Can a gamified, rapid, online assessment of letter encoding ability in kindergarten and first grade children predict future reading development?

Abstract

There is a long, contentious history of visual processing theories of dyslexia. Understanding the processes that reliably correlate with reading ability is particularly appealing for developing a dyslexia screening tool since visual processes can be measured, reliably, far before children begin to learn to read. The multi-element processing (MEP) task has the strongest evidence for identifying a subgroup of struggling readers not captured by conventional, language-based measures. In this task, a string of letters (or symbols - pseudo letters) are flashed briefly (240ms) and participants are asked to identify a post-cued element from six choices. Our goal is to investigate if the MEP task predicts future reading development. First, we developed and validated a web-based MEP task to reliably measure task performance in young children. We leveraged the Multitudes platform to collect large and diverse data from kindergarten and first-grade children. We administered the MEP task with 2, 4 and 6 letter strings (N = 231) and observed that task performance is significantly correlated with reading ability [r = 0.36, p = 7.5e-07], replicating our previous finding from cross-sectional data (N = 187, 6-17y) of r = 0.42; p = 9.2e-09. We then used Item Response Theory to calibrate item difficulty (Ө) on an interval scale. We observed similar item difficulty for the four (Ө = 2.122) and six (Ө = 2.269) letter string trials. We removed redundancy in the next pilot of K/1 students (N=51) and observed similar performance in the task with letters (d’ = 0.89) and pseudo letters (d’ = 0.83). Based on our results, we hypothesize that MEP task performance predicts future reading development and if pseudo letters are letter-like, they can potentially contribute to a language agnostic dyslexia screener.