In this dissertation, I propose a novel analysis of so-called Split Topicalization (ST), focusing on German. ST, which seemingly splits constituents into two parts, has been a recalcitrant problem for syntactic theory. The present dissertation argues that it follows directly from fundamental principles of syntactic computation.

Chapter 2 presents the central properties of ST. A brief sketch of its pragmatics leads to the conclusion that ST is not "information-structurally driven," contrary to what is typically assumed in the literature. While ST exhibits all properties of an Ā-dependency, in many cases there is no identifiable base constituent from which the two parts could be derived. This is the empirical problem that so far no analysis of ST has been able to solve.

Chapter 3 develops a novel analysis of ST, based on the idea that the two separated parts are underlyingly related in a "bare-predication structure," i.e. they directly merge as DP subject and NP predicate ({DP, NP}). I argue that this structure is locally unstable: it must be broken by movement to be endowed with a label. This analysis explains why the two parts, while not forming a constituent, nevertheless agree in Case (the result of Multiple Agree) and are obligatorily separated. I show that the analysis correctly accounts for the locality conditions on ST, including its circumvention of the CED, and discuss various implications and extensions of the analysis.

In chapter 4 I propose to extend the analysis developed in chapter 3 to Quantifier Float (QF). I show that QF has in common with ST the property that the two separated parts do not necessarily originate in a single source constituent. In this case, too, the two parts are each an independently generated XP, related to one another by predication: I analyze floated quantifiers as predicates that merge with their DP associates, again creating a locally unstable structure ({DP, QP}) that requires movement.

Chapter 5 concludes by summarizing the theoretical implications of the proposal, pointing to future avenues of research.