Accurate reading rate: Validations of machine scoring

Accurate oral reading rate is a common measure of early reading proficiency (grades K-4). A concurrent validation study compared a fully automated measure of accurate reading rate (words read correctly per minute) to DIBELS, a well-established, human-scored assessment. We report data from 174 students at a parochial school in Delaware and 130 students at a public school in Texas. Each student took three forms of an automatically scored test (yielding 898 test forms automatically scored) and also took the grade-appropriate form of the standard DIBELS test, which was scored by a trained person. Test administration order was counterbalanced. Correlations were computed for Machine-Machine and Human-Machine score pairs, for each grade separately and for the total dataset. When all grades are combined, Pearson r = 0.87, and correlations for single grades are all above 0.85, thus at or above the reliability ceiling of the DIBELS criterion score. Also, the automated tests have higher test-retest reliability (r = 0.91) than reported in an independent DIBELS study (Goffreda and DiPerna, 2010). [This research was supported by IES, U.S. Department of Education, via contract ED-IES-17-C-0030 to Analytic Measures, Inc. Disclosure: This report of an accurate automatic method may potentially benefit the authors financially in the future.]