Passage Difficulty Modeling in Oral Reading Fluency Assessments

In oral reading fluency (ORF) assessments, passages contain rich textual information that may inform their levels of difficulty. In this study, we estimated passage parameters using a speed-accuracy measurement model and investigated whether those parameters could be predicted from observable textual characteristics, thereby extending item difficulty modeling from discrete items to reading passages. Despite the dismal overall performance of the models, results showed that baseline linguistic features, including several readability indices were good predictors of passage difficulty and time-intensity parameters. Surprisal features, which are related to the unexpectedness of words in a passage, improved the prediction of time-intensity parameters; however, they did not contribute to the prediction of passage difficulty. These findings demonstrate the potential of derived textual features to explain psychometric passage parameters and support passage calibration, selection, and development in ORF assessments.