UCLA

A constituent containing the main predicate of a sentence can go unpronounced, as in "Mary will leave before John will [--]", when certain syntactic, semantic, and discourse conditions are met. This process has come to be known as "VP Ellipsis" (VPE), but this term is misleading: it implies that non-verbal predicates cannot be omitted in the same fashion (they can be), and that VP is the constituent undergoing the operation in question elsewhere (it isn't). This dissertation focuses on the second point, recast here as a research question that has received surprisingly little attention in an otherwise robust literature: exactly what constituent(s) does VPE operate on?

I argue that VPE is a non-uniform operation: two distinct "sizes" of VPE can be diagnosed according to the amount and variety of material that can be omitted under identity with some salient antecedent. I provide a handful of diagnostics that reveal this distinction in VPE size, and I show that, surprisingly, this distinction tracks a previously-known but ill-understood observation in the VPE literature: namely, in certain environments, VPE can apply within a clause whose grammatical voice does not match that of its antecedent (e.g. passive vs. active), while in other environments, such instances of VPE are unacceptable. The diagnostics of VPE size that I present suggest that smaller instances of VPE correlate with those environments that allow voice-mismatches in VPE, whereas larger instances of VPE correlate with configurations that resist such voice-mismatches. I argue that this follows if grammatical voice is encoded in the syntax on a dedicated functional projection located at the edge of the main predicate, and the different sizes of VPE are distinguished by whether the ellipsis site is large enough to include this functional projection ("high-VPE") or is not large enough to include it ("low-VPE"), such that this head remains intact, and thus free to differ featurally from its antecedent (leading to voice-mismatch). This analysis posits a principled, fine-grained distinction in ellipsis size where only coarser distinctions (e.g. VP vs. TP ellipsis) were thought to exist.