Autism is diagnosed on the basis of a triad of impairments in social interaction, communication, and flexible imaginative functions (with restricted and repetitive behaviors and interests; RRBIs). There has been a strong presumption that these different features of the syndrome are strongly intertwined and proceed from a common cause at the genetic, cognitive and neural levels. In this review we examine evidence for an alternative approach, considering the triad as largely 'fractionable'. We present evidence from our own twin studies, and review relevant literature on autism and autistic-like traits in other groups. We suggest that largely independent genes may operate on social skills/impairments, communication abilities, and RRBIs, requiring a change in molecular-genetic research approaches. At the cognitive level, we suggest that satisfactory accounts exist for each of the triad domains, but no single unitary account can explain both social and nonsocial features of autism. We discuss the implications of the fractionable-triad approach for both diagnosis and future research directions.