| This dissertation proposes a theory of syntactic case and examines its synchronic and diachronic consequences. The framework is a unification-based approach to direct linking, in which grammatical functions are derived from argument structure and linkers (cases, word order positions, and agreement). The theory consists of three main constraints on the association of NPs and arguments. Thematic structure is shown to be non-essential in capturing morphosyntactic generalizations. Two of the constraints are based on the notion of restrictiveness, a measure of relative specificity. This theory is first of all shown to handle basic case patterns, including some case alternations in Germanic languages. The same set of distributional features is referred to by all three constraints and so forms a classical natural class.; With two parameters added, this theory predicts a four-way typology of case systems, all four types of which are attested. Data from Classical Greek, Japanese, Warlpiri, and Kabardian are analyzed in terms of the typology. Cases receive cross-linguistic definitions independent of traditional labels (nominative, accusative, ergative, absolutive). The theory, in conjunction with the analysis of a particular language as belonging to one of these types, makes a tightly interconnected set of predictions for each linker. These issues are examined more closely in the syntax of Icelandic, which has rich case and relatively strict word order.; The theory makes a number of diachronic predictions about the relative likelihood of changes in linking, on the assumption that restrictiveness correlates with simplicity. A class of changes that are not parametric is identified and accounted for. Finally, some unusual case patterns are analyzed in terms of the theory. |
