Sichel at Berkeley SS-Circle

Professor Ivy Sichel was invited to give a talk at the Berkeley Syntax and Semantics Circle on Friday (Feb 2), titled “Parasitic gap licensing without movement”.

A leading empirical claim in the study of parasitic gaps (henceforth, PGs) is that they are licensed by an A-bar movement chain which stands in a particular structural relation to the PG (Engdahl 1983, and many others). Here I argue, based on the distribution of PG licensing by Hebrew RPs, that the requirement is actually weaker, and should be recast in terms of a licensing A-bar chain, which may or may not involve actual movement. I will show that a non-movement A-bar dependency culminating in a resumptive pronoun (RP) may also license a PG, as long as the structural condition on the relation between the licensing chain and the PG is met. This structural condition is identical to classical cases of PG licensing (Nissenbaum 2000) dubbed by Arregi & Murphy (2023) Parasitic Scope (following Barker 2007):

(1) Parasitic Scope for PG licensing: The PG-containing XP must take the scope of the binder of the licensing variable.

Crucially, I show that the variable may be realized as an A-bar Operator bound RP *which does not involve movement*. In the talk, I present differences and similarities in distribution between Hebrew PGs licensed by gaps and PGs licensed by non-movement RPs. The differences follow, I argue, from differences in the placement of the Operator heading the chain and determining the placement of the PG-containing XP (1). Whereas in movement chains launched from the complement of V, the immediate binder is the trace of movement left in specvP (Nissenbaum 2000), in A-bar binding chains which do not involve movement, the immediate binder is in specCP. Since the PG-containing phrase must take the scope of the licensor’s binder, this derives the core difference in PG-licensing: PGs contained within vP-level clausal adjuncts are licensed by a movement chain but not when the RP is not associated with movement (Sells 1984, and more recently Hewett 2023, based on much broader cross-linguistic coverage) – the binding Operator is too high. In other environments, where the Op binder does satisfy (1), these RPs do license PGs, showing that a movement derivation is not absolutely required.