Embedding Preschool Assessment Methods into Digital Learning Games to Predict Early Reading Skills

Abstract

The aim of this pilot study was to explore the predictive accuracy of computerbased assessment tasks (embedded within the GraphoLearn digital learning game platform) in identifying slow and normal readers. The results were compared to those obtained from the traditional paper-and-pencil tasks currently used to assess school readiness in Finland. The data were derived from a cohort of preschool-age children (mean age 6.7 years, N = 57) from a town in central Finland. A year later, at the end of first grade, participants were categorized as either slow (n = 11) or normal readers (n = 46) based on their reading scores. Logistic regression analyses indicated that computer tasks were as efficient as traditional methods in predicting reading outcomes, and that a single computer-based task—the letter–sound knowledge task,—provided an easy method of accurately predicting reading achievement (sensitivity 95.7%; specificity 81.8%). The study has practical implications in classrooms.