First, there must be some kind of filtering process that circumscribes the kinds of thoughts that can act as a reward signal. Second, the control must be self-limiting, that is, it must be relatively short-lived.