its system of voice alternations and verb morphology. This study presents a novel analysis of the forms and functions of voice marking in Minangkabau, incorporating naturalistic data into the analysis as well as taking the findings of recent typological and theoretical studies of Austronesian languages into consideration. The study makes use of naturalistic, conversational and narrative data from a database maintained by the Max Planck Institute for Evolutionary Anthropology Field Station in Padang. The study also makes use of elicited data collected in Perth and during fieldwork in Padang. Naturalistic and more formal, elicited Minangkabau data reveals different kinds of linguistic patterns, therefore this study makes a distinction between Colloquial Minangkabau and Standard Minangkabau. The study concludes that Minangkabau has a pragmatically motivated voice system encoded by the alternation between active voice, passive voice and the pasif semu construction. In addition, the study concludes that Minangkabau also has a conceptually motivated voice system that is encoded by a series of semantic and lexical/derivational affixes (ta-, pa-, and ba-) which show how the action originates and develops. The Minangkabau applicatives -an and -i are for the most part valency changing devices but operate within both the pragmatic and conceptual domains of Minangkabau voice. The active voice marker maN- also operates in both pragmatic and conceptual domains whereas the use of the passive voice marker di- is primarily motivated by pragmatic and syntactic factors. This analysis is supported by the finding that di- is a morphosyntactic clitic whereas the conceptual voice markers are affixes and have mainly lexico-semantic properties.