UC Berkeley

This thesis offers a systematic treatment of discontinuous exponence, a pattern of inflection in which a single feature or a set of features bundled in syntax is expressed by multiple, distinct morphemes. This pattern is interesting and theoretically relevant because it represents a deviation from the expected one-to-one relationship between features and their morphological expressions. I consider cases of discontinuous exponence in verb agreement, TAM morphology, pronoun formation, and negation, showing the relationships among these various types and arguing that a unified analysis is in order.

The empirical foundation of the work is a typological survey of discontinuous exponence in the inflectional systems of 40 genetically and geographically diverse languages. This study establishes discontinuous exponence as a robust phenomenon, worthy of study in its own right, and brings to light new generalizations about the behavior of agreement features.

Working within the framework of Distributed Morphology I develop an analysis of discontinuous verb agreement that accounts for both the robustness and the noncanonicality of the phenomenon and extends naturally to other types of discontinuous exponence. My theory of Cyclic Insertion includes substantial revisions to Distributed Morphology; it rejects key assumptions such as the idea that feature insertion is feature discharge and it offers a view of vocabulary insertion that is compelled and constrained in very different ways than those assumed in the standard theory. Specifically, I assume that morphological insertion operates relative to meaning targets: insertion is motivated when it brings a form closer to its target meaning and is blocked if it cannot do so. The modifications I propose push Distributed Morphology in the direction of deriving discontinuous exponence more naturally. The noncanonicality of the phenomenon is explained with reference to greater complexity in its characteristic derivations.

I argue throughout the thesis for a view in which phi-features (agreement features) are bundled into sets. This view combines two independently motivated ideas -- that feature categories stand in hierarchical relations with one another and that categories themselves can be decomposed -- to develop a rich, two-dimensional phi-set structure. Along one dimension are the fine-grained primitive features and entailments within feature categories, and on the other are hierarchical relations among the categories. These phi-sets have both descriptive and explanatory power; viewed as meaning targets they derive the patterns of discontinuous exponence, and within the system I propose they predict the phenomenon's cross-linguistic tendencies.

A thorough study of discontinuous exponence can illuminate much about the typology and theory of agreement. I will show that a commitment to accounting for the syntax and morphology of an agreement system -- and the interface between the two modules -- can lead to some very interesting insights about the necessary features of a good theory of agreement.