The Psychological Reality of Grammar : the Theta Principle in Parsing Performance

Abstract

This dissertation presents evidence for the psychological reality of a grammatical principle, the Theta Principle. It adopts a grammar-derived theory of human natural language processing ? the thematic parser. Parsing phenomena such as garden path effects (The horse raced past the barn fell) and center-embeddings (The mouse the cat the dog chased bit died) show that whenever the Theta Principle is locally violated, severe processing difficulty is experienced. A unique prediction of the thematic parser is provided, along with predictions of other prominent parsing models. Experiments were carried out to establish which prediction was borne out. The experimental paradigm was Magnitude Estimation, used to measure subtle differences in grammaticality judgments in offline experiments. However, the method was validated as a technique suitable for real-time parsing research. It was found that empirical evidence was compatible only with the thematic parser, but not with other models of parsing performance. A grammatical account is suggested for the difficulty of center-embeddings, in contrast to what has been thought about these constructions in the past fifty years. The explanation provided relies on the discovery that center-embeddings are instances of strong islands. Given that processing difficulty in garden paths and center-embeddings is regarded as a local violation of the Theta Principle, these seemingly disparate phenomena are united within a single account. On the basis of the parallel between processing difficulty and the local violation of the Theta Principle, it is argued that the Theta Principle is psychologically real.