Islands in the grammar? Standards of evidence

When considering how a complex system operates, the observable behavior depends upon both architectural properties of the system and the principles governing its operation. As a simple example, the behavior of computer chess programs depends upon both the processing speed and resources of the computer and the programmed rules that determine how the computer selects its next move. Despite having very similar search techniques, a computer from the 1990s might make a move that its 1970s forerunner would overlook simply because it had more raw computational power. From the naïve observer’s perspective, however, it is not superficially evident if a particular move is dispreferred or overlooked because of computational limitations or the search strategy and decision algorithm. In the case of computers, evidence for the source of any particular behavior can ultimately be found by inspecting the code and tracking the decision process of the computer. But with the human mind, such options are not yet available. The preference for certain behaviors and the dispreference for others may theoretically follow from cognitive limitations or from task-related principles that preclude certain kinds of cognitive operations, or from some combination of the two. This uncertainty gives rise to the fundamental problem of finding evidence for one explanation over the other. Such a problem arises in the analysis of syntactic island effects – the focus