In this dissertation, I provide an interface model that accounts for semantic and information-structure effects of constituent order variation and relative prominence in a free constituent order language. I test this model on Serbian, a free constituent order language with flexible relative prominence. Building on Diesing's (1992) Mapping Hypothesis, I argue that the relationship between constituent order variation and information structure in a free constituent order language is mediated by the Quantification structure. The driving force behind constituent order variation is a principle that requires that constituents which participate in domain restriction be overtly moved into the restriction clause of the Quantification structure. This movement has an information structure effect in Serbian in that it determines the domain of the common ground that is relevant for the assertion in the nuclear scope. I confirm the predictions of this model experimentally by testing for speakers' acceptability ratings of simple transitive sentences in short story contexts. I further argue that flexible relative prominence in Serbian is best captured by constraints on F-marking and GIVENNESS of Schwarzschild (1999). I show that neither of these notions can be a factor in constituent order variation, and that the full paradigm of Serbian data can be accounted for only if the F-marking structure is freely mapped on the syntactic structure. I conclude that constituent order variation and flexible relative prominence in a language like Serbian must be driven by independent modules of the grammar (contra Godjevac 2000, 2006). Finally, I use my interface model to explain two widely discussed interface phenomena. First, I account for the distinct distribution and pragmatics of the socalled A-accents and B-accents (Bolinger 1965, Jackendoff 1972) in terms of the Quantification structure. Second, I account for bipartite NPs, a phenomenon observed in a number of free constituent order languages. I argue that the two members of a bipartite NP are base-generated independently of one another, and that a binding relation between them is established via a secondary-predicate relation. Crucially, the two members of a bipartite NP belong to different partitions of the Quantification structure, which explains their special information structure properties.