FINITENESS AND THE NATURE OF ISLAND CONSTRAINTS* 12-pt 12-pt

Two competing views of island constraints have emerged in recent work. In one, which we will call the accumulation view, islands result from the accumulation of several independently documented processing difficulties, such as the presence of a filler-gap dependency and an especially complex syntactic structure (cf. Kluender (1998, 2004), Hofmeister and Sag (2010)). In a sentence with an island violation, these difficulties are claimed to exceed the processor’s capacity and the sentence is perceived as unacceptable. In the second view, which we will call the disruption view, islands result from an otherwise unproblematic element that may disrupt a filler-gap dependency and render it illicit (cf. Ross (1967), Chomsky (1986), Rizzi (2004)). Bounding nodes and certain syntactic features, for instance, have been claimed to be of this type; they are not inherently bad, but they lead to unacceptability when they intervene in dependencies of certain types. Though typically seen as competitors, these two approaches to island constraints are not in principle mutually exclusive. It could turn out to be the case that certain island phenomena are due to one and that others are due to the other, or even that a single island constraint stems both from the accumulation of difficulties that overtax the processor and from some element that is disruptive only when it intervenes in a dependency. It is also worth keeping in mind that although the accumulation view is typically associated with processing accounts and the disruption view with grammatical accounts, these associations are not logically necessary. It is not difficult to imagine small grammatical violations accumulating to result in very low acceptability or some element that disrupts the processing of a filler-gap dependency but does not produce any special processing burden on its own. In this paper, we use the above two views of islands as a backdrop as we explore the role of finiteness in island constraints. Ross (1967) first noted that finiteness seems to make island domains even more resistant to extraction, particularly in the case of wh-islands, as in (1).