ABSTRACT
Scholars of language and psycholinguistics have been among the first to stress the importance of rules in describing human behavior. Since the early work of Chomsky, a great deal of progress has been made by viewing language structure in terms of a system of rules. These insights clearly contribute to what Marr has called the computational theory of language - an abstract description of language itself and of the tasks facing the language learner and user. Our approach to the question of language representation and learning does not arise from within the tradition of linguistic theory. We are not linguists, but psychologists who have discovered that a new class of models based on highly parallel processing architectures appears to provide rather good accounts of a number of aspects of cognition. We call these models Parallel Distributed Processing (PDP) models.
