Virtual Reality-Social Cognition Training (VR-SCT) is a behavioral training program designed to improve social communication skills in young adults with ASD and has been empirically validated to be effective with behavioral evidence. In this project, neuroimaging-based discoveries will be applied to promote the development of VR-SCT into a more effective intervention method. Specific Aims: (1) To demonstrate clinical efficacy of a VR-SCT training program versus active control in a randomized, controlled trial; (2) To identify pretreatment brain characteristics that are associated with greater behavioral response to VR-SCT, relative to active control; (3) To validate neuroimaging-based measures as sensitive outcome measures of treatment response to VR-SCT. Methods: A 5-week randomized controlled trial of VR-SCT is proposed to actively treat social communication deficits versus an active control condition, in 48 cognitively-able (IQ ? 80) young adults (18-35 years of age) with ASD and will perform behavioral, Magnetic Resonance Imaging (MRI), and Electroencephalography (EEG) testing before and after the intervention. The active control group will be offered a similar kind of VR-based computer game with online interactions (“Minecraft”) for matched sessions. Expected results: (1) adults with ASD in the active treatment (vs. control) group will show greater behavioral gains in social skills from baseline to treatment endpoint; (2) The pretreatment neural level of activation, pattern of activation, or functional connectivity of brain regions implicated in theory of mind and facial affect processing, or cortical thickness of key social brain regions, or white-matter fiber tract integrity, will be positively associated with the rate of response to VR-SCT relative to the active control; (3) adults with ASD in the VR-SCT relative to the active control group will exhibit enhanced neural activity in the brain networks implicated in theory of mind or facial affect processing from baseline to treatment endpoint.