Mechanisms of visual-cognitive aging: A neural network account

Abstract

Two neural network simulations were run for the purpose of specifying the network parameters that best describe age-related interindividual and intraindividual variability in figure completion performance. the network architecture was modified in three ways that corresponded to plausible computational and neuro-anatomical constraints. In two simulations in which the rate of increase of activation within network units was constrained during testing, the performance of the network closely resembled the relevant empirical data. These findings supported a computationally specific model of age-related differences in figure completion performance rather than a diffuse model of age-related slowing. It is suggested that the analysis of fit between the results of computer simulations and analogous empirical data provides a useful method for the explication of aging mechanisms, and for distinguishing among competing hypotheses and constructs in the study of cognitive aging.