Theories of cortical dynamics have traditionally relied on trial-averaged neuronal responses, largely ignoring neural variability beyond its most basic aspects. Here we developed novel nonlinear analysis techniques to utilise a growing body of data on the variability and co-variability of cortical responses allowing us to delineate the operating regime of cortical dynamics. We show that in a simple yet physiologically realistic cortical circuit model, the stochastic supralinear stabilized network, external inputs dynamically modify the effective connectivity, thus modulating the variability of neuronal responses in a stimulus-dependent manner. This accounts for variability quenching following stimulus onset, a phenomenon ubiquitously observed across multiple cortical areas, as well as more intricate interactions between the external input and the stimulus tuning of individual neurons in determining Fano factors and noise correlations in area MT [1]. The results challenge a commonly accepted interpretation of variability quenching as a hallmark of noisy attractor dynamics.

[1] A. Ponce-Alvarez, A. Thiele, T. D. Albright, G. R. Stoner and Gustavo Deco, Proc. Natl. Acad. Sci. USA 110:13162-13167 (2013).