CTN: Shaul Druckmann

Title: Neural dynamics of short term memory, from mice to human speech

Abstract: Neural dynamics represent the hard-to-interpret substrate of circuit computations. Advances in large-scale recordings have highlighted the sheer spatiotemporal complexity of circuit dynamics within and across circuits, portraying in detail the difficulty of interpreting such dynamics and relating it to computation. Indeed, even in extremely simplified experimental conditions, one observes high-dimensional temporal dynamics in the relevant circuits. This complexity can be potentially addressed by the notion that not all changes in population activity have equal meaning, i.e., a small change in the evolution of activity along a particular dimension may have a bigger effect on a given computation than a large change in another. We term such conditions dimension-specific computation.

I will first briefly discuss our work on motor preparatory activity in a delayed response task in mice were we utilized neural recordings performed simultaneously with optogenetic perturbations to probe circuit dynamics. We first revealed a remarkable robustness in the detailed evolution of certain dimensions of the population activity, beyond what was thought to be the case experimentally and theoretically. Second, the robust dimension in activity space carries nearly all of the decodable behavioral information whereas other non-robust dimensions contained nearly no decodable information, as if the circuit was setup to make informative dimensions stiff, i.e., resistive to perturbations, leaving uninformative dimensions sloppy, i.e., sensitive to perturbations.

I will then present unpublished work on understanding the neural dynamics underlying preparation of speech. First, we found that as words or brief sentences are being prepared to be spoken, we can decode not just the first phoneme to be spoken, but rather multiple components of the speech sequence are decodable, i.e., prepared in parallel. Second, we found that unlike some previous descriptions of sequence preparation, components were not always separated into distinct subspaces, but rather were found in overlapping subspaces with a structured organization of their neural geometry by element and sequence position.