Do autistic individuals show atypical performance in probabilistic learning? A comparison of cue-number, predictive strength, and prediction error.

The study used adults recruited online, and the researchers could not verify whether autistic participants had a clinical autism diagnosis. The online sample was likely well-educated and had typical intelligence, so the results may not apply to children, people with intellectual disability, or autistic adults with different education levels. The weakly predictive multi-cue group difference appeared only in the late phase of a task with 16 associations and 8 repetitions per association, so it may reflect the task design rather than a stable learning difficulty. In the single-cue task, autistic adults discriminated x/o stimuli better than non-autistic adults, and perceptual ability was related to learning accuracy, which may have affected the lack of group difference. The study did not test therapy, real-world social communication, or clinical outcomes, and the authors state that a direct link to social cognition needs future research.

The supplied text lists funding from H2020 Marie Skłodowska-Curie Actions and HORIZON EUROPE European Research Council. It does not report author conflicts of interest or a statement that funders designed or wrote the manuscript.

Do not conclude that autistic adults have a general probabilistic learning difficulty. The two groups learned strongly predictive associations similarly; the group difference appeared only late in the weakly predictive multi-cue task, and the reinforcement learning result was about how feedback errors were weighted, not a tested therapy outcome.