Towards an information structure analysis of ergative patterning in the Inuit language

X.1 The introduction The Inuit language is often characterized as an ergative language (Bok-Bennema 1991; Manning 1996; Johns 1992, 2000, 2006, among others). Interestingly, the Inuit language exhibits a case assignment variability which, unlike traditional split ergativity, does not affect argument alignment, but instead concerns what – and how many – arguments trigger φ-feature agreement on finite verb. This paper investigates the question of what exact grammatical property triggers this variability and the corresponding agreement properties. The case variability has not gone unnoticed in the current literature. Some proposals attribute the split to aspect, for instance, Clarke (2009); Spreng (2006, 2010, 2012), while others to information structure, scope, or incorporation (Kalmár 1979; Bittner 1987, 1994; Bok-Bennema 1991; Manga 1996; Bittner and Hale 1996; Hallman 2008; Berge 2011, among others). All these proposals point out some important property of the observed case variability. Yet, it is not always clear how the proposals derive the morphosyntactic properties of the Inuit language, neither do they provide much insight into how these seemingly disparate properties might relate to each other. We argue that the ergative clause structure of the Inuit language is conditioned by information structure properties, more precisely by its topic-comment properties. We will provide evidence that the topic dimension is the primary trigger of the morphosyntactic make up. Furthermore we will show that other relevant properties discussed in the literature, i.e., scope properties of objects and aspect, are derivative of this information structure characteristics of the ergative split. In order to execute the argument, we will first provide evidence that the difference between so called singular and double agreement is a difference of φ-agree vs cliticization (Johns to appear contra Compton 2014). We will tie this conclusion to our second core observation, namely, that absolute objects must be (aboutness) topics in the sense of Reinhart (1981), following Berge (2011). Since topics are associated with clitic doubling (Dočekal and Kallulli 2012), these two observations provide an explanation of the clitic nature of the double agreement. Furthermore,