Swinburne Romine, Russell E.2013-06-202013-06-202013-04https://hdl.handle.net/11299/151489University of Minnesota Ph.D. dissertation. April 2013. Major: Educational Psychology. Advisors: S. Jay Samuels & Sashank Varma. 1 computer file (PDF); viii, 106 pages.This research study examines the relationship between reader-level variables and text-level variables in a large-scale assessment of reading comprehension for Grade 10 students in Minnesota, administered in 2006. Six narrative passages and the associated multiple-choice reading comprehension items were examined and coded for variation in propositional density, causal connectedness, cognitive demand, item type, and distracter features. A series of Rasch models were used to calculate individual item difficulty for different groups of students, divided on proficiency level. The relationship between student proficiency, item characteristics and item difficulty was investigated. A significant effect of whether multiple choice items demanded an inference from readers on item difficulty was observed for low-ability readers. Significant effects were also observed for the presence of highly causally connected, true but irrelevant information in item distracters for students of each of four categories of ability level. Results are interpreted in light of a causal model of reading comprehension, the Construction-Integration Model and the Simple View of Reading.en-USAssessmentCausal networkPropositional densityPsycholiguisticsReading comprehensionThesis or Dissertation