Extraction Asymmetries in Ellipsis Sites: Phase-based Approach

This study investigates the intriguing asymmetries in overt extraction from ellipsis sites, with emphasis on how the grammaticality of extraction varies depending on the ellipsis target type. Using data from Korean and English, I examine various cases in which overt extraction from ellipsis sites is realizable or hindered. I argue that the phase-based generalization governing extraction from ellipsis sites (Bošković 2014) applies to Korean; namely, extraction from an ellipsis site is acceptable only if the ellipsis target is a phasal complement. To test this generalization, I examine NP argument ellipsis, clausal argument ellipsis, NP-ellipsis in numeral-classifier constructions, fragment answers, and VP-ellipsis in Korean. The findings show that most ellipsis types in Korean align with this generalization; however, the VP-ellipsis poses a unique challenge. In Korean, predicates such as sayngkak-ha(-ta) ‘think’ decompose into a nominal form sayngkak ‘thought’ and the light verb ha ‘do’ (Park, in press), thus complicating extraction from VP-ellipsis sites. I propose resolving this complication by introducing verb movement as a factor in the Korean syntax, which results in ungrammatical cases involving extractions from phasal ellipsis sites.