ABS 015: On-Line Pupillometric Indices of Perceptual Learning

Soph Bililies ¹ , Aaron Cochrane ¹ , Takeo Watanabe ¹ , Yuka Sasaki ¹

¹ Brown University

Van Wickle (2025) Volume 1, ABS 015

Introduction: Our cognitive neuroscience research investigates the neural and behavioral dynamics underlying perceptual learning—how the brain processes and adapts to visual information to acquire new skills. By examining the interplay between feedback, perceptual sensitivity, and pupil-linked arousal systems, this study seeks to uncover mechanisms that enhance learning and generalization efficacy.
We employ pupillometry, a non-invasive proxy for noradrenaline function, alongside dynamic behavioral modeling to explore how the brain amplifies its own critical signals ("gain") to support perceptual improvements. Training participants across six days on a visual motion discrimination task, we assessed how varying feedback conditions—ranging from none to dense or initial —affect learning outcomes and pupillary responses. Results reveal that none or minimal feedback will still facilitate self-supervised learning, with participants improving their perceptual thresholds. Notably, feedback appears to enhance the ability to allocate cognitive effort selectively, as evidenced by differential pupillary responses to task difficulty.

Our findings contribute to a growing body of research linking physiological markers, such as pupil dilation, to learning processes. By demonstrating the role of initial feedback in catalyzing self-supervised learning and identifying potential links between pupil dynamics and evidence-accumulation parameters like drift rate, this study advances our understanding of perceptual plasticity. These insights have broad implications for optimizing learning strategies, from educational settings to neurorehabilitation programs, paving the way for more effective and adaptive skill-building interventions.

Methods: Participants completed a coherent motion discrimination task over six consecutive days. On each trial, they judged the direction of motion (left vs. right) of noisy dot displays. Training consisted of four blocks of 128 trials per day (~512 trials/day, ~3,072 total). External feedback (correct/incorrect) was manipulated across groups. Performance thresholds (accuracy vs. motion coherence) and pupil dilation were recorded during stimulus presentation and response. Testing sessions were incorporated on Days 3 and 6 alongside training blocks, following a fixed training timeline.

Results: Adaptive pupil responses tracked learning in the full feedback condition, suggesting some generation of endogenous feedback. In the condition without external feedback, this association was absent or reversed, indicating that external feedback was necessary for initial learning. Critically, in the initial-only feedback condition, the association between learning and pupil response persisted even after feedback was removed, confirming that adaptive pupil responses became self-generated

Discussion: Our findings suggest that brief, targeted feedback can train learners to generate internal signals that support continued improvement, even after feedback is removed. This has important implications for education and rehabilitation. Additionally, pupil dilation maintains a promising biomarker for tracking cognitive effort, engagement, and learning plasticity. Future work could explore how to optimize feedback schedules across different populations and settings to enhance endogenous learning processes.