Korean is a nominative-accusative language that permits case-stacking - the realization of multiple case-markers on a single noun phrase. Consider the examples in (1):
(1) Korean case-stacking a. Cheli-hanthey-ka ton-i. isse. C.-dat-nom money-nom have ‘Cheli has money.’ b. Swunhi-ka Yenghi-hanthey-lul chayk-ul cwuesse. S.-nom Y.-dat-acc book-acc gave ‘Swunhi gave Yenghi the book.’ c. Sensayng-nim-tul-kkeyse-man-I kulen il-ul teacher-HON-PL-h.nom-only-nom that.kind work-acc hasipnita do ‘Only teachers do such work.’
In this talk, I argue that data like those in (1) have two strong implications for the proper modeling of the assignment and realization of case morphology.
First, I demonstrate that the full range of case-stacking possibilities strongly supports adopting a Dependent Case model of case assignment; case morphology indicates the position of nominals with respect to one another within a local domain, i.e. within a phase (e.g. Yip et al 1987, Marantz 1991, Baker 2015). In this respect, (1c) is of particular importance. Nominative-nominative stacking is incompatible with the more commonly adopted Agree model of case assignment, whereby C/case is assigned to DPs by nearby functional heads (Chomsky 2000, 2001). It is impossible for a nominal to enter into two Agree relationships with the same functional head for the purpose of valuing the same features. However, if case reflects the position of a DP in a local domain, it is quite easy to imagine that a DP could occupy the same position (relative to other DPs) in two domains, yielding multiple instances of the same case morphology.
Second, in order to capture the full range of licit and illicit case combinations in Korean case-stacking constructions, I argue that nominative case is best modeled as the absence of a Case-feature value. That is to say, nominative case is the overt realization of [Case:__] (see also Kornfilt & Preminger 2015, Levin 2017). This conclusion challenges a central tenet of Minimalist syntax – the position that all syntactic operations are triggered by legibility requirements imposed by the PF and LF interfaces (Chomsky 2008; see also e.g. Pesetsky & Torrego 2007, Preminger 2014).