Different fields of research within the cognitive sciences have investigated basic processes in reading, but progress has been hampered by limited cross-fertilization. We propose a theoretical framework aimed at facilitating integration of findings obtained via these different approaches with respect to the impact of visual factors on reading. We describe a specialized system for parallel letter processing that assigns letter identities to different locations along the horizontal meridian within the limits imposed by visual acuity and crowding. Spatial attention is used to set up this system during reading development, and difficulty in doing so has repercussions in terms of efficient translation of the orthographic code into its phonological counterpart, and fast access to semantics from print.