A Derivational Approach to Wh-Scrambling

In this paper we see that Wh-phrases in Korean (and Japanese as well) need not necessarily be licensed at LF, a crucial assumption taken by the LF reconstruction approach of scrambling. We will see that Saito's original examples indeed receive a better account on our new assumption that Wh-phrases are licensed in the course of derivation, but not at LF. The derivational approach is also shown beneficial in providing a unified and principled account as to otherwise awkward phenomena across languages (including a weak crossover and a sentence with a Wh-phrase outside a relevant CP domain in Korean and Japanese). Based on these findings, we conclude that the LF-undoing hypothesis of scrambling is not strongly tenable.