This paper synthesizes findings and theoretical propositions across behavioral, cognitive, and neuropsychological theories, with significant new conceptualizations bearing upon processes of learning and performance. There is a need to explain ultrarapid learning within the framework of cognitive science. In video self modeling and in challenging circumstances, the speed of behavior change appears to be derived from feedforward, in which component behaviors (in the repertoire) are reconfigured to produce a new skill or level of performance. It is argued that 'self modeling' is fundamental to learning, and peer/other modeling serves as an alternative. Learning in this way produces a cognitive self-simulation which can be accessed to trigger a behavioral response in a future context. Related neurological processes are indicated by 'mental time travel' (MTT) and specific brain activity during the imagination (simulation) of future personal events. There is evidence that some brain mechanisms (mirror neurons), involved in immediate imitation, are differentially responsive to images of self versus other. MTT (to future events) in cognitive neuroscience has so far been discussed only in terms of prediction and planning not behavior change. These issues are brought together by self model theory. Conclusions drawn in this paper include discussions of the value in 'learning from the future' as a ubiquitous human ability. Overall, the propositions of this theory should stimulate diverse future research, linking neurological and behavioral contributions to cognitive science. WIREs Cogn Sci 2012, 3:215-230. doi: 10.1002/wcs.1156 For further resources related to this article, please visit the WIREs website.