It is a reward signal limited to the sense of “reward for having made a correct prediction”. This is, indeed, the kind of reward needed for reinforcement learning. In that case (for example, in songbird, recent paper: http://science.sciencemag.org/content/354/6317/1278?ijkey=9dfV3EVHS3nbA&keytype=ref&siteid=sci) instead of a “prediction” you have a correct behavioural outcome, and you are rewarded if the action you learn corresponds to it. So it is a reward in this sense. But great article!