Abstract

The development of connectionist, or neural network, models has provided a new and promising way of understanding cognitive functions. Reading processes, particularly those associated with the identification of words, have been the subject of extensive study through computational modeling with connectionist systems. The general goal of this enterprise has been to develop a theory of skilled word recognition processes, including accounts of the development of that skill (Seidenberg & McClelland, 1989) and of neurological damage that leads to disorders of reading ability (Hinton & Shallice, 1991; Plaut, McClelland, Seidenberg, & Patterson, 1996; Plaut & Shallice, 1993). Connectionist models of these phenomena have come to rely on a division of labor between different sources of knowledge to account for successful and patterns of impaired word reading performance (e.g., Plaut et al., 1996). The general theme pursued in this chapter is the issue of how knowledge sources interact to bring about fluent word reading. The proposition that multiple knowledge sources interact in the process of identifying words has been of fundamental importance in the development of models of reading (e.g., Lesgold &

Citations