Most behavioral training regimens in autism spectrum disorders (ASD) rely on reward-based reinforcement strategies. Although proven to significantly increase both cognitive and social outcomes and successfully reduce aberrant behaviors, this approach fails to benefit a substantial number of affected individuals. Given the enormous amount of clinical and financial resources devoted to behavioral interventions, there is a surprisingly large gap in our knowledge of the basic reward mechanisms of learning in ASD. Understanding the mechanisms for reward responsiveness and reinforcement-based learning is urgently needed to better inform modifications that might improve current treatments. The fundamental goal of this review is to present a fine-grained literature analysis of reward function in ASD with reference to a validated neurobiological model of reward: the 'wanting'/'liking' framework. Despite some inconsistencies within the available literature, the evaluation across three converging sets of neurobiological data (neuroimaging, electrophysiological recordings, and neurochemical measures) reveals good evidence for disrupted reward-seeking tendencies in ASD, particularly in social contexts. This is most likely caused by dysfunction of the dopaminergic-oxytocinergic 'wanting' circuitry, including the ventral striatum, amygdala, and ventromedial prefrontal cortex. Such a conclusion is consistent with predictions derived from diagnostic criteria concerning the core social phenotype of ASD, which emphasize difficulties with spontaneous self-initiated seeking of social encounters (that is, social motivation). Existing studies suggest that social 'wanting' tendencies vary considerably between individuals with ASD, and that the degree of social motivation is both malleable and predictive of intervention response. Although the topic of reward responsiveness in ASD is very new, with much research still needed, the current data clearly point towards problems with incentive-based motivation and learning, with clear and important implications for treatment. Given the reliance of behavioral interventions on reinforcement-based learning principles, we believe that a systematic focus on the integrity of the reward system in ASD promises to yield many important clues, both to the underlying mechanisms causing ASD and to enhancing the efficacy of existing and new interventions.