It has been widely accepted that it is difficult to characterize scrambling in terms of A/A'-distinction. One of the most challenging examples in Korean is binding-related examples that are closely connected with (anti-) reconstruction effects. This paper points out that it is problematic to attempt to define scrambling depending on only the availability of reconstruction effects on the basis of the binding principles(Condition B and C). Instead, we need to consider non-canonical relations among anaphors, pronouns, and Rexpressions in understanding the interplay between scrambling and binding in Korean; the complexity of binding relations is disguised as the inconsistent behaviors of scrambling with regard to (anti-)reconstruction. Based on the assumption that scrambling is not a semantically vacuous movement and the non-trivial fact that anti-reconstruction effects are observed in a wider variety of data while reconstruction effects are limited to cases in which anaphors are scrambled, I argue that (at least in binding-related constructions) scrambling disallows ‘reconstruction effects’ and examples that have been considered to be involved in reconstruction effects can be accounted for within the frame of ‘anti-reconstruction effects’.

This paper asks why scrambling is recognized as 'transparent' movements in showing the remedy effect on the intervention phenomenon (e.g. 'LF'-intervention effects, island effects, Weak Crossover effects). Focusing on the non-trivial fact that scrambling induces D-linked interpretation in whquestions, I analyze scrambled elements with the same perspective to D-linked ones. When a wh-word is scrambled, as a D-linking effect, its base copy is inactive (incomplete before [+D] is realized) and its higher copy is active as the complete copy (complete at the outer vP,Spec or TP,Spec). Under this new analysis, we can account for well why scrambling across offending interveners is legitimate. Further, it provides a new perspective on scrambling.