Free choice and negative polarity: A compositional analysis of Korean polarity sensitive items

This dissertation addresses the following three questions regarding polarity sensitivity. (1) What is the nature of the relationship between NPIs and FCIs? What forms the common core of Polarity Sensitive Items (PSIs)? (2) If there is a common core, is it derived from only one source or could there be more than one source? How many sources can be detected in natural language? (3) What makes the distinction between NPIs and FCIs? In an attempt to answer these questions, this dissertation investigates Korean PSIs, where it is transparent from the morphology which part is common to FCIs and NPIs and which components are responsible for differentiating them. Korean PSIs consist of one of the two indefinite roots, amwu- and wh-, and one of the three particles, -to \u27also/even\u27, -lato \u27even\u27 and -na \u27or\u27. Combining these components results in the six items: amwu-(N) -to/-lato/-na and wh-(N)-to/-lato/-na. I propose that the Korean indefinite root amwu- induces proper domain-widening à la Kadmon and Landman (1993), and that the other root wh- does not induce domain-widening but rather ranges over a regular or a contextually salient domain. From the fact that all the wh-PSIs, which lack the property of domain-widening, still function as NPIs or FCIs, I suggest that the proper domain-widening of amwu- is not a necessary condition for polarity sensitivity, and propose that it is the particles -na, -to, and -lato that are responsible for deriving NPI-hood/FC-ness in Korean. All these particles commonly yield a no matter wh… or indifference reading. Furthermore, I argue that two kinds of sources of PSIs are observable in Korean. The first source is the disjunctive particle -na \u27or\u27, which I analyze as equivalent to English -ever in a subtype of -ever free relatives. Their semantic contributions are formulated with the presupposition of counterfactual variation (von Fintel, 2000), from which an indifference reading is inferred. The second source is the scalar focus particles -to, which is analyzed as PPI -even (Rooth, 1985), and -lato, which is analyzed as NPI-even (cf. Guerzoni, 2003). I show that the combination of the presuppositions that the even- particles trigger derives an indifference reading