Neural response variability and divisive normalization

Abstract

Cortical responses to repeated presentations of a stimulus are variable. This variability is sensitive to experimental manipulations that are also known to engage divisive normalization: a widespread description of neural activity as the ratio of a numerator (the excitatory stimulus drive) and denominator (the normalization signal). Yet, we lack a framework to quantify the effects of normalization on response variability. We extended the standard normalization model, treating the numerator and denominator as stochastic quantities, and derived a method to infer the single-trial normalization strength, which cannot be measured directly. The model revealed a general reduction of response variability in macaque primary visual cortex for neurons that were more strongly normalized, and during trials in which normalization was inferred to be strong. This framework could enable a direct quantification of the impact of single-trial normalization on perceptual judgments, and can readily be applied to other sensory and non-sensory factors.