Computational Models of Age-Associated Cognitive Slowing

Background Cognitive slowing accompanies normal aging, yet understanding of the mechanisms of slowing is limited at the network and neuronal level. Relating the pathophysiological factors responsible for cognitive slowing, and interpreting its relationship to working memory, requires multiscale computer modeling. Objective The aim of this research is to explore multiple mechanisms of cognitive slowing using computational modeling of the cortex to link neuronal activity with cognitive content. Method We developed multiscale computer models of a simple cognitive task - Condition 1 of the Stroop recognition task - using the Nengo system, a cognitive simulation environment with a semantic pointer architecture developed to model cognitive tasks using spiking neural networks. We explored how changes associated with aging such as increased input noise, axonal loss, neuronal loss, and feedback would affect the function of the models. Results Axonal loss and increased input noise produced profound slowing. High levels of neuronal loss severely impaired memory and paradoxically decreased slowing via the ability to respond more quickly by “releasing” a prior memory. Increased feedback improved memory at the cost of increased slowing. Conclusion Our simulations suggest that significant slowing could be caused by white matter loss (axonal loss) or input signal degradation (which could be caused by visual or other afferent system worsening). As neuronal loss markedly decreased the duration of working memory, we propose that physiological feedback is increased to preserve working memory at the cost of further cognitive slowing.