In a study that brings to mind any number of science fiction films, researchers have taught monkeys to feed themselves using a robotic arm, controlled by the monkeys' brain power alone. This is the first experiment to demonstrate the use of a brain–machine interface for a practical task; until now such studies have been limited to moving a cursor across a screen. The results suggest that sophisticated prosthetic devices that can be controlled as naturally as normal limbs may one day be feasible for amputees and people suffering from paralysis.

Andrew Schwartz and colleagues of the University of Pittsburgh and Carnegie Mellon University (PA) implanted grids of microelectrodes in the brains of two rhesus macaques, in the region of the cortex known to control arm movement (Nature, published online 28 May 2008; doi:10.1038/nature06996). The electrodes transmitted monkeys' cortical signals, which were translated into the motions of a humanlike robotic arm.

The researchers trained monkeys to use the robotic arm to reach for a treat, grip it, bring it to the mouth and release it. To become familiar with the robot's movements, monkeys first learned to control it with a joystick. Monkeys' arms were then gently restrained so that they would not grab the food manually, and scientists taught them to manipulate the robot using their 'thoughts'. When training began, many of the robot's motions were automated, and this assistance was gradually reduced until monkeys controlled the arm independently.

Rapid computing of monkeys' brain signals enabled fluid and natural control of the prosthesis. Monkeys seemed to interact with the arm as if it were their own, discovering how to avoid obstacles and even uncovering certain features of which scientists were previously unaware (for example, certain gooey treats would stick to the arm, so monkeys didn't bother gripping them when retrieving them to their mouths). Additionally, monkeys seemed to carry out normal motor functions such as chewing and moving their heads without disrupting the motion of the machine.

There are still many challenges to overcome before such brain–machine interfaces are practical in humans. First, implantable electrodes are still unreliable for the long term; in fact, signal power in one of the monkeys in this study faded before the experiment's conclusion. In addition, the available technology for operating such devices is too cumbersome to transport. Still, despite these limitations, this proof-of-principle study suggests a bright future for human prostheses.