Adults with Autism Learn a Better Bet Just as Fast, Their Pupils Pay a Different Price
- In a probabilistic reward-learning task with no volatility (one option consistently paid off more), 46 adults – half with high-functioning autism spectrum disorder (ASD), half neurotypical – learned to prefer the better option at the same rate, with no significant group difference in choice accuracy or reaction time.
- Pupil-linked arousal, a physiological marker tied to noradrenergic signaling and to the brain's precision-weighting of predictions, diverged by group as learning progressed: it rose across the session in the ASD group and fell in the neurotypical group.
- Stronger pupil dilation during higher-value choices correlated with higher self-reported intolerance of uncertainty, a relationship that held across both groups rather than being ASD-specific.
- The study was run at the MEG Center, Moscow State University of Psychology and Education, using a probabilistic choice task combined with pupillometry and mixed-effects modeling – not self-report or observational rating alone.
A behavioral test can return an identical score for two groups and still hide a real difference in what it costs each nervous system to produce that score. That gap is what a new pupillometry study out of Moscow found in adults with autism spectrum disorder, and it lands on a question clinicians already face: when a client "does fine" on a task, has the underlying difficulty actually resolved, or just gone quiet.
The Choice Task Looked the Same in Both Groups
Forty-six adults – half with high-functioning ASD, half neurotypical – worked through a probabilistic reward task with no volatility: one option paid off more often than the other, consistently, across the whole session. Both groups learned to prefer the better option at the same rate, with comparable accuracy and reaction times. On the behavioral read-out alone, autism drops out of the data entirely.
Pupil Dilation Diverged as Learning Progressed
It does not drop out of the eyes. As both groups converged on the better option, pupil-linked arousal – tied in predictive-processing accounts to how much weight the brain assigns an updated prediction – moved in opposite directions across groups: climbing in the ASD group, falling in the neurotypical one. The more a choice carried higher expected value, the larger the pupil response, and the larger that response, the higher a person's self-reported intolerance of uncertainty. For practice, the separation matters more than the correlation itself: a client can hit every behavioral marker of successful adaptation while running a costlier internal process to get there, and only one of those two facts shows up without measuring it directly.
Learning to choose well and paying a physiological price for it are not the same outcome, and only one of them shows up on a behavioral checklist.
The sample is autism spectrum disorder, not depression, anxiety, or psychosis – the diagnoses this issue otherwise centers on – and the finding should not be generalized across them. The sample is small (N=46), and the link between pupil response and intolerance of uncertainty is correlational, not causal.