Causal and distributed mechanisms of cortical computation

Abstract

Perception and behavior emerge from the coordinated activity of neurons within local circuits and across distributed brain regions. This view is supported by extensive recurrent and long-range connectivity, as well as large-scale recordings of brain-wide activity. Open questions include which cortical interactions arise from direct causal relationships, and the relative magnitude of shared versus private activity in sensory areas. In this dissertation, we address both questions using simultaneous mesoscale calcium imaging and cortex-wide optogenetic manipulation, high-density electrophysiology, and statistical modeling of population activity and spike counts. In Chapter 1, we review the evidence that cortical activity is distributed across areas and the supporting anatomical connectivity, and then discuss methodological advances for circuit manipulation. In sensory areas, movement, arousal, and other cognitive signals have been proposed as candidate sources of neural variability in sensory responses. Next, we introduce a system for simultaneous imaging and optogenetic manipulation across dorsal cortex (Chapter 2). We show that direct input did not account for homotopic correlations, both at the level of single-neurons and population coupling. In Chapter 3, we developed a Poisson reduced-rank regression framework to partition trial-to-trial variability in primary visual cortex (V1) spike counts, and found that a low-dimensional pattern of activity shared among V1 neurons accounted for most of their variability beyond Poisson. Cortex-wide activity, measured with simultaneous electrophysiology or mesoscale imaging, predicted more than half of this shared variability and subsumed the variability predicted by behavior. Variability in V1 therefore substantially reflects activity shared with the rest of dorsal cortex rather than fluctuations private to V1. Finally, in Chapter 4, we discuss the implications of these findings for distributed neural computation, the limitations of measuring global signals in dorsal cortex alone, and future directions for dissecting causal pathways along which cortical activity is shared.