Abstract
Neuronal populations in primary visual cortex adjust their responses to the statistical structure of the environment, including both stimulus contrast and the probability of occurrence of visual patterns. Here we show that, across a wide range of adaptation states, the distribution of population responses is well described by a zero-inflated log-normal model with three parameters: the probability of "silence" P0 , the log-response mean μ , and its variance σ2 . Adaptation produces coordinated changes in P0 and μ , whereas σ2 remains approximately invariant. These coordinated shifts collapse the family of response distributions onto a one-dimensional manifold, consistent with the existence of a common gain mechanism underlying both contrast and pattern adaptation. We further demonstrate that μ obeys power-law relationships with stimulus contrast and with orientation probability, and that P0 varies linearly with μ . Finally, we show that these empirical relations arise naturally in a population of linear-nonlinear neurons driven by Gaussian inputs whose mean, but not variance, is modulated by the environment. Together, these results suggest that contrast and pattern adaptation rely on a shared mechanism that adjusts the mean input to cortical populations while preserving the overall structure of their response distribution.