Subject Islands are different

Syntactic islands vary in the degree of their opacity, with the well-known contrast between strong and weak islands (Cinque 1990; Rizzi 1990, 2003; Szabolcsi and Zwarts 1993, a.o.). Until recently, decisions about the strength of particular islands relied on individual judgments of the researcher or cursory interviews with fellow linguists; most judgments have been based on English. Some islands, for instance adjuncts, have come out uniformly strong and, as a result, have given researchers confidence in the notion of syntactic opacity (but see Truswell 2007, 2011 for a subset of adjuncts which are transparent for semantic reasons). Other islands show much more variation, both within English and across the few other languages that linguists have considered. Subject islands belong to this latter category, and their degree of opacity has been the cause of disagreement among linguists. Starting with English, extraction out of subjects shows a range of acceptability depending on the predicate; cf. the following examples based on Chomsky (2008): It was the CAR (not the TRUCK) of which the driver ___ arrived late/was awarded a prize. *It was the CAR (not the TRUCK) of which the driver __ caused a riot. The acceptability of examples such as (1a) has also been supported by experi-mental work (Hiramatsu 1999, 2000), which compared adjunct islands with subjects of unaccusatives and showed that the latter were fairly transparent.