The goal of this paper is to show that strong islands can be derived from the way derivation is linearized, as long as we assume that the derivation proceeds in a topdown fashion.To begin with, I will present one advantage of adopting a top-down approach regarding linearization issue: the Linear Correspondence Axiom (henceforth, LCA, Kayne 1994) can be reformulated in a more derivational and minimalist way.In particular, Kayne assumes that the notion Sushi Mold that derives precedence is the asymmetric c-command.Because of the asymmetric c-command, the LCA rules out the head-complement configuration in bare phrase structures.I show that with a top-down derivation, the problematic asymmetric c-command relation can be eliminated.
With our Top-down LCA, precedence relations are derived from the way phrase structures have been built: roughly, if Y enters the derivation right after X, then X precedes Y.In a second part, I will argue that this non-standard approach to derivation1 and linearization can D-pad/board capture CED effects: it will be demonstrated that subjects and adjuncts are islands because they have to be built in a parallel derivation.