Influence of truthful and misleading instructions on statistical learning across the autism spectrum

Abstract Bayesian studies of perception have documented how the brain learns the statistics of a new environment and uses them to interpret sensory information. Impairments in this process have been hypothesised to be central to autism spectrum disorders. However, very few such studies have differentiated between implicit and explicit learning. We manipulated the instructions given before a cue-stimulus association task to investigate their effects on statistical learning. The task was conducted online, in 335 participants with varying autistic traits. In the implicit condition, where no information was provided, participants acquired weak prior beliefs about the task regularities. Conversely, explicit information about the presence of regularities resulted in strong priors, correctly reflecting the task’s statistics, regardless of the information’s veracity. Contrary to our hypothesis, autistic traits did not correlate with the influence of priors in any condition. Author Summary Perception is greatly influenced by the brain’s prior knowledge of the environment, through a process called Bayesian inference. Recent theories of psychiatric disorders and particularly autism view them as impairments in this process. A crucial aspect of this process is how individuals form their knowledge of the environment. However, previous studies have not differentiated between learning that occurs when participants are aware of what they are learning and learning that happens implicitly. In the present study, we conducted an experiment with four conditions, each varying in terms of what participants were trying to learn and whether they were aware of the general form of the regularities. Our findings revealed that participants form much stronger beliefs about the regularities when they are informed about their presence. Additionally, we discovered that participants with strong autistic traits did not differ in their beliefs about the regularities in any of the conditions. Competing Interest Statement The authors have declared no competing interest. Footnotes Nikitas Angeletos Chrysaitis: s1791272{at}ed.ac.uk New model comparison, including a static model and new analysis. The new static model was found to be best fitting in all conditions. In this new analysis, autistic traits did not correlate with the influence of priors in any condition.