Higher cognitive functions are reliably predicted by working memory measures from two domains: children's performance on complex span tasks, and infants' looking behavior. Despite the similar predictive power across these research areas, theories of working memory development have not connected these different task types and developmental periods. The current project works to bridge this gap with a processoriented theory, focusing on two tasks assessing visual working memory capacity in infants (the change-preference task) versus children and adults (the change detection task). Previous results seem inconsistent, with capacity estimates increasing from one to four items during infancy, but only two to three items during early childhood. A probable source of this discrepancy is the different task structures used with each age group, but prior theories were not suffi ciently specific to relate performance across tasks. The current theory focuses on cognitive dynamics, that is, the formation, maintenance, and use of memory representations within task contexts over development. This theory was formalized in a computational model to generate three predictions: 1) increasing capacity estimates in the change-preference task beyond infancy; 2) higher capacity estimates in change-preference versus change detection when tested within individuals; and 3) correlated performance across tasks because both rely on the same underlying memory system. Lastly, model simulations tested a fourth prediction: development across tasks could be explained through increasing real-time stability, realized computationally as strengthening connectivity. Results confi rmed these predictions, supporting the cognitive dynamics account of performance and development changes in real-time stability.