Right Edge Sharing: Finding Structural Constraints in the Post-syntax

How much syntactic information is required for post-syntactic processes?

This dissertation pursues this question through Right Edge Sharing: a recategorization, building on my earlier work (Serova forthcoming, Serova 2026), of a class of well-studied shared exponence phenomena — right node raising, backwards gapping, and suspended affixation.

Right Edge Sharing must be derived in the post-syntax (Chapters 1–4). Shared pivots need not be syntactic constituents, and the constraint governing all of these structures — the Right Edge Restriction, requiring all non-final sharing targets to be linearly rightmost within their host constituents — is stated over linear order, not hierarchy. Since no movement or constituent ellipsis analysis can generate such pivots, I argue these structures arise through PF-Deletion — deletion of form-identical morpho-phonological exponents after syntactic structure is linearized and exponed.

Yet this post-syntactic process remains governed by syntactic structure (Chapters 5–6). If a target within the sharing domain is contained in a coordination, every coordinate of that coordination must contain an identical target. I argue that this Exhaustivity Requirement is the Coordinate Structure Constraint operating in the post-syntax.

This paradoxical mix of linear and structural constraints requires a novel analysis: an exponent-blind syntactic theory cannot enforce the linear order and identity constraints; a syntax-blind phonological theory cannot enforce the structural ones. I therefore propose a structure-sensitive PF-Deletion Algorithm that derives both, and demonstrate that checking these constraints is incompatible with theories of phasal spell-out that destroy syntactic structure.

A second empirical contribution (Chapter 6) is the novel observation of cross-linguistic variation in sharing domains: Right Edge Sharing is unrestricted across syntactic domains in English, German, and Dutch, but strictly confined to true coordination structures (&P) in Turkish. I propose two formal derivations of this distribution, and ask whether coordination is the default or minimal sharing domain, and if so, why.

In short, this dissertation's answer to its opening question is that post-syntactic processes require considerable access to syntactic constituent structure. Specifically, Right Edge Sharing, analyzed as PF-Deletion, requires that the following structures be preserved after linearization and exponence: coordination maximal projections, coordinates of coordinations that contain shared strings, and all non-overlapping constituents containing exactly one occurrence of a shared string. These requirements count against any theory of spell-out, linearization, or the Syntax-Phonology interface that disrupts access to this structural information at the relevant derivational stages.