Spatiotemporal Divergence Between Intrinsic And Evoked Cortical Activity Predicts Visual Detection.

Abstract

The threshold for sensory detection varies with fluctuations in intrinsic cortical activity. When stimulus-encoding cortical populations are more excitable, stimuli elicit stronger neural responses that are more likely to be detected. However, the detection of a stimulus is also more likely when cortical populations are less excitable because there is less background "noise". Therefore, it is unclear how the variable states of intrinsic and sensory-evoked cortical activity and their interactions impact sensory detection. We hypothesize the answer depends on the spatiotemporal structure of intrinsic activity states across sensory encoding and non-encoding populations. To test this, we examined intrinsic and target-evoked population activity across cortical Area MT in common marmosets while they performed a threshold visual detection task. We compared detection performance based on target-evoked responses and the state of intrinsic activity in the larger surrounding population. We find that the intrinsic activity in the surrounding, non-encoding population predicted trial-by-trial detection performance better than the population encoding the target-evoked response. Furthermore, we find that the detection performance of the monkey was best predicted by the divergence in excitability between the encoding and surrounding non-encoding population. These findings suggest that, rather than a source of noise or irrelevant to sensory processing, the distributed spatiotemporal state of intrinsic activity directly influences how sensory signals are represented in cortical populations and can influence perceptual thresholds in visual detection. Significance StatementPrior research into how variability in neural activity impacts perception has often focused on neural populations that encode relevant sensory information. However, the role of variable intrinsic activity in nearby, non-encoding populations and their contribution to sensory representations is less well understood. We found that the state of intrinsic activity in non-encoding populations was a better predictor of performance on a threshold visual detection task than the evoked-response magnitude. These results suggest that the state activity in broader neural populations plays a larger role in sensory computations relevant to perceptual decisions than previously regarded.