Comparative markedness in Korean palatalization

Abstract

Two kinds of Korean palatalization ? t- and s-palatalization ? contrast in that while t-palatalization applies only across morpheme boundaries, s-palatalization applies across the board. McCarthy (2003) accounts for this with the ranking OO -N PAL ≫ IDENT ≫ PAL wherein palatalization is usually blocked because of IDENT ≫ PAL. However, if a PAL violation is created by morphological concatenation, then the ranking OO -N PAL ≫ IDENT triggers palatalization. The problem with this analysis is that it does not distinguish between t- and s-palatalization ? it incorrectly predicts that s-palatalization should also be blocked in non-derived environments. We show that the difference between the two kinds of palatalization lies in the faithfulness violations involved. In the mapping /s/→[?], only IDENT[place] is violated; however, /t h /→[t∫ h ] violates both IDENT[place] and IDENT[strident]. The constraint that blocks palatalization of /t/ in non-derived environments in McCarthy’s analysis must therefore be the IDENT[strident], a constraint that is not involved in s-palatalization. IDENT[place], the only faithfulness constraint involved in s-palatalization, has to rank below O PAL, so that s-palatalization is triggered even in non-derived environments.