The implementation of reinforcement learning (RL) theory in the brain is discussed. In particular, how cortical regions represents value function. We consider positive (in the case of reward) and negative (in the case of punishment) values to be calculated in different ways. Orbitofrontal cortex is known to calculate value of both signs. We consider that negative value calculation don’t use RL methods. We propose the algorithm of amygdala participation in negative value calculation. In the whole, full algorithm of state-value calculation include reinforcement learning in the case of reward and learning by means of amygdala influence in the case of punishment.