The mammalian brain is equipped with reward-harvesting mechanisms that efficiently categorize and value the behavioral choices that lead to rewards necessary for survival. In this context, 'efficiency' embodies the idea of achieving maximum returns for minimal energetic investments and places a premium on how an animal represents its available options. But the capacity to efficiently represent choices is a profoundly difficult problem because representations for behavioral choice depend intimately on the statistics of information arriving not just from the sensory world and but also from within the creature itself. Any complete account of decision-making in mammals must efficiently connect the internal needs to the perceptual apparatus available to a creature moment-to-moment.