Actor-critic models of reinforcement learning in the basal ganglia: From natural to artificial rats