Different state-dependence of population codes across cortex

Abstract

During perceptual decision-making tasks, behavioral performance and its neural correlates vary with changes in internal states such as arousal, motivation, and strategy. It is not yet clear whether these state fluctuations differently impact the distributed neural processes that underlie task performance, from sensory processing to control of motor outputs. The posterior parietal cortex (PPC) is an association-level cortical region implicated in sensorimotor transformations, but evidence of its causal requirement in decision-making tasks has been contradictory. We trained mice to perform a navigation-based sound localization task and asked how natural fluctuations in behavioral state related to neural processing in association areas during decision-making. Behavioral performance in expert-level mice was not static but instead transitioned between periods of near-perfect performance and biased, less accurate performance. Using a hidden Markov model, we could reliably define these as distinct strategies that included a high-performance state where mice used relevant stimulus information to inform choices, and two biased states where mice weighted stimulus information less strongly. Optogenetic inactivation of PPC decreased task performance accuracy, and the model captured the resulting change in behavioral strategy as a reduction in the weighting of the auditory cues and an increase in behavioral bias, in predicting mices decisions. Two-photon calcium imaging revealed that performance states strongly influenced population activity patterns in PPC, but not primary auditory cortex (AC). Surprisingly, activity of individual PPC neurons was better explained by external inputs and behavioral variables during biased behavioral performance, while shared variability across neurons in PPC was strongest in the high-performance state. Together, these findings suggest that neural activity in parietal cortex is causally required for decisions and is linked to behavioral performance states.