Effects of textual and cohesive structure on discourse processing

Abstract

Results from two experiments are presented that tested the hypothesis that processing is facilitated when there is a correspondence between discourse focus and two types of markers of text coherence: cohesion (pronominalization) and textual structure (topicalization). Participants read well‐structured stories in which the coherence marking in the target sentence was either congruent or incongruent with the discourse, or they read randomized lists of sentences drawn from the stories, in which the target sentence appeared in the same serial position, In all conditions, the local environment was held constant. Results from some of the stories indicated that there was a facilitation in reading time for congruent text marking for both cohesion and textual structure. However, this facilitation effect manifested itself at different points in the sentence: at the end of the sentence for congruent pronominalization, and within the sentence for congruent topi‐calization. A slight modification of the original hypothesis can account for the stories that did not show this pattern of results. These results strongly suggest that: (1) readers are highly sensitive to coherence marking devices; and (2) strictly local coherence models cannot completely account for what readers are doing as they read well‐structured discourses. These findings are consistent with models of discourse processing that assume that readers are using coherence marking in the text to help them construct a semantic representation of the discourse.