I'm sorry, HAL. Robot movements don't compute in our premotor cortex.

Don't Imitate the Robot

Being able to mimic the actions of others is an important skill, and not just for budding ballerinas and baseball players. It was critical for the development of tool use in humans. Research reported this week in Current Biology suggests that our ability to imitate others is hard-wired into our brains, but the use of tools isn't.

When we see someone else make a move--such as picking up their glass--our brains unconsciously prepare us to imitate. In a kind of mental short-cut, a group of brain cells in the premotor cortex are thought to encode the observed action so that our bodies can be immediately commanded to perform it. These so-called mirror neurons were first discovered in monkeys where they not only fire up when a monkey grasps an object, but also when it merely observes a human being grasp an object. However, when a monkey watches a human using a pair of pliers to grasp an object, the mirror neurons remain inactive, suggesting a hard-wired preference for imitating biological grasping movements.

Unlike monkeys, we humans have evolved as tool-users. So to test whether humans perceive nonbiological movement differently, Umberto Castiello, a neuroscientist at Royal Halloway, University of London, and colleagues used positron emission tomography (PET) to monitor the activity of people's mirror neurons while watching someone either grasp a cylinder or manipulate a robotic arm to do so. The robotic arm was dressed and gloved to look human, although the operator was visible to the subjects in the experiment.

Like those of monkeys, human mirror neurons lit up when a person grabbed the cylinder. But when the robotic arm was doing the grasping, the neurons remained indifferent. In spite of our tool-using ways, our mirror system, like that of monkeys, apparently only keeps track of limb movements, not tool movements. The ability to imitate tool use, therefore, must rely on other brain regions. Besides helping us to understand how action is encoded in the brain, Castiello says the finding should provide insight for engineers trying to design humanoid robots.

"The study is important," says Patrick Haggard, a cognitive neuroscientist at University College London, because it shows that our unconscious brains seem to actively distinguish between actions made by living beings and those made by a machine. Haggard would like to see the team do experiments to tease apart what parameters--the steadiness of hand velocity, for example--the brain uses to decide whether a real person is doing the moving.

Related sites
Castiello's home page
About Mirror Neurons