Description:

The theoretical investigation of ellipsis has long been concerned with understanding the internal structure of ellipsis sites. Two main approaches have emerged: the structural approach, which posits a fully-fledged but phonetically unrealized structure within the ellipsis site, and the non-structural approach, which suggests the presence of a null-pronominal element or no syntactic structure at all. Therefore, the focus of the syntactic component in this thesis is to determine whether an unpronounced syntactic structure exists at the ellipsis site and, if so, what that structure is hiding behind the ellipsis site.In the field of online sentence processing, previous studies have established that comprehension is an incremental process, with the parser constructing hierarchical syntactic structures as it encounters new words. However, in the context of ellipsis, the parser lacks overt material to support the construction of such structures. This leads to a crucial question: How does the parser construct structure within the ellipsis site in the absence of overt material? The sentence processing component of this thesis investigates the nature of the structure built within the ellipsis site and the mechanisms employed by the parser for its construction.To address these questions, a series of offline and online experiments are conducted. First, the study demonstrates that ellipsis sites exhibit grammatical "connections" to their unelided counterparts, known as connectivity effects. Experimental studies specifically focus on the requirement of binominal each in English to be bound by a plural noun phrase in a C-Commanding position, and these structural/relational requirements are observed in sluicing constructions when binominal each is embedded in a sluiced wh-phrase. Another set of experiments explores whether the online processing of the ellipsis site is sensitive to the processing complexity associated with the supposed structure underlying the ellipsis site. Drawing from well-established configurations in previous studies, the research reveals that the processing cost of the ellipsis site corresponds to the processing cost of the antecedent. Thus, the parser constructs the structure of the antecedent within the ellipsis site. Overall, the findings provide compelling evidence that the ellipsis site contains a detailed syntactic structure that parallels the antecedent, and the parser incorporates structural information stored in memory during the online process of ellipsis resolution. These results are largely inconsistent with the cue-based memory retrieval mechanism, which does not account for the inclusion of structural or relational information in the retrieval process.