Abstract

This thesis presents a model of inverse-marking in the Nishnaabemwin (Algonquian) agreement system. An abstract 'person hierarchy' ranks discourse participants as 2nd person > 1st person > 3rd person > inanimate, which seems to determine the form of a suffix on the verb stem often called a 'theme sign' (Valentine 2001). The highest ranked discourse participant is marked by a prefix on the verb. A so-called 'direct' theme sign appears if the subject outranks the object, while a so-called 'inverse' theme sign appears if the object outranks the subject. However, these descriptions imply that agreement mechanisms in the verbal domain obligatorily consult an abstract hierarchy in order to encode a concrete relationship between the subject and object. This is problematic as it requires the controversial property that abstract hierarchies are directly encoded in the grammar.

A growing body of research analyzes languages with person hierarchy effects as complex systems of person agreement (e.g. Béjar and Rezac 2009, Lochbihler 2012, Oxford 2014, Preminger 2014, among others). The model proposed in this paper is an application of Preminger (2014), an analysis originally meant to capture person hierarchy effects in Kichean, a Mayan language. Although Preminger's analysis elegantly accounts for the person hierarchy effects in Kichean, it cannot straightforwardly be extended to Nishnaabemwin inverse-marking.---namely, Preminger's model fails in contexts with two arguments that are speech act participants (i.e. either the speaker or addressee). To fill these gaps, this thesis builds on Oxford (2014) in analyzing Nishnaabemwin inverse-marking patterns as a complex form of object agreement. The proposed model is an agreement system in which two probes work together to license arguments. Though Nishnaabemwin inverse-marking reflects a complex form of object agreement in most cases, the morphosyntactic consequences of failed agreement show that, in some contexts, object agreement is overridden as a result of failed agreement.