MODELS OF GAP-LOCATION IN THE HUMAN LANGUAGE PROCESSOR

Laurie Stowe

Ph.D. Dissertation, 1983

Abstract

A review of models of gap-location proposed in the literature shows that these models do not make specific enough predictions to be evaluated by experimental evidence; they must be realized as gap-location procedures within a specific parsing model. Specific parsing models are outlined and evaluated in terms of existing expeirmental results and of two experiments performed as part of this research. These experiments replicate and extend an experiment performed by Crain and Fodor (1983), which demonstrated that an object noun phrase takes longer to process in a WH-question than in the corresponding declarative sentence. Using the same self-paced reading paradigm, Experiment 1 demonstrates that this result also obtains when embedded WH-clauses and if-clauses are compared. This replication suggests that WH-constructions of distinct types are classed together by parsing procedues. Lengthened reading times appear to be characteristic when the processor expects a gap and an over NP appears. On the other hand, subject noun phrases are not more difficult to comprehend in a WH-clause than in an if-clause. This suggests that gap-location procedures differ crucially between subject and object positions. A grammar-related explanation and a parsing-related explanation of this subject/object asymmetry are advanced. The second experiment reported here shows that the processing mechanisms are sensitive to syntactic context. Noun phrases occurring in a prepositional phrase complement to a subject noun phrase do not differ in difficulty between a WH-clause and an if-clause. This result suggests that gap-location procedures are not used in contexts where they are not appropriate, since a gap cannot grammatically occur in this position.