Agreement, Coindexing and Reactivation

Of the many interesting issues raised by Barss (this volume) in his reply to Fodor’s paper, we will address just one. This is his objection to Fodor’s proposed explanation, based on phrase structure grammar (Generalised Phrase Structure Grammar, Head-Driven Phrase Structure Grammar), for the results of the cross-modal priming experiments on antecedent reactivation by empty categories. It is agreed on all sides that the experimental data are still incomplete and uncertain in various ways. But Barss is prepared to accept, for the purposes of this argument, the generalisation over the data suggested by Fodor: when the task is cross-modal on-line lexical decision or naming, reactivation is rapid and accurate for overt pronouns and anaphors and for WH-trace, but is slow and possibly inaccurate for passive, raising, and controlled complement constructions, which are analysed within Government Binding theory in terms of NP-trace and PRO. ¹ Fodor argued that the weak effects for NP-trace and PRO are understandable if the cross-modal task addresses the syntactic structure of a sentence, and if GPSG/HPSG is right about the inventory of empty categories in natural language. Overt pronouns and anaphors are present in syntactic structures, and so is WH-trace. But PSG theories do not recognise NP-trace and PRO as syntactic constituents; the facts of passive, raising, and control constructions are explained in PSG without positing empty categories. (See, inter alia, Gazdar, Klein, Pullum, & Sag, 1985; Klein & Sag, 1985; Pollard & Sag, 1987, 1993; Sag & Pollard, 1991).