This article presents a meta-analysis of spelling-to-sound regularity effects in individuals with reading disabilities and reading-level comparison groups. The phonological-deficit model of reading disabilities, coupled with the classic dual-route model of word recognition, has led to two predictions: (a) a specific deficit in the pseudoword reading of those with reading disabilities and (b) an absent or reduced regularity effect for those with reading disabilities relative to reading-level controls. Previous reviews confirm the first prediction. The present meta-analysis tested the second prediction. There was a clear effect of word regularity for individuals with reading disabilities, the magnitude of which did not differ from the word regularity effect for reading-level controls. The authors explore how the contradictory support for these 2 predictions is inconsistent with classic dual-route models of word reading and how connectionist models are consistent with the empirical findings on reading disability.