Ian Hocking's Website

[Home] [Déjà Vu] [Proper Job] [Writings] [Talking Stories] [About Me] [PhD] [Resources]

PhD Thesis: 'Resources and Parsing'

On this page you can look at the abstract or full text (Adobe Acrobat PDF) of my PhD thesis.

My thesis falls within experimental psychology, which is the scientific study of mind and behaviour. More specifically, it concerns psycholinguistics. This is normally defined as the study of mental representations and mechanisms used by the Human Sentence Processing Mechanism (HSPM).

Abstract

For some time, psycholinguists have been interested in individual differences in sentence comprehension. This interest centres on the processing of ambiguous sentences. In the sentence, Someone shot the servant of the actress who was on the balcony, there is now evidence to suggest that natives speakers assign syntax – or ‘parse’ – in a manner more consistent with the quirks of their native language than universalist parsing principles. In another instance of the same ambiguity, people have been shown to prefer one reading strategy to another (Corley, 1995). To be sure, these findings of individual differences are overshadowed by the mainstream literature, which typically holds that such differences reflect fixed language effects or noise. The case for differences is further hampered by the labours involved in obtaining them. But in some cases where they have been obtained – for example, the Main Verb/Reduced Relative ambiguity (Just & Carpenter, 1992) – an hypothesis has been proposed that may tie individual differences in reading to individual differences in memory resources. These data are not without criticism, but they provide a working hypothesis that bears pursuit. In this thesis, the working hypothesis is tested in the domain of the RC-attachment (Someone shot the servant…), which has already elicited individual differences.

This thesis, which comprises one off-line study, a second two-phase off-line study, a computational model, another off-line study, and a final on-line study, found no strong relationship between reading span and individual differences in attachment with the most strict measures of working memory. This casts doubt on the role of resource differences in parsing. The first off-line study was exploratory, whereas the second targeted a difference between two major parsing-resource theories. Both studies offered an opportunity for readers with a higher verbal working memory to demonstrate differences in attachment behaviour, but no strong evidence was found - while the second study showed a relationship between one measure of reading span and preference, another stricter measure did not show such a relationship. The final study, which was on-line, followed the creation of a computational model that suggested that looking for the immediate use of contextual information by higher span readers (where the consensus is, at present, that contextual information is used non-immediately). In sum, the role of resources was systematically investigated and no definitive evidence was found. This thesis concludes that resources are an unlikely candidate for individual differences in RC-attachment.