Neural Sequence Generation Using Spatiotemporal Patterns of Inhibition

Jonathan Cannon, Nancy Kopell, Timothy J. Gardner, Jeffrey E. Markowitz

Abstract: Author Summary Sequences of stereotyped actions are central to the everyday lives of humans and animals. It was hypothesized over half a century ago that these behaviors were enabled by linking together groups of neurons (or “cell assemblies”) into a feedforward chain using correlation-based learning rules. These chains could then be activated to generate particular behavioral sequences. However, recent data from HVC (the songbird analogue of premotor cortex) paint a more complicated picture: inhibitory and excitatory cells lock to different phases of a rhythm, with inhibitory cells providing windows of opportunity for the excitatory cells to fire. This study puts forward a mathematical model that uses both a feedforward chain geometry and local feedback inhibition to generate stereotyped neural sequences. The chain conducts an excitatory pulse through multiple spatial regions, arriving at each as local inhibition dips. Our simulations and analysis demonstrate that such patterned local inhibition can synchronize the firing of pools of neurons and stabilize spike timing along the chain. Our model provides a new way of thinking about sequence generation in the songbird and in neural circuits more generally.