We explore the application of principles derived from statistical learning theory to model the role of executive processes in cognitive learning. We formulate a model of learning that brings together into a computationally coherent whole a number of empirical observations that in the existing literature have not been systematically related to each other. These observations concern the form of cortical map plasticity during learning, the phenomena associated with hyperacuity in humans, and, on the executive side, the experimental demonstration of the existence of a signal generated by conflict in a decision task.