Instructional control of reinforcement learning: A behavioral and neurocomputational investigation