Korean particle <-ka> has long been simply known as a subject marker. But subjects without <-ka> are very often realized in sentences, and they are always accompanied by the constraint on the correspondence between linear order and thematic role of the constituents. Besides, the subjects without <-ka> are sometimes not allowed in accordance with several syntactic/semantic properties of the subjects and their predicates, in very complicated ways. Traditional view on <-ka> cannot explain these phenomena. In this article, we adopt Bowers (2001)'s PrP Hypothesis and many-folded predicate structure to solve these puzzles. It's claimed that [Spec, PrP] is the position of the optional realization of <-ka>. The nominal which merges with <-ka> has theta positions of [ComP, VP] or [Spec, PrP], and its case is licensed in-situ through agreement under c-command of an Agr <-si/∅> which is placed directly after a verb stem as a subject honorific marker, in derivations of finite clauses. The optionally projected morphological features of <-ka> attract the nominal, and the lexical meaning of <-ka>(focus etc.) is added after the merge. It is the structurally determined position of <-ka> and the theta positions varied by the argument structures of different predicates that explain why some subjects without <-ka> are not allowed and the linear orders of other constructions are constrained. On the other hand, the projection and merge of <-ka> is necessary when the nominals are semantic restrictions of operators of which arguments are propositions. It seems quite natural that the meaning of <-ka>(focus etc.) is necessarily added to the restrictions because the semantic constituents contain important informations which prevent domains of operation from expanding massively.

• 김신회(연세대학교)