Large-scale cross-brain recordings during cognition: initial steps towards finding meaning in all that variability in neural activity

Abstract:
Neural firing rates are known to be highly variable, responding in different ways to different repeats of identically prepared trials. It has been known for decades that this variability is highly structured across neurons and brain regions, but whether it is meaningful or simply noise has remained unclear. Recent advances in recording technology are now making it possible to record simultaneously from many individual neurons across multiple regions of the brain, opening new opportunities to study coordination of neural activity across brain regions. I will describe recent experiments from my lab in which we are pushing the envelope of such simultaneous electrophysiological recordings, with ~80+ neurons recorded simultaneously from each of more than 30 brain regions (~2,700 simultaneous units). We are carrying out these recordings in rats performing a sensory decision-making task. We found an internal neural signature of commitment to a decision—in other words, a neural signal that indicates when the subject makes up their mind. It is an internal signal in the sense that it occurs and can be detected even if the subject makes no outward motor act to indicate that they have made up their mind. In terms of behavior, the signal indicates when sensory evidence stops affecting the subject’s upcoming decision, and in terms of neural activity, the signal indicates major state changes in decision-related spiking across the brain. Seeking to identify further such internal signals, which we hypothesize could explain a substantial fraction of the apparently noisy structured variability in neural firing rates, I will also describe new machine learning-based methods that use simultaneous recordings to identify and succinctly describe structure in joint neural activity.