Memory retrieval dynamics and storage capacity of a modular network model of association cortex with featural decomposition

Cognitive Networks (Cognits) Process and Maintain Working Memory

Ever since it was discovered in the monkey’s prefrontal cortex, persistent neuronal activity during the delay period of delay tasks has been considered a phenomenon of working memory. Operationally, this interpretation is correct, because during that delay those tasks require the memorization of a sensory cue, commonly visual. What is incorrect is the assumption that the persistent activity during the delay is caused exclusively by the retention of the sensory cue. In this brief review, the author takes the position that the neural substrate of working memory is an array of long-term memory networks, that is, of cognitive networks (cognits), updated and orderly activated for the attainment of a behavioral goal. In the case of a behavioral task, that activated array of cognits has been previously formed in long-term memory (throughout this text, the expression “long-term memory” refers to all experiences acquired after birth, including habits and so-called procedural memory, such as the learning of a behavioral task). The learning of a task is the forming of synaptic associations between neural representations of three cognitive components of the task: perceptual, motor, and reward-related. Thereafter, when needed, the composite cognit of the task is activated in an orderly fashion to serve working memory in the perception-action cycle. To make his points on a complex issue, which has been the focus of his work, and to delineate a frontier for future research, the author refers to several of his own publications and previously published reviews.