Although accumulated evidence has demonstrated that autism is found with many varied brain dysfunctions, researchers have tried to find a single brain dysfunction that would provide neurobiological validity for autism. However, unitary models of autism brain dysfunction have not adequately addressed conflicting evidence, and efforts to find a single unifying brain dysfunction have led the field away from research to explore individual variation and micro-subgroups. Autism must be taken apart in order to find neurobiological treatment targets. Three research changes are needed. The belief that there is a single defining autism spectrum disorder brain dysfunction must be relinquished. The noise caused by the thorny brain-symptom inference problem must be reduced. Researchers must explore individual variation in brain measures within autism.