Comparing activations in artificial control networks with brain activity in real animals enables long-standing theories of motor control at the level of neural circuits to be probed.