Cognitive Functions (Normal) and Neuropsychological Deficits, Models of

Abstract

Researchers interested in human cognitive processes have long used computer simulations to try to identify the principles of cognition, by building computational models that embody a set of principles and then examining how well the models capture human performance in cognitive tasks. Early models were based on information-processing diagrams in which cognitive processing is carried out by a series of discrete stages. However, formalisms based more closely on neural computation—including connectionist or neural-network models—have proven more effective at capturing the effects of brain damage on cognition. In such models, cognitive processes take the form of cooperative and competitive interactions among large numbers of simple, neuron-like processing units. Many such models use distributed representations in which cognitive entities like words, objects and concepts, are encoded by alternative, overlapping patterns of activity. On this approach, the internal representations needed to perform various tasks are not stipulated in advance but are learned though feedback and interaction with the environment. Connectionist models have been applied successfully to a wide range of neuropsychological domains, including perception, attention, language, memory, executive control, and action. Considerable work remains, however, in extending such models to address more complex temporal phenomena.