How contextual information modulates eye movements during natural sequential behavior in a dynamic scene.

Our motor system has a remarkable ability to generate precise motor action and the flexibility to adjust in changes in the environment. In conditions where inflexibility in the motor domain has been reported, such as in Autism Spectrum Disorders, motor failures and sensory overload are exacerbating. Here we used a naturalistic interception task to investigate the interplay between autistic traits and the use of contextual information during natural sequential behavior in a dynamic scene. Participants engaged in a modified version of the 'Pong' video game, where they were asked to maintain a continuously moving ball within the game field by striking it with the paddle to prevent it from traversing beyond bounds. By manipulating the speed of the ball (fast vs slow) trial by trial, we demonstrated that participants scored low for autism were able chose the most efficient eye movement strategy, according to the ball physics. However, participants reporting higher autistic symptoms showed an inflexibility in accurately switching eye-movement strategy according to contextual information provided: on one hand, accuracy in adjust prebounce predictions of the ball’s post-bounce trajectory were negatively correlated with the autism symptomatology; on the other hand, more frequent pursuit eye movements for the moving object, irrespective of the ball speed, were linked to more pronounced autistic symptoms, which have been interpreted as indices of gaming appraisal. Finally, pupil diameter, a remote indicator of load that is strongly associated to subjective gaming experience, varied with the autistic score of our participants: when predictive saccades were elicited, the control group showed larger pupil dilations compared to the high AQ group, presumably support a higher performance in the game. Overall, our results support atypical sensorymotor inflexibility in high autistic symptomatology that may be influenced by differential use of relevant contextual information in motion prediction.