Autism is generally accepted as the most genetic of all the developmental neuropsychiatric syndromes. However, despite more than several decades of genetic study, the etiology of autism remains unknown, largely due to the genetic and phenotypic diversity, or heterogeneity, of this disorder, and the lack of biologically based classification systems. At the same time, in the neuroimaging literature, the body of research identifying candidate neural systems underlying aspects of autistic impairment has grown considerably, fueled by the advent of technologies such as functional magnetic resonance imaging (fMRI). Yet the findings from these neuroimaging studies have not been incorporated to inform the collection of samples for genetic studies of autism, which are predominantly based on a diagnosis of the disorder. This article presents a review of the genetics of autism and describes the genetic approaches that have been applied, including the phenotypic strategies that have been used to address heterogeneity and optimize the power of these genetic studies. With the increasing recognition that there may be different "autisms" (Geschwind and Levitt, 2007) with unique neural mechanisms, it is argued that neural systems research, using technologies such as fMRI, currently allows for the identification of more biologically informative phenotypes for genetic studies of autism and is positioned to identify informative neuroimaging markers for "neurogenetic" studies of the disorder. To illustrate this, we describe several candidate neural systems for the social communication impairment seen in autism, and the characteristic behavioral and physiological manifestations associated with these that could be incorporated into phenotypic assessments.