Working in Movement

The University of Pittsburgh has trained a monkey to feed itself using a prosthetic arm. What makes this interesting is the arm is not attached to the monkey, and the monkey controls the arm with its thoughts alone. Wired News: Advent of the Robotic Monkeys sums up the project, but gives no details on whether the monkey paid in-state or out-of-state tuition.

Though monkeys have been trained to move prosthetic arms before (see More on Thought Controlled Movement), this experiment used a "closed loop" approach in which the monkey is conscious of the robotic arm and is making an effort to control it.

The unique aspect of Schwartz's research is that he conducted what is known as "closed loop" brain experiments. In a "closed loop" experiment, the monkey is conscious of the robotic arm and is making an effort to control it. Monkeys in previous experiments did not understand that they were having an effect on the world at all. Duke University performed such prosthetic arm experiments as far back as 2000. In one case they even sent the electrode signals over the internet, allowing the monkey to move an arm 600 miles away at MIT.

"The open loop experiment was really very crude," said Schwartz. "The closed loop introduces us into a whole new field because the animal actually sees the arm and the consequence of what it is doing." For Schwartz's monkey the robotic arm is incorporated into its mental body representation, making it an extra limb.

In the experiement, the monkey learned (though biofeedback and virtual reality) to incorporate the arm into its body image so that it can control it with thought. What makes this difficult is that the fake arm doesn't "feel" like a real arm. This is because it takes many fewer neurons to move the fake arm than the real arm, so the new, unfamiliar movement needed to be learned.

In the virtual space the monkey learned through biofeedback how to modify the firing rates of the neurons that are being recorded and sent to the robotic arm for directions. By the end of its "brain control" lessons the monkey mastered this new form of movement and could control its phantom limb in virtual reality by knowing how to fire the few key neurons needed.

Like most learned movements, there is room for improvement in the feeding gesture.

"The initial movement to the mouth is pretty good, but when it gets to his mouth he is concentrating on the food and not on the arm movements so it gets a little clumsy," said Schwartz.

In fairness to the monkey, the clumsiness is at least partially related to the design of the arm, with its simple one-movement tool at the end.

One of the interesting things about this is the idea of mapping an external device into the body image, and then using thought to initiate and control movement. But I wonder what part sensing plays in all this. After all, it might seem that a device not attached to the body could not be felt or sensed. And that might be accurate if the sensing function depended solely on the muscular/skeletal system. It doesn't. The nervous system integrates lots of complex stuff together to make movement seem like it's one smooth action. So I suppose that the monkey somehow does learn to sense the arm; at any rate, the important thing is that the monkey can use the arm to get food.