Erin J. Lightman, Philip M. McCarthy, David F. Dufty, Danielle S. McNamara
Recent research in reading comprehension supports the hypothesis that readers are aided by textual cohesion. Traditional readability formulas are not able to effectively assess levels of textual cohesion, nor do they account for potential comprehension obstacles caused by differences in genre. This research employs the computational tool, Coh- Metrix, to assess distributions of both cohesion and difficulty within chapters of expository and narrative texts. We sampled representative sections from the beginning, the middle, and the end of each chapter in three science textbooks, four history textbooks, and six narrative texts. The results suggest that the three domains differ significantly in terms of structural organization and levels of difficulty and cohesion. Differences between these results are discussed.
Subjects: 1. Applications; 13. Natural Language Processing
Submitted: Feb 11, 2007