Towards a Derivational, Labeling-based Approach to ‘Floating’ Numeral Classifiers in Korean

This paper investigates the well-known subject-object asymmetry in the ‘floating’ numeral classifier construction in Korean: Unlike objects, subjects cannot be separated from their related floating numeral classifiers by, for example, an intervening object. We first review Shin’s (2017) semantic composition–based analysis of this asymmetry, showing that it confronts non-trivial problems. Taking an alternative approach, we adopt Chomsky’s (2013, 2015) ideas on labeling, answering why unlike object-related ones, subject-related numeral classifiers cannot be ‘floated’ or left behind in the vP or TP domain. Specifically, it is shown that the merger of in-situ subject-related numeral classifiers with vP (i.e., the latter being the merger of v and its complement VP) cannot be labelled properly, or the interleaving of a scrambled object between the double subjects composed of the subject and its related numeral classifier in the TP domain is banned in light of the labeling uniformity or the Subject Condition.