Altered category learning and reduced generalization in autistic adults.
Autistic adults learn categories more slowly and extend them less easily, especially when they dislike uncertainty.
01Research in Context
What this study did
Van Overwalle et al. (2025) asked autistic and neurotypical adults to learn new shape categories.
The task looked simple: press left or right when you see a shape. But the shapes shared hidden rules.
Researchers also tracked brain waves to see how each group processed the pictures.
What they found
Autistic adults needed more trials to sort the shapes correctly.
When new shapes followed the same rules, they generalized less often.
Slower learning lined up with higher scores on an intolerance-of-uncertainty scale.
How this fits with other research
Laurie-Masi et al. (2022) saw a similar slow-down. Their autistic adults learned cue-outcome links fine, but lagged when the links flipped. Together the studies say: changing rules, not basic learning, is the hurdle.
Goris et al. (2021) seems to disagree. They found that neurotypical people with high autistic traits did worse on a volatile reward game. The twist: they studied traits, not diagnosis. Mixing trait and clinical samples explains the clash.
Beaurenaut et al. (2024) add nuance. Autistic adults showed normal social learning online, but anxiety and autism severity weakened performance in women. Learning differences in autism often hinge on extra factors like sex or mood, not autism alone.
Why it matters
When you teach new concepts, expect autistic adults to need more examples and clearer signals. Build extra practice for rule changes. Check intolerance of uncertainty; if it is high, preview upcoming shifts and give wait-time before answers. These small tweaks can cut frustration and speed mastery.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Add five extra exemplars and one probe set when you teach a new sorting or matching skill.
02At a glance
03Original abstract
Individuals with autism spectrum condition (ASC) are suggested to experience difficulties with categorization and generalization. However, empirical studies have mainly focused on one process at a time, and neglected underlying neural mechanisms. Here, we investigated categorization and generalization at a behavioral and neural level in 38 autistic and 38 neurotypical (NT) adults. By presenting shapes sampled from an artificial multidimensional stimulus space, we investigated (1) behavioral and neural underpinnings of category learning and (2) behavioral generalization of trained categorization to both an extended version of the stimulus space and a novel stimulus space. Our previous findings showed that individuals with autism were slower in category learning. In this study, we demonstrate that this slower learning in autism was not related to differences in applied categorization strategy. In contrast, electroencephalography recordings during training did reveal a reduced amplitude of the N1 component in the right occipital temporal cortex after stimulus presentation in autistic participants, which suggests atypical categorical proficiency. In addition, we observed delayed and higher activation in the frontal regions after receiving (negative) feedback in the autistic group, potentially suggesting more explicit feedback processing or a higher salience of prediction errors in autism. Finally, autistic and NT individuals were able to generalize their learned categorization after training. However, when generalizing to a novel set of shapes, autistic individuals were significantly less accurate. Reduced generalization significantly correlated with increased intolerance to uncertainty scores. This multi-level approach reveals behavioral and neural differences in learning and generalization that could be related to clinical symptoms in autism.
Autism research : official journal of the International Society for Autism Research, 2025 · doi:10.1002/aur.3286