On the Variability of Island Effects in Korean Fragments and Its Implications

Abstract

This paper investigates the variability of island-(in)sensitivity in Korean fragments (Park 2005a,b). It is widely assumed that fragments in Korean exhibit island-insensitivity. However, it is observed that in multiple fragments (MF) environments, not all fragments exhibit the island-insensitivity: Some MFs are island-sensitive. It is further shown that some instances of “island sensitivity” are only apparent, not resulting from island-violations. I argue that the apparent island-sensitivity is due to intricate interactions of ellipsis and processing. In particular, I claim in line with Bae and Park (2018) that the fragments strongly tend to be processed with the predicate that they first encounter linearly, ignoring the intervening covert/elided arguments (cf. Frazier and Foder 1978, Yoon 2009). In certain contexts, this leads to wrong parsing, giving rise to the apparent island-sensitivity. The analysis is supported by the fact that the same variability is observed in non-island involving contexts. However, I also show that some MFs exhibit genuine island-sensitivity and present directions towards an account and related implications. I suggest that the genuine island-sensitivity arises when the structure is not fully repaired by ellipsis. (Merchant 2001, 2008, Fox and Lasnik 2003).