Autism is diagnosed by the criteria for Autism Spectrum Disorder (ASD): impairments in social interaction and communication and repetitive and restricted behaviors. Although for over seventy years autism has been viewed as one disorder, autism as ASD is currently understood to define an undetermined set of subgroups called “the ASDs.” This paper proposes that discovery of existing individual ASD subgroups useful for translational research requires studying ASD symptoms separately because three lines of evidence demonstrate that ASD diagnostic symptoms do not identify a bounded entity. These three lines of evidence are reviewed here. One, the extent and variety of neurodevelopmental symptoms found with ASD risk factors do not support the assumption that ASD diagnostic symptoms isolate a natural convergence of two of the many symptoms. Two, the ASD diagnosis lacks neurobiological validity: many varied brain dysfunctions have been found with the ASD diagnosis. Three, ASD diagnostic symptoms have no consistent empirical association with one another. This brief review of these three lines of evidence concludes that discovering subgroups of valid etiopathophysiologies requires exploring ASD diagnostic symptoms separately and forming study samples defined by an ASD risk factor, or an ASD brain dysfunction, or defined by a specific risk factor linked to a specific brain dysfunction.