A major problem in understanding mechanisms of pathogenesis in autism spectrum disorder (ASD) is deciphering how risk factors act via the brain to influence the behavioral symptoms of this disorder. We may start to bridge this gap in our understanding by systematically examining the structure and function of cell types that make up circuits underlying behavioral endophenotypes in animal models for ASD. A confluence of advances in basic behavioral neurobiology, in ASD mechanisms and animal models, and in genetic tools for imaging and manipulating brain circuits will make this possible. Through a process of elimination of candidates and comparison across models, we may hope to understand how ASD risk factors influence the development and function of neural circuitry at the level of genetically defined cell types. As an example of how this candidate circuit approach may be applied to investigating ASD, here I focus on social behavior as an endophenotype, and I discuss recent findings regarding the development and function of the oxytocin system, which is implicated in both normal social behavior and ASD pathogenesis. I stress the importance of a collaborative, multidisciplinary approach to probing candidate cell types and circuits across mouse models of ASD.