A Minimal Model of Neural Computation with Dendritic Plateau Potentials

Abstract

AbstractOver the last two decades, advances in neurobiology have established the essential role of active processes in neural dendrites for almost every aspect of cognition, but how these processes contribute to neural computation remains an open question. We show how two kinds of events within the dendrite, synaptic spikes and localized dendritic plateau potentials, interact on two distinct timescales to give rise to a powerful model of neural computation. In this theoretical model called dendritic plateau computation, a neuron’s computational function is determined by the compartmentalization of its dendritic tree into functionally independent but mutually coupled segments. We demonstrate the versatility of this mechanism in a simulated navigation experiment, where it allows an individual neuron to reliably detect a specific movement trajectory over hundreds of milliseconds with a high tolerance for timing variability. We conclude by discussing the implications of this model for our understanding of neural computation.