Verb phrase ellipsis and anaphoric deaccenting, although impressionistically quite distinct from one another, present similar theoretical challenges for the interfaces of syntax, semantics, and discourse. For both phenomena, the prior literature has raised questions about the precise division of labor between information from local linguistic antecedents and information from the discourse at large in their licensing and interpretation. The goal of this dissertation is to use novel experimental methodologies to investigate the empirical status of antecedent versus broad discourse information in these constructions, and to use the experimental results to constrain the hypothesis space for the mechanisms underlying their licensing and interpretation. In the literature on verb phrase ellipsis, there is substantial debate as to whether ellipsis sites stand in a relation of syntactic identity with a linguistic antecedent, or are resolved under semantic identity or along broader discourse lines. This dissertation develops an experimental paradigm to assess verb phrase ellipsis interpretation in "complex discourse contexts," where the linguistic antecedent and broader discourse context make different information available for inclusion in the interpretation of an ellipsis site. This is followed by the development of a preliminary probabilistic model of verb phrase ellipsis interpretation in context. The results of both investigations agree that verb phrase ellipsis in context is best understood as hybridizing interpretation with respect to the linguistic antecedent and with respect to the general discourse salience of the competing propositions. In the case of deaccenting, it is widely accepted that material can be deaccented on the basis of being notionally "given" not only when it is identical to an antecedent, but also when its meaning can be inferred, formally or informally, from the meaning of an antecedent. There is disagreement in the literature on whether to represent the deaccenting of antecedent-nonidentical material in the grammar of deaccenting, or as an extragrammatical operation that "rescues" deaccented tokens that fail to meet a grammatical requirement for identity with an antecedent. The dissertation presents a production and perception investigation of the prosody of two types of inferable verbs: verbs whose meanings are entailed by an antecedent, and verbs whose meanings are made available by an informal inferencing relation linking them to their antecedents. The results reliably indicate that repeated and inferable material have a different empirical status, supporting a licensing account that generates deaccented identical material in the grammar and accounts for deaccented inferable material through an extragrammatical process. The results of each investigation indicate that each construction instantiates more complex interactions between linguistic structure and discourse information than have previously been considered at length in the literature. The dissertation separately outlines the grammatical and extragrammatical mechanisms that remain viable as explanations for the behavior of verb phrase ellipsis and anaphoric deaccenting in light of the new data. Following this, the conclusion outlines the prospect for a unified analysis of both constructions, with both holistically considering the discourse at large in the grammar or both augmenting a grammatical identity requirement with extragrammatical consideration of discourse.