My research is focused on how domain-general perceptual and learning processes give rise to specialized language functions in reading and speech perception. My work on reading is driven by large-scale PDP models that learn to map among the written forms of words, their pronunciations, and their meanings. The models instantiate the theory that there is a universal functional architecture for reading across languages, despite large surface differences in writing systems. Testing predictions from the models involves collecting behavioral and functional neuroimaging data as part of a network of researchers in eight countries. Because the kinds of models I use have an important learning component, they can be used to address questions about typical and atypical development. I have recently begun work on a number of large-scale collaborative projects focused on this translational application of modeling. One exciting new direction in this work is the development of new techniques for relating computational models to neurobiological models that provide an alternative to the “box and arrow” approach still common in cognitive neuroscience.