Morphology matters: Case licensing in Basque

The aim of this dissertation is to provide a unified account of a wide range of morphosyntactic phenomena within a single case licensing theory. Few case theories have investigated phenomena that fit beyond the canonical mould. For example, quirky case has been well studied, but has nearly always been analyzed as a lexical stipulation. Similarly, argument-adding operations, such as causatives, can potentially create four-place predicates which never occur in underived verbs. I explore the intersection of these complex areas by investigating quirky case and causatives in Basque, an ergative language isolate. Basque is particularly well suited for this study as it is morphologically ergative, has a structural dative, and has 'quirky' case (which I define as the non-canonical use of structural case, not idiosyncratic case). Moreover, Basque allows causatives of all basic clauses, including its ditransitives, which creates a situation where the language is forced to assign non-structural case to exactly one of the four resulting arguments as Basque does not allow case doubling: I show that Optimality Theoretic Lexical Decomposition Grammar (OT-LDG; Kiparsky, 1997, 2001; Wunderlich 1997) is best suited to modeling these data.; The main results of this dissertation include a new classification of case which defines a category of semantically generalizable non-canonical structural case arrays. I show that this is necessary to account for the Basque data, and provide an analysis of the non-canonical case arrays. Having identified this category as distinct from lexically specified case, I show that there is a Case Ordering Generalization (COG) for all canonical and non-canonical structural case arrays. Another contribution is the Causee Case-marking Principle (CCP) which I propose to account for the case marking in causatives. I show how the CCP is naturally implemented within OT-LDG using the same set of constraints motivated for other case phenomena. This improves upon earlier work by correctly predicting both the morphological form and the structural status of each argument's case. I show that this implementation of my proposal generates the observed typological variation.