Presented at the 15th Annual World Congress on Learning Disabilities, Burlington, MA, 26–28 October 2006

Reading disability (RD) refers to unexplained failure in learning to read and write despite adequate instruction and no primary sensory, emotional, or behavioral disorders. We present a fully automated screening battery that evaluates critical skills relevant for RD diagnosis, designed for unsupervised application. Psychometric validation in 261 children 12–14 years old (7th grade) indicated that computer-based screening can detect children likely to be diagnosed with RD if professionally evaluated, using a linear discriminant function derived from measures taken without supervision within a 30-minute “computer game” interaction. The test-retest reliability in classification (83%) was comparable to the inter-rater reliability among experts (85%). The validity of RD detection, referenced to an expert-judgment criterion, was very high (up to 96%). The sensitivity and predictive value of detection can be balanced against each other to achieve optimal results. We present additional data from 143 children in the 3rd and 4th elementary grades, using a new, age-adapted automated screening battery, demonstrating that computer-based screening can also be applied at this younger age, in an attempt to increase the probability that children in need of remedial intervention will be reliably detected and attended to as early as possible, to maximize the benefit of special instruction. We review the existing approaches to computer-based RD assessment and argue for the value of fully automated screening solutions as a means to address shortage of expert personnel without compromising the reliability of screening. We discuss the implications and caveats of widespread application of computer-based screening in the educational system.