Non-incremental parsing strategies: the role of short-term working memory
OA Version
Citation
Abstract
This study sought to investigate further the findings of Caplan et al. (2011) that heightened self-paced reading [SPR] times may cmrelate positively with comprehension measures at points of significantly increased processing load, indicating use of atypical non-incremental parsing methods. Reading comprehension tasks were presented in SPR and eye-tracking paradigms, and three measures of short-term working memory [ST- WM] were administered. Experimental stimuli consisted of early-closure and late-closure sentences, as the former require the reader to revise his or her initial parse of the sentence at a disambiguating word (creating a point of significantly increased processing load) while the latter do not. Independent variables included WM capacity and sentence type, and dependent variables included self-paced reading time [RT], comprehension accuracy, and five measures of eye fixations: first fixation duration, go past time, dwell time, regression out, and regression in. It was hypothesized that participants would show heightened RT correlating positively with comprehension accuracy at the disambiguating point of the early-closure sentences. It was further predicted that this would correspond with a heightened probability of regression out of the area at this same point. Finally, it was predicted that participants with higher ST-WM capacity would be more likely to show these effects, as they were more likely to use them successfully.
No correlation between RT and accuracy was found in the SPR paradigm, and while participants did show heightened probability of regression out of the critical phrase in the early-closure than in the late closure sentences, this phenomenon was also present at nearly all other points in the sentence. A significant interaction between ST-WM capacity and sentence type on RT, such that participants with higher WM capacity showed supra-additively high RT at the critical point in early-closure sentences, was observed. This finding indicates that readers with higher ST-WM capacity may persist in processing sentences longer than their lower ST-WM capacity peers, as argued by Stine-Morrow, Ryan, and Leonard (2000).
Description
Thesis (M.S.)--Boston University