ScholarWorks@숙명여자대학교

초록

This paper aims to explain the peculiar behavior of the so-called ‘/s/-irregular stems’ of Korean in terms of a new version of Lexicon Optimization which I propose. It will be shown that when there is real evidence for the underlying form -- such as alternations within a paradigm -- learners must attend to that evidence and ignore Lexicon Optimization. From this arises a problem of abstractness in positing the input forms. In order to prevent this undesirable situation, I propose that the lexicon be constrained in some way. I argue that learners posit some prominent prosodic information in the input forms in the course of their learning. I will show that the relevant prosodic information is abstracted away from output forms.

키워드