Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster C in Poster Session C: Wednesday, August 5, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Dissociating different sources of uncertainty in mice

Ishan Kalburge¹, Theoklitos Amvrosiadis², Nathalie Rochefort², Máté Lengyel¹; ¹University of Cambridge, ²University of Edinburgh

Presenter: Ishan Kalburge

Adaptive decision-making requires representing uncertainty often about several latent variables. While previous studies suggest that animals might represent uncertainty, it remains unclear whether such representations extend beyond task-relevant decision variables to include auxiliary perceptual variables. Here, we test whether mice represent both perceptual and decision uncertainty in a Go/No-Go orientation discrimination task while recording from area V1. We independently manipulated decision uncertainty (via stimulus orientation) and perceptual uncertainty (via stimulus contrast and dispersion). We then developed a Bayesian ideal observer (BIO) model that estimates trial-wise posterior distributions over stimulus orientation and choice. The model, fit to individual mice, outperformed both stimulus-based and behavior-only baselines in predicting choices. Estimated posteriors revealed dissociable structure: decision uncertainty was more strongly modulated by orientation, while perceptual uncertainty was driven by contrast and dispersion. These results provide evidence that animals might track multiple sources of uncertainty, and establish a framework for linking behaviorally inferred uncertainty to neural population activity. This approach enables direct tests of whether and how cortical circuits encode uncertainty during complex decision-making.

Topic Area: Computational Models of Vision & Visual Cortex