Delayed Spell-Out and MaxElide

MaxElide, which constrains the size of elliptical constituents, raises a non-trivial question in ellipsis resolution, namely whether ellipsis takes place cyclically or in one fell-swoop fashion. As MaxElide needs to choose the largest constituent for the ellipsis target, to determine the target, the computational system needs to wait until the whole sentence is derived - i.e. the non-cyclic view. However, there are several pieces of empirical evidence that support the cyclic approach. One is so called Modal Complement Ellipsis, found in Germanic languages, and another comes from sloppy ellipsis puzzles. Thus, a tension exists between the definition of MaxElide and the empirical facts. The tension is relieved when Bachrach & Katzir's (2007) delayed spell-out hypothesis is adopted, which allows the computational system to delay spell-out until a variable is bound within a phase. It provides a conceptual ground that the entire conjunct may fall in a single spell-out domain. Within the same domain, PF would choose the largest [E]-marked constituent. The PF account together with the delayed spell-out hypothesis resolves the paradox between MaxElide and cyclicity of ellipsis resolution.