Abstract The earliest observable symptoms of autism spectrum disorders (ASDs) involve motor behavior. There is a growing awareness of the developmental importance of impaired motor function in ASD and its association with social skill. Compromised motor function requires increased attention, leaving fewer resources available for processing environmental stimuli and learning. This knowledge suggests that the motor system—which we know to be trainable—may be a gateway to improving outcomes of individuals living with ASD. In this review, we suggest a framework borrowed from machine learning to examine where, why, and how motor skills are different in individuals with ASD.