Yuxuan Dai¹, James Cooke¹, Ilona Bloem^1,2, Sam Ling³, Janneke Jehee¹; ¹Donders Institute for Brain, Cognition and Behavior, Radboud University, ²Netherlands Institute for Neuroscience, ³Boston University

Percepts of our visual environment are uncertain, due to noise and ambiguity in our sensory signals. While previous work has shown that this uncertainty can be extracted from cortical activity as the width of a probability distribution, these studies mainly focused on spontaneous fluctuations in uncertainty. Here, we ask whether experimentally manipulated sources of noise, such as changes in stimulus contrast, are similarly reflected in visual cortical activity. Participants viewed gratings of random orientation (0-179 degrees), while their brain activity (areas V1-V3) was measured with fMRI. Observers reported the orientation of each grating on a trial-by-trial basis. Stimulus contrast was varied to manipulate uncertainty (contrast levels: 1%, 5%, 10%). Using a generative model-based analysis (van Bergen et al., 2015; van Bergen & Jehee, 2021), we extracted from each trial of cortical activity the posterior distribution of stimulus orientation. The width of the distribution was taken as a measure of uncertainty on that trial. We found that the degree of uncertainty about orientation in visual cortical activity was reliably greater for lower levels of stimulus contrast. Moreover, these contrast-related changes in decoded uncertainty predicted the observer’s behavior: behavioral orientation judgments were less precise when stimulus contrast was lower, and the width of the decoded distribution was broader. Interestingly, the decoded distribution also tracked spontaneous trial-by-trial fluctuations in the precision with which the viewed orientation was encoded in cortical activity. Further analyses revealed that these flexible across-noise adjustments in uncertainty appear to rely on mechanisms of neural gain. Taken together, our results suggest that the neural code for uncertainty is remarkably robust across diverse sources of noise – whether that noise arises from the physical environment or is intrinsic to the observer.