The trial-to-trial variability of neuronal responses provides detailed information on how the circuit structure connecting neurons drives brain activity and behavior. This idea combined with the widespread use of population recordings has prompted deep interest in how variability is distributed over a population. A pervasive yet puzzling feature of cortical circuits is that despite their complex wiring, population-wide shared spiking variability is low dimensional with all neurons fluctuating en masse. Previous model cortical networks are at loss to explain this low dimensional shared variability. Rather than explain this variability within the network, these models typically explain it by appeal to external sources such as global alertness of the animal, wandering of attention, or just noise from other brain areas.
Further, attention-mediated modulations in population variability provide constraints on how shared variability is distributed within and between neuronal populations. Attention reduces within area correlations (Cohen & Maunsell, 2009, Nat Neurosci 12, 1594–1600) while simultaneously increasing between area correlations (Ruff & Cohen, 2016, J Neurosci 36, 7523-7534). In a recent paper in Neuron “Circuit Models of Low-Dimensional Shared Variability in Cortical Networks”, a collaboration between the labs of CNBC faculty Marlene Cohen and Brent Doiron, the authors show that such a differential correlation modulation is a difficult constraint to satisfy with a model where the source of the fluctuations are strictly external to the network.
Lead author, postdoctoral fellow Chengcheng Huang notes: “We argue that a global source of fluctuations is not able to explain the differential modulation of within and between area correlations and that the observed low-dimensional variability can be internally generated within the recurrent circuit.” The group develop spiking neuron network models where population-wide shared variability is internally generated.
The authors studied networks of spiking neurons with spatially ordered connections, meaning that nearby neurons are connected with higher probability. These networks showed that if the spatial and temporal scales of inhibitory coupling match known physiology in monkeys, the network exhibits spatiotemporal dynamics over large spatial scale, which results in low dimensional shared variability. This matches the observed variability of in vivo population recordings in cortex. A top-down modulation of inhibitory neurons in the authors’ network provides a parsimonious mechanism for attentional modulation on both within and between area correlations, providing support for the author’s theory of cortical variability. The work provides a critical and previously missing mechanistic link between observed cortical circuit structure and realistic population-wide shared neuronal variability and its modulation.
