ORGaNICs: A Canonical Neural Circuit Computation

Abstract

A theory of cortical function is proposed, based on a family of recurrent neural circuits, called ORGaNICs (Oscillatory Recurrent GAted Neural Integrator Circuits). Here, the theory is applied to working memory and motor control. Working memory is a cognitive process for temporarily maintaining and manipulating information. Most empirical neuroscience research on working memory has measured sustained activity during delayed-response tasks, and most models of working memory are designed to explain sustained activity. But this focus on sustained activity (i.e., maintenance) ignores manipulation, and there are a variety of experimental results that are difficult to reconcile with sustained activity. ORGaNICs can be used to explain the complex dynamics of activity, and ORGaNICs can be use to manipulate (as well as maintain) information during a working memory task. The theory provides a means for reading out information from the dynamically varying responses at any point in time, in spite of the complex dynamics. When applied to motor systems, ORGaNICs can be used to convert spatial patterns of premotor activity to temporal profiles of motor control activity: different spatial patterns of pre-motor activity evoke different motor control dynamics. ORGaNICs offer a novel conceptual framework; Rethinking cortical computation in these terms should have widespread implications, motivating a variety of experiments.