Explorations in ellipsis: The grammar and processing of silence

Abstract

Grammatical insights about ellipsis structures reveal that intricate syntactic and informational structure constraints govern ellipsis. But there continues to be a fairly large gap between the predictions of grammatical theories of ellipsis and the actual data. I will pursue a processing approach in which ellipsis is strictly regulated in the grammar but the grammar is embedded in a theory of language processing. The first part of the talk will explore information structure constraints on processing preferences in cases of ellipsis with multiple permissable antecedents, examining the role of ‘main assertion’ (Frazier and Clifton, 2005), contrastive focus (conveyed by pitch accent, or by syntactic clefting) and information structure expectations (Carlson, Dickey, Frazier and Clifton, to appear). It will be shown that ellipsis is marginal in cases where information structure and syntactic conditions are at odds, e.g., with elided constituents taking antecedents in conditional sentences. Various effects support a speculation that ellipsis, though clearly constrained syntactically, may take an antecedent that lies ‘just beyond’ LF. For example, under conditions where combining information from the antecedent of a conditional with prior information would allow cancellation of the antecedent ( Assume X. If X, Y...), then the consequent clause acts as if it is the main assertion of the conditional sentence (Clifton and Frazier, in progress). The second part of the talk will sketch an account of ellipsis in which, say, active elided clauses may not take passive antecedents in the grammar, nor may ellipsis (VP ellipsis or Sluicing) violate island conditions. Examples of both structures have been reported in the literature. It will be suggested that they can be explained in a theory which allows the processor to repair antecedents at LF, subject to the usual constraints on syntactic reanalysis of garden path sentences (Frazier, 2007), and allows the discourse processor to substitute a variable for a focused constituent. The account will be contrasted with accounts where the grammar freely permits active-passive mismatches and permits island violations in sluicing.