Social and Nonsocial Visual Prediction Errors in Autism Spectrum Disorder.

Impaired predictive coding has been proposed as a framework to explain discrepancies between expectations and outcomes in autism spectrum disorder (ASD) that may contribute to core symptoms of the disorder. However, no eye tracking study has directly addressed this framework in the context of visual predictions of social and nonsocial stimuli. The current study used eye tracking to examine violations of learned visual associations of both social and nonsocial stimuli. Twenty-six adolescents with ASD and 18 typically developing control (TDC) adolescents completed an outcome expectation eye tracking task in which predictive cues correctly (80% of trials) or incorrectly (20% of trials) indicated the location (left or right) of forthcoming social or nonsocial stimuli. During violation trials, individuals with ASD focused their gaze relatively more often on stimuli presented on locations that violated the learned association and less often on locations that corresponded with the learned association. This finding was not moderated by stimulus type (i.e., social vs. nonsocial). Additionally, participants who looked at incorrectly predicted locations more often had significantly greater ASD symptom severity. These results are consistent with theories that characterize ASD as a disorder of prediction and have potential implications for understanding symptoms related to prediction errors in individuals with ASD. Autism Res 2019, 12: 878-883. (c) 2019 International Society for Autism Research, Wiley Periodicals, Inc. Lay Summary Individuals with autism spectrum disorder (ASD) exhibit impairments making predictions that may impact learning. In this study, we used eye tracking methodology and found that individuals with ASD were less likely to look at the predicted location when a visual routine was violated. This pattern was evident for both social and nonsocial images and was associated with greater ASD symptom severity. These findings provide additional support for predictive challenges in ASD.