Korean has an ambiguous construction between passive and causative.  Whitman and Han (1988) claim that this ambiguity is resolved by assuming two different derivations: the passive interpretation involves possessor ascension while the causative interpretation does not.

            In this paper, I propose that the motivation for the possessor ascension movement is the EPP feature if we use the Minimalist Framework.  If Whitman & Han (1988) and Kim & Pires (2002) are on the right track, their analyses lead us to ask two questions arising from the Minimalist Program view: (i) what is the feature that licenses the possessor ascension? (ii) how can the VP-internal possessor raise to [Spec, IP], skipping the DP baby-DAT in [Spec, VP], apparently violating the Minimal Link Condition (MLC)?  As an answer to the first question, I propose the EPP feature licenses the possessor movement.  For the second question, I propose that V moves to T and it makes the two DPs (VP-internal possessor and DP baby-DAT in [Spec, VP]) equi-distant.

I assume that the causative and passive suffixes are placed in v and project vP.  In the causative reading, NOM-marked DP is base-generated at [Spec, vP] and moves to [Spec, IP].  In contrast, a NOM-marked subject in the passive interpretation is base-generated with genitive case in the specifier position of the ACC-marked DP and raises to [Spec, IP].  EPP is an uninterpretable c-selectional feature on T and it requires the presence of nominal in the specifier of T to be checked off (Chomsky 2000, 2001).  In the causative, EPP is checked when the causer John moves from [Spec, vP] to [Spec, TP].  In the passive, the genitive DP from the specifier of the ACC-marked DP moves to [Spec, TP] to check EPP. One possible problem arising from this analysis might be that we have another candidate, baby-DAT, to check the EPP.  But, I will argue that it is PP, not a possible candidate to check EPP.