What do we wonder is not syntactic

Abstract

This paper discusses the so-called halting aka criterial freezing phenomenon (Rizzi 1997, 2006, 2011, 2014) (analyzed in Epstein 1992 as an arguably deducible (“last resort”) effect of Chomsky’s Strong Minimalist Thesis (SMT), encapsulated as “computationally efficient satisfaction of bare output conditions”). Rizzi’s (2014) insightful and attractive deduction-seeking analysis explains freezing phenomena-the inapplicability of movement to certain syntactic objects-by appeal to a particular (re-)formulation of the independently motivated hypothesis that X projections are invisible for movement, the so-called X invisibility hypothesis. Once a phrase moves to a criterial position, it is argued that given modifications (see below) of Chomsky’s (2013) labeling-by-Minimal-Search analysis, movement halting can also be explained since a phrase moved to a criterial position becomes “an X projection,” hence invisible to movement, thereby explaining criterial freezing. Despite the elegance of Rizzi’s analysis, we present empirical and conceptual arguments disfavoring it. First, the possible empirical problem is that although the analysis prevents the Narrow Syntax (NS) from generating (a) * Which dog do you wonder John likes? by appeal to independently motivated labeling theory and X invisibility, it nonetheless fails to exclude (b) *You wonder John likes this dog . We suggest (i) that (a) and (b) are the same phenomenon, hence are to be captured by the same law; (ii) that (b) is not excluded by label-based X invisibility; (iii) that morphophonological, conceptual-intentional (CI) requirements are independently needed to exclude (b); and (iv) that these requirements automatically exclude (a) as well, rendering label-based X invisibility unnecessary. If on track, this would suggest that halting of movement is in fact an illusion. Despite appearances, there is no syntactic prohibition on the application of successive cyclic movement that prevents (a) (contra Epstein 1992; Rizzi 1997, 2006, 2011, 2014). Rather, it is generated by an allowable movement (freely applied simplest Merge, as discussed below) but results in a violation of the independently needed morpho-phonological, CI requirements. As we will discuss, our re-analysis, shifting the burden from a Universal Grammar(UG)-specific syntactic constraint on Internal Merge to independent interface conditions, allows us to maintain freely applied simplest Merge-Merge (, )  {, }—subject only to 3rd factor laws (such as the No-Tampering Condition and inclusiveness). By contrast, enforcing syntactic halting in NS (movement inapplicability) necessitates in this case departure from (simplest) Merge in two important respects: (i) A dedicated syntactic (non-3rd factor, i.e. UG-specific) constraint is imposed on movement (a label-based incarnation of “X invisibility”); and (ii) this dedicated syntactic constraint must be applied only to Internal Merge (IM), not to External Merge (EM). This asymmetrical constraint prevents the adoption of Chomsky’s important unification of EM and IM as simply two instantiations (which come for free) of the single rule (simplest) Merge. As Chomsky notes, this unification in turn (i) reduces the complexity of the object, namely UG, which must be evolutionarily explained (if indeed it is humanly explicable; see Chomsky 2013); (ii) for the first time renders the hitherto curious property of displacement unremarkable; and (iii) by intertwining IM and EM application, eliminates the long-standing two-stage, level-ordered model within which phrase structure representations are first generated and then undergo transformational rule application (see Epstein, Kitahara, and Seely (EKS) 2014b for further discussion).