Syntactic islands by Cedric Boeckx (review)

Syntactic islands (henceforth SI) opens with the bold statement that ‘island effects are perhaps the most important empirical finding in modern theoretical linguistics’ (ix). Indeed, it seems that hardly any linguist with an interest in clausal syntax would question the central importance of the study of locality. In Chomsky’s (1980:194) informal definition, ‘a phrase is an island if it is immune to the application of rules that relate its parts to a position outside of the island’. Given that we can turn 1a into 1b by replacing the object with a wh-phrase and moving it to the front, why does the same mechanical transformation fail when applied to 2a, yielding the decidedly odd 2b?

1. 1.

   1. a. John likes apples.

   2. b. What does John like?

2. 2.

   1. a. John likes apples and oranges.

   2. b. *What does John like and oranges?

This is the problem of islands as standardly assumed: why do reorderings from some domains but not from others yield sentences we perceive as deviant? And how do we come to know this without ever having encountered direct evidence for it? Regrettably but understandably given the book’s theoretical orientation, SI focuses almost exclusively on the first question; see Berwick et al. 2011 and Pearl & Sprouse 2013 for recent discussion pertaining to the second.

Boeckx pays ample tribute to Noam Chomsky and John Ross as pioneers of the study of syntactic locality. A concluding chapter rightly praises Ross for not merely compiling a catalog of island domains but also asking much harder why and where questions: why do islands exist, and what is their locus—grammar itself or interfacing domains of cognition? SI surveys a variety of approaches to these questions, beginning with the origins of the study of locality: Chomsky’s (1964) A-over-A condition and Ross’s (1967) seminal study it inspired, which in turn informed groundbreaking works by Chomsky (1973, 1977). It was here that islands were first recognized as a diagnostics for movement and that a systematic attempt was made at formulating rules of some generality that would capture the impenetrability of certain domains across constructions, culminating in the theory of subjacency (restricting rule application to subjacent cycles, supplemented with provisos to permit ‘long’, successive-cyclic movement). B briefly reviews related notions such as weak islands, relativized minimality, and freezing, but the focus of the book is on strong islands and Huang’s (1982) condition on extraction domain (CED). The background discussion in this opening chapter is helpful to understand the genesis of more contemporary notions such as barriers and phases.

Ch. 2 (whose title is blatantly misspelled in the copy I reviewed) discusses ‘reductionist’ approaches to islands that seek to reduce locality effects to extragrammatical factors such as parsing, premised on the idea that dependencies straddling island boundaries exceed certain cognitive limitations (cf. Miller & Chomsky’s 1963 classic discussion of multiple center-embedding). B rightly notes that this mode of explanation is a priori attractive in that it reduces UG-internal complexity, much in line with ‘minimalist’ ambitions (Chomsky 2007). At the same time, however, he makes it clear that he sees little hope for a comprehensive reduction to performance factors; he deplores the ‘absence of a theory of cognitive constraints’ (38, his emphasis) and points out that no such account has provided a satisfactory explanation of cases like 2b above. B cites two arguments against nongrammatical approaches but concedes that ‘reductionist attempts … may be the best explanation we can hope for in at least some cases’ (51).

This reconciliatory note notwithstanding, the tone remains pessimistic, and Ch. 3 is devoted to showing that grammatical accounts of island effects hardly fare any better. B pursues the right strategy to demonstrate this negative result: he discusses at length one of the most explicit formalist proposals to date, namely Müller’s (2010) attempt to derive CED effects from a version of Chomsky’s phases, supplemented with an intricate theory of ‘edge features’. While misrepresenting some of the predictions of this analysis...