Implicit arguments in English double object constructions

Pesetsky (1995) argued that both objects in the double object construction must be selected arguments of the lexical verb, based on patterns of optionality. A closer examination shows that this is not correct. The second object of the double object construction and both the NP and PP of the PP frame behave like selected arguments of the lexical verb: the lexical verb determines both whether they can be implicit or not, and how they are interpreted when they are (indefinite versus definite). In contrast, particular lexical verbs determine whether the first object of the double object construction can be dropped, but not how it is interpreted. All implicit first objects are interpreted as pragmatically recoverable definites. Implicit first objects also do not license sluicing, unlike all other implicit objects. I propose a purely syntactic account of these patterns, using the ApplP analysis of double object constructions (Marantz 1993; Bruening 2001). In the analysis, arguments of functional heads like Voice and Appl can only be implicit in the presence of a Pass(ive) head, but arguments of lexical verbs are left implicit through adjunction of an operator to the lexical V. I show that implicit arguments are not projected as NPs in the syntax, contra works like Landau (2010). The failure of sluicing with implicit first objects is then analogous to the failure of sluicing with active-passive pairs. I also suggest that the identity condition on ellipsis makes reference to maximal projections, not to heads as in Rudin (2019).