Given that both Chapters 3 and 4 take the CED generalization to be what requires a syntax-based analysis (the syntactic residue of islands, as I have called it), I feel that I should dispel a few doubts cast in the recent literature on the correctness of the CED-generalization, which are distinct from Stepanov's critique already discussed (and dismissed) in the main text (cf. Chapters 1 and 2).

In particular, Chomsky (2008) has argued that some instances of extraction out of A-chains are licit. Likewise, Rizzi (2006), in part building on original observations by Esther Torrego, has argued that extraction out of A-bar chains is sometimes licit. Both Chomsky's and Rizzi's arguments threaten the generalization put forth at the end of Chapter 4, according to which no chain can embed another chain. For this reason, I want to examine both types of situations, and show that the counterexamples to the original CED are only apparent. There will thus be no need to qualify the account of (sub)extraction possibilities defended in the main text.

Let us consider Chomsky’s (2008) cases first. Chomsky focuses on the paradigm in (1). (The judgments are Chomsky’s. CAPS are meant to indicate that the relevant elements are (contrastively) focused. I should note that many speakers consulted regard (1b-b') as degraded.)