Local and global stimuli in reinforcement learning