The Extended Reach: How the brain codes for tool use
A team of Italian neuroscientists has illuminated how the brains of monkeys adjust from grasping with hands to grasping with tools


by Jim Schnabel

April 16, 2008

From tying shoelaces to turning screwdrivers to clacking away at laptops, we humans would be lost without our tools. Our facility with tools appears even to be ingrained upon our brains: The cleverest, most nimble-thumbed monkeys can take months to learn tool-related tasks that human children somehow pick up in minutes.

Neuroscientists have long wondered how the human brain manages this. Now one group has taken a big step towards finding out. Reporting Feb. 12 in the Proceedings of the National Academy of Sciences, researchers in the laboratory of Giacomo Rizzolatti at the University of Parma in Italy suggest that the brain adjusts to the handling of some tools simply by treating them as if they were part of the body.

Monkey see, monkey do

Studying the working human brain in fine detail is difficult. Ethical considerations preclude, for example, inserting electrodes into the brain of a healthy person to study his detailed neural workings as he learns to use a tool. Researchers have had to approach the subject somewhat indirectly in three ways: by studying tool use among people with brain damage; by studying human brain activity using magnetic-resonance and other not-so-fine-grained imaging methods; and by implanting neural electrodes into monkeys to study the firing patterns of individual neurons.

In recent years, Rizzolatti’s group has arguably been unmatched in their work in the latter category. Their electrode studies suggest that muscle-related tasks are generally controlled by two categories of motor neurons. The first are the primary motor neurons that connect to individual patches of muscle. The second category appear to represent the broader goals or tasks the muscles are supposed to carry out, such as “grasp food.”

“Rizzolatti's group has previously identified neurons in the ventral premotor cortex of monkeys that code grasping actions,” explains Scott Frey, a specialist in tool-use neuroscience at the University of Oregon. “In some cases they found that single cells within this area, known as F5, code the act of grasping regardless of whether it’s accomplished with the hand or mouth.”

Iraki Intskirveli, who was part of Rizzolatti’s group until recently and now works at University of California at Irvine, says that such findings strengthened the hypothesis “that some neurons in the F5 area, instead of coding for specific muscle movements associated with action as the primary F1 motor area does, are coding for the more abstract goal of the action.”

Rizzolatti’s group tested this hypothesis by fitting macaque monkeys with electrodes that eavesdropped on their neurons in the F5 and F1 areas. The monkeys were then trained to use two kinds of pliers to grasp an object they normally would grasp by hand. The first kind of plier was the normal one in which the hand has to close in order for the plier ends to grip the object. The second was a “reverse pliers,” in which the hand has to open to allow the plier ends to grip the object. By forcing the monkeys to use two tools that involved the same abstract grasping goals, but with different sequences of muscle movements, they could answer the question—as Intskirveli puts it— “are the neurons following the muscle movements, or the abstract goals of the action?”

Rizzolatti and his colleagues found, as expected, that some of the neurons they measured in F1 fired according to actual hand movements, and so changed when the pliers were replaced by reverse-pliers. But other neurons they measured in F1, along with all the neurons they measured in F5, kept firing, and in the same sequence, regardless of whether the monkeys were grasping an object with their hand, grasping with normal pliers or grasping with the reversed sequence of muscular movements needed for the reverse pliers.

“This result elegantly demonstrates that those neuron units were coding the abstract goals of grasping, not the movements required for their implementation,” says Frey.

By maintaining a distinction between abstract goals and specific muscle actions, in other words, the brain stays flexible and goal-oriented, allowing itself to accomplish a given task in any number of ways, including by tool-use.

The brain in a sense doesn’t care whether it is manipulating a live limb or an inanimate tool, says Michael Arbib, a neuroscientist at the University of Southern California. “The monkeys eventually learned to transfer their attention from what their hands were doing to what the jaws of the pliers were doing. So the crucial point about tool use is that you can displace or extend your ‘self’ from your body ending at the end of your hand to your body ending at the end of the tool.”

What remains to be determined is how humans came to be so much more adept at this sort of self-extension than all other species. “Even in chimpanzees we do not see universal use of technology, whereas human material culture is a universal characteristic that, as far as we currently know, emerged abruptly about 2.6 million years ago in our hominid ancestors,” Frey says.

Mirroring and imitation

One clue may be the mirroring property of some primate motor-related networks, first described by Rizzolatti’s group in the mid 1990s and widely seen as a key element in the ability to learn new motor skills by imitation. These “mirror neurons” are found in F5 and other regions but are wired to sensory areas of the cortex, and become active whenever the monkey performs a given action or recognizes it being performed by another.

Arbib, Rizzolatti and others think that the mirroring property of monkey brains and the imitation it enables are keys to the evolution of more complex tool use in primates and hominids, which culminated eventually in the multi-purpose, symbol-manipulating toolkit of human language.

This hypothesis now underlies a great deal of research on tool use and imitation. In humans but not in lower primates, notes Frey, tool use is generally localized to the left hemisphere of the brain, as is also true for language areas. “And we know that when a human has an injury to the left hemisphere, he often has a language difficulty, and often too there will be a problem with these high-level motor skills, including those involving tools and doing gestures and things like that.”

Frey also points out that the F5 area of the monkey premotor cortex, where a relatively large number of goal-directing and mirroring neurons are located, shows anatomical similarities to Broca’s area, a key language-processing area in the human brain. “Many researchers,” he says, “think F5 is an evolutionary precursor to Broca’s area.”

But the “mirror system hypothesis,” as it’s known, still has some unfinished business. Functional-MRI and other imaging studies of mirror neurons in humans haven’t had the neuron-specific resolution that is possible in neural-electrode studies of monkeys, so they haven’t been able to yield much in the way of definitive data. And as Arbib notes, “there have been almost no attempts to study mirror neurons in other creatures.”

The first such attempt was reported Jan. 17 in Nature. Researchers from Duke University found what appeared to be auditory-vocal mirroring neurons in a species of songbird. The finding suggests that mirroring and the imitative potential that goes with it are not the exclusive properties of higher mammals such as primates, and may be just as highly developed in birds—as perhaps every talking-parrot owner already knew.

“It greatly complicates an evolutionary account of tool manufacture and use,” says Frey. But research continues to move forward, he adds. “It’s still early.”