I suggest that the current study of autism is problematic, due to: (1) its failure to pursue a medical model of disease causation, with protocols for differential diagnoses of causes; (2) a notable incidence of unrecognized false positive diagnoses in children; (3) the conceptual equating of autism with sets of traits that have been shown to be genetically and phenotypically unrelated to one another; and (4) the expansion of use of the terms “autism” and “autism traits” to psychiatric conditions that have no substantive etiological or symptomatic overlap with autism. These problems can be alleviated by, like Kanner, considering autism as a syndrome, a constellation of traits, conceptualized as differences rather than deficits, some set of which is found in each affected individual to some degree. The original, prototypical form of autism can be delineated based on the “hallmarks” of autism: a set of core traits, originally explicated by Kanner, that defines a relatively-homogeneous group, and that connects with the larger set of autism symptoms. The hallmarks of autism provide a touchstone for research that is unambiguous, historically continuous to the present, and linked with major theories for explaining the causes and symptoms of autism. Use of the hallmarks of autism does not impact recognition and treatment of individuals with DSM diagnosed autism, or individuals with the many disorders that involve social deficits. This perspective is compatible with the research domain criteria approach to studying autism, via analyses of autism's constituent traits and the differential diagnosis of its individual-specific causes.