The goal of this study was to design a small set of phonological, letter, and memory tasks that would reliably identify children likely to develop reading disabilities (RD). We tested children in kindergarten and followed them through 1st grade, layering the investigation by testing various cohorts from diverse geographic, community, and economic conditions. Our strategy involved establishing selection measures and criteria by calibrating indicators of RD on 1 cohort of children, testing the parameters on a new cohort, exploring the relative accuracy of predictors gathered over time, and testing the contribution of dynamic segmentation in which children received varying levels of prompts in performing the task. Discriminant analysis based on a small set of predictors yielded high hit rates in distinguishing children who exhibited an RD profile at the end of 1st grade. Measures taken early in 1st grade were more accurate discriminators than were measures taken late in kindergarten, which were more discriminating than measures early in kindergarten. Whereas segment phonemes and rapid letter naming qualified as primary discriminators of RD at all 3 screening windows, other tasks were discriminators at some but not other screening windows. (author abstract)