Human probability learning in changing environments relies on distinct pupil dilation patterns that reflect internal computations of uncertainty, according to a computational study conducted by Aaron Greenhouse-Tucknott, Cedric Foucault, Anne Buot, and Florent Meyniel. By tracking arousal-mediated changes in pupil diameter during a dynamic estimation task, the researchers investigated how the brain represents and responds to statistical shifts when changes are difficult to detect.
In the experiment, human participants completed a probability learning task where they estimated a hidden generative probability of stimuli that changed abruptly and unpredictably over time. To analyze learning dynamics, the team applied a Bayesian ideal observer model to estimate trial-by-trial learning rates alongside two specific uncertainty-associated factors: change point probability, defined as the likelihood that the generative process had abruptly shifted, and prior uncertainty regarding current beliefs about that process.
The behavioral data showed that participants relied on prior uncertainty to adjust their learning rates in line with normative theory. Pupillometric analyses revealed that the nervous system organized these computational signals into temporally distinct physiological representations. Specifically, rapid phasic pupil dilations tracked change point probability during stimulus presentation, while slower baseline tonic pupil dilations tracked the participants’ accumulated prior uncertainty.
Statistical mediation analyses further indicated that both phasic and tonic pupil signals partially carried the respective effects of change point probability and prior uncertainty on subsequent learning adjustments. These physiological signals demonstrated that autonomic arousal systems actively contribute to the translation of latent statistical assessments into measurable updates in human belief formation during volatile learning conditions.
Critically, the investigators discovered that high tonic arousal weakened the coupling between phasic pupil responses and change point probability, which was accompanied by a diminished influence of change point probability on learning updates. The authors concluded that probability learning in volatile settings is underpinned by computationally rational integration of latent uncertainty factors regulated through distinct arousal-associated physiological responses.
Subject of Research: Pupillary representations of uncertainty during dynamic probability learning in humans
Article Title: Pupil dynamics reflect uncertainty-driven adjustments of probability learning
Article References: Greenhouse-Tucknott, A., Foucault, C., Buot, A., & Meyniel, F. (2026). Pupil dynamics reflect uncertainty-driven adjustments of probability learning. PLOS Computational Biology, 22(11), e1014064. https://doi.org/10.1371/journal.pcbi.1014064
Image Credits: AI Generated
DOI: 10.1371/journal.pcbi.1014064
Keywords: pupillometry, probability learning, uncertainty, change point probability, prior uncertainty, Bayesian ideal observer model, learning rate, tonic arousal, phasic pupil dilation, tonic pupil dilation, belief updating, cognitive computational neuroscience, arousal systems, dynamic environments, mediation analysis, normative theory
Cite Scienmag News
Drew Townsend. (November 20, 2026). Pupil Dilation Patterns Track Uncertainty During Human Probability Learning. Scienmag. https://scienmag.com/pupil-dilation-patterns-track-uncertainty-during-human-probability-learning/
Drew Townsend. "Pupil Dilation Patterns Track Uncertainty During Human Probability Learning." Scienmag, 20 November 2026, https://scienmag.com/pupil-dilation-patterns-track-uncertainty-during-human-probability-learning/. Accessed 20 November 2026.
Drew Townsend. "Pupil Dilation Patterns Track Uncertainty During Human Probability Learning." Scienmag. November 20, 2026. https://scienmag.com/pupil-dilation-patterns-track-uncertainty-during-human-probability-learning/








