ABERRANT NEURAL ADAPTATION TO EXPECTED UNCERTAINTY IN ADULTS WITH AUTISM SPECTRUM DISORDER

The ability to adjust brain resources to manage expected uncertainty is hypothesized to be impaired in autism spectrum disorder (ASD), though the evidence remains limited. To investigate this, we studied 23 neurotypical (NT) and 23 high-functioning adults with ASD performing a probabilistic two-alternative value-based task while undergoing magnetoencephalography (MEG) and pupillometry. The task comprised five sequential blocks with stable reward probabilities (70%:30%), but varying stimulus pairs and reward values, enabling the assessment of behavioral and neural adaptation to expected uncertainty. We analyzed the hit rate of advantageous choices, response times, and computational measures of prior belief strength and precision. To examine cortical activation during decision-making, we used magnetoencephalographic (MEG) source reconstruction to quantify α-β oscillation suppression in decision-relevant cortical regions within the pre-decision time window. Linear mixed models were applied to assess trial-by-trial effects.

Behaviorally, ASD participants exhibited lower overall belief precision but intact probabilistic rule generalization, showing gradual performance improvement and strengthening of prior beliefs across blocks. However, unlike NT individuals, they did not show a progressive downscaling of neural activation during decision-making or a reduction in neural response to feedback signals as performance improved. Furthermore, on a trial-by-trial basis, increased belief precision in ASD was not associated with reduced cortical activation, a pattern observed in NT individuals.

These findings suggest an atypically rigid and enhanced allocation of neural resources to advantageous decisions that individuals with ASD rationally judge as optimal. This pattern may reflect an aversive response to the irreducible uncertainty inherent in probabilistic decision-making.