Abstract
This paper reconsiders so-called freezing effects within Chomsky’s (2004 and sub.) Phase Theory. I argue that freezing (or halting) should not be seen as the consequence of an exocentric {XP,YP} structure in which the heads of XP and YP share some feature (cf. Chomsky 2013, 2015) or as the invisibility of X’ projections (cf. Rizzi 2015). Instead, I submit that A-freezing (Chomsky’s 2000, 2001 Activity Condition) and A-bar freezing (Rizzi’s 2006 Criterial Freezing) should be dealt with by different principles: the former follows from an independently motivated rule of efficient computation (the application of cyclic Transfer; cf. Chomsky 2000, Uriagereka 1999), coupled with Labeling Theory (cf. Chomsky 2013, 2015), whereas the latter is simply syntactically vacuous. In line with previous proposals (cf. Gallego 2009; Epstein, Kitahara & Seely 2016), I claim that XPs in edge positions are not frozen in the narrow syntax (they can always move, unless affected by cyclic Transfer). Nevertheless, such XPs may be part of a configuration and thus receive an interpretation at the semantic component (cf. Chomsky 2001, 2004). Therefore, if they move from an edge, the relevant interpretation at that edge (be it topic, focus, etc.) will be lost, as interpretations of the relevant kind (thetaroles, criterial-roles, etc.) cannot accumulate, which I ultimately attribute to a Principle of Interface Freezing, whose effects can be subsumed under the Principle of Full Interpretation (cf. Chomsky 1986a).
