Processing of LF structure: Evidence from reconstruction for binding

Abstract

This thesis investigates whether the human sentence processing mechanism (the parser) is sensitive to effects of interpretation in real-time processing. We aim to better understand how the parser assigns meaning to a given string of words as the words are presented one at a time in left-to-right parsing. We assumed Chomsky (1995)’s Y-model of the grammar following which syntax plays a central role. In this model, the phonological component (Phonological Form, PF) and the interpretative component (Logical Form, LF) are considered separate in the architecture of the grammar. In processing experiments, the parser is presented with the string of words in its pronounced form (roughly, the PF form). To interpret the string of words, it must derive an interpretable structure for that string (LF structure). This thesis examines whether the parser is sensitive to the LF component of the grammar and if so, how it derives interpretable structures in processing. We investigated constructions that are argued to have distinct PF and LF representations (PF-LF mismatch). In the theoretical literature, these cases are known under the label reconstruction. To systematically investigate potential effects of reconstruction, we used constraints on referential dependencies (Binding Theory, Chomsky, 1981), which are argued to be an LF constraint (Fox, 2000; Nissenbaum, 2000; Fox and Nissenbaum, 2004). More precisely, we examined constructions in which binding constraints seem to be violated based on the PF level of representation but are satisfied at LF. In a series of processing experiments, we manipulated whether the phrase containing a determiner phrase could be interpreted upon first encounter or whether it needed to reconstruct to a syntactically lower position to receive an interpretation. Our results suggest that the parser is sensitive to reconstruction effects in the domain of referential processing, providing evidence that the parser is sensitive to PF-LF mismatches. To our knowledge, the experiments reported in this thesis are the first to directly investigate whether the parser is sensitive to reconstruction operations required for semantic interpretation in processing.