Representation and utilization of sensory uncertainty in the primate visual system Accurate perception is vital for survival, but available sensory information can often be impoverished. Observers act with seeming knowledge of their own uncertainty when making perceptual decisions as well as reflecting on the accuracy of those decisions. This implies that neural circuits which process sensory information also represent the uncertainty of this information. How does this work? This proposal seeks to answer this question and investigate its implications for perception and behavior. By combining theoretical neuroscience, electrophysiological recordings, and perceptual experiments, this proposal will integrate results from human and nonhuman primates to combine the characterization of sensory uncertainty and its neuronal basis. This proposal comprises three specific research and training aims. First, we will leverage theory and physiology to develop and test a novel computational theory of how sensory neurons encode stimulus uncertainty (Aim 1). We propose a view that stimulus features are encoded by the response mean of visual neurons, but stimulus uncertainty is encoded by variance in the response gain. We will conduct electrophysiological recordings in the early visual cortex while presenting stimuli with varying levels of uncertainty to further develop and test this theory.
This aim will provide strong training in computational neuroscience and the design of experiments with a theory-driven approach. Second, we will identify how downstream circuits decode stimulus uncertainty for perception (Aim 2), by recording neurons from animals trained to judge the perceptual reliability of noisy sensory stimuli. Achieving this aim will require recording from large neuronal populations across multiple areas simultaneously, so will provide training in cutting-edge, large-scale electrophysiological techniques. Finally, we will connect the insights gleaned from neurophysiological experiments in nonhuman primates to human behavior through a novel approach to eliciting perceptual confidence (Aim 3). This approach will involve training in and application of concepts from value-based decision-making to rigorously isolate confidence behavior from the potential influence of reward preferences, such as risk attitude. These experiments will be vital as a potential foundation for exploring the possible disruption of uncertainty encoding in neuropsychiatric disorders. Overall, results from this proposal will help to reveal the role of sensory circuits in representing and utilizing uncertainty to guide visual perception.
Observers act with seeming knowledge of their own sensory uncertainty when making perceptual decisions as well as reflecting on the accuracy of those decisions, but the neural basis of this ability is unknown. This proposal develops and tests the predictions of a new theory for the encoding of stimulus uncertainty by populations of sensory neurons. By combining theoretical neuroscience, neurophysiology, and novel behavioral procedures, this proposal will help to uncover how sensory circuits guide visual perception in an uncertain world.