| The construct of working memory has played a key role in cognitive psychology's understanding of individual differences on a wide range of tasks, especially differences between young and older adults. Despite this importance, working memory has received little attention from computational modelers, especially in regard to the results from the cognitive aging literature. The thesis presents a computational theory of working memory called SPAN, which tries to take into account what is known about the working memory mechanisms of decay, displacement, and especially processing speed. In many ways, SPAN is a production system instantiation of a slowing theory of working memory and aging much like that advocated by Salthouse. The motivation and background for the theory is laid out and the theory described. To test the theory, models of three tasks are presented: the fan effect sentence verification task, the Digit Symbol speed test, and the Computation Span working memory measure. Not only can SPAN produce the kinds of young vs. old differences observed in empirical studies of these tasks, but the SPAN models mirror the quantitative differences as well. The results provide further support for slowing theory and demonstrate that empirical results from the cognitive aging literature are amenable to the formal techniques usually associated with cognitive science. SPAN presents a promising new way of looking at working memory, and prospects for future work in modeling, perception/action, and learning are also discussed. |
