Abstract
The non‐Pama Nyungan languages of the north of Australia are notable for the often elaborate pronominal prefixing that is found on verbs. Most descriptions of this prefixing describe it as following a nominative‐accusative pattern, and note that the ordering of the affixes is not fixed: sometimes object precedes subject, sometimes subject precedes object. In this article I demonstrate that the pronominal prefixing found in Maung is broadly a split‐ergative system, and that through reference to the alignment categories distinguished by the prefixes, not the syntactic roles (A, S or O), we can find a set of principles behind the variations in prefix order, which are modelled in a set of constraints following the principles of Optimality Theory. Some speculations on the historical origins of this system are presented, suggesting that this aberrant system arose from the reanalysis of a voice marker, and the development of the accusative prefixes from the nominative forms. The Maung data is used to argue that Bittner and Hole's proposals regarding agreement ordering do not represent a set of universal conditions.
Notes
I would like to thank Alan Dench for his detailed and insightful comments on an earlier version of this paper, which substantially improved both the analysis and presentation, Charles Reiss for his perceptive comments which have improved the clarity of the presentation, and Cathryn Donohue for her technical support of this work.