Inter-neuronal correlation distinguishes mechanisms of direction selectivity in cortical circuit models

Direction selectivity is a fundamental physiological property that arises from primary visual cortex (V1) circuitry, yet basic questions of how direction-selective (DS) receptive fields are constructed remain unanswered. We built a set of simple, plausible neuronal circuits that produce DS cells via different mechanisms and tested these circuits to determine how they can be distinguished experimentally. Our models consisted of populations of spiking units representing physiological cell classes ranging from LGN cells to V1 complex DS cells. They differed in network architecture and DS mechanism, including linear summation of non-DS simple-cell inputs or nonlinear pairwise combinations of non-DS inputs. The circuits also varied in the location of the DS time delay and whether the DS interaction was facilitatory or suppressive. We tested the models with visual stimuli often used experimentally, including sinusoidal gratings and flashed bars, and computed shuffle-corrected cross-correlograms (CCGs) of spike trains from pairs of units that would be accessible to extracellular recording. We found that CCGs revealed fundamental features of the DS models, including the location of signal delays in the DS circuit and the sign (facilitatory or suppressive) of DS interactions. We also found that correlation was strongly stimulus-dependent, changing with direction and temporal frequency in a manner that generalized across model architectures. Our models make specific predictions for designing, optimizing, and interpreting electrophysiology experiments aimed at resolving DS circuitry and provide new insights into mechanisms that could underlie stimulus-dependent correlation. The models are available and easy to explore at www.imodel.org.