Autistic traits relate to speed/accuracy trade-off but not statistical learning and updating.

Cognitive and social alterations characterize Autism Spectrum Disorder (ASD), yet comprehensive explanations are challenged by ASD's heterogeneity. One candidate framework is predictive processing, which posits that predictive processes are altered in ASD (e.g., slower internal model updating). We tested this framework using the spectrum approach, which suggests that subclinical autistic traits are continuously distributed in the general population, with most diagnosed individuals above a threshold. We recruited neurotypical adults (N = 296) to examine the relationship between autistic traits and predictive processing. Using an implicit statistical learning task, we tested model updating in an unsupervised, ecologically valid manner, and assessed speed/accuracy trade-off to control for potential visuomotor performance confounds in ASD. We found no difference in model updating rate along autistic traits, suggesting no relationship between these traits and model updating in the general population, contrary to the slow updating hypothesis. However, our results reveal a difference in the evolution of speed/accuracy trade-off along the degree of autistic traits, potentially indicating a shift in the balance of goal-directed and habitual systems related to autistic traits. These findings set the stage for further research on the interaction between executive functions and predictive, habitual processes related to autistic symptoms.