A Critique of Phase Extension, With a Comparison to Phase Sliding

ing away from the data for a moment, observe that nothing precludes subject raising after PI. Concretely, the possibility remains for the in situ subject to become an outer-Spec-F. Unless we stipulate otherwise, which is what DD does by endorsing (10): (10) Adjunction to meaningless categories is disallowed. [from Den Dikken 2006b: 5] Assuming F is a meaningless category, (10) does give a way to address the data in (9), preventing the undesirable scenario just described. Now consider each of the central presuppositions in turn: “equidistance” and (10). First, “equidistance” can be dispensed with if, as Chomsky (2001) argues (contra Chomsky 1993; 1995; 2000), strict c-command and evaluation at the phase level are the relevant notions when calculating intervention effects (see the next section). Second, although there may be principled ways of questioning the existence of multiple specifiers (see Boeckx 2006), (10) comes with its own difficulties. To begin with, it is not obvious how a “meaningless category” is to be understood (a placeholder? a featureless item? one with merely uninterpretable features?). More importantly, if F is truly meaningless (making no semantic contribution whatsoever), why should it exist? We underscore this latter point since, after all, it was Chomsky’s (1995) main argument to rule out agreement projections altogether. If, given the limitations just noted, PE is in the end not relevantly involved in the examples in (9), their ungrammaticality must of course follow from something else. 7 Chomsky (1993) invoked equidistance to solve an apparent superraising paradox under the assumption that objects raise to a position above the base position of subjects: be it Spec-AgrO or Spec-v*. DD appears to assume this when discussing Object Shift (pp. 10-11), but as noted by Lasnik & Boeckx (2006: 119) the problem goes away under the Koizumi (1995)/Lasnik (1999) analysis of object raising, in which these dependents move to a position that doesn’t have to bypass the in situ subject. See Hiraiwa (2005) for empirical evidence against “equidistance”. A Critique of Phase Extension, With a Comparison to Phase Sliding 69 Moro (2006) provides an answer to (9a), which Gallego (2007) attempts to push to the other examples in (9). We will not have space to go into this analysis here.