³³In fact, it is often difficult to define what is the actual reward and what is differences in state-values. I’m here assuming that the sweet taste is a reward in itself, and not a question of a high state-value (i.e. predicted future reward), but this can be disputed. It is less controversial that an Olympic medal does not produce a reward in itself, but even this is not so clear. To solve this problem, Singh et al. (2009) propose that the rewards should evolve so that they are correct in most environments, while state-values are then learned during an individual’s lifetime for the particular environment where the individual is living.