The previous chapter revealed the serious explanatory limitations of (currently available) accounts of (strong) islands couched in purely syntactic terms. Although phase-based analyses tend to resort ultimately to more general, non-syntactic considerations to motivate the existence of phases (e.g., computational efficiency), I was at pains to show that phases in and of themselves do not impose conditions that can account for islands, if only because the periodic ‘forgetting’ (opacification) that phase-based derivations impose on the syntactic computation applies to the most transparent domains (complements) in the island landscape. As a result, phase-based analyses, like previous bounding nodes/barriers-inspired attempts, are forced to resort to syntax-internal, lexical properties like ‘L(exical)-marking,’ or ‘edge features’ and formulate constraints on the basis of these (proper government, edge feature condition, and the like) to prevent the generation of island-violating structures.

It goes without saying that such accounts are compelling only insofar as these lexical properties and the conditions imposed on them can be given a natural explanation. To date (after close to 40 years of trying), this is not the case. For this reason alone I think that it is worth looking outside pure, narrow syntax to seek properties of the external (mental) systems with which syntax interacts that could explain why certain structures generated by the syntactic component are judged less acceptable than others. Put differently, I am suggesting we exploit the sort of interface-based accounts that linguistic minimalism favors because such accounts tend to maximize the cognitive resources independently made available by the external systems of thought and sound/sign, leaving the content of narrow syntax maximally minimal and simple (hence, easier to rationalize a priori).