Duke University Alumni Magazine


Brain chips: neurobiologist Nicolelis examines a specially designed probe to record animal brain-wave activity
Photo: Chris Hildreth

By sifting through storms of neural signals, neurobiologists are making discoveries that not only yield potential practical benefits but also challenge old notions of the brain as a static, passive computer.

he rat was thirsty. Whiskers quivering, it busily nosed its way around the test chamber, considering whether to press the small bar in the chamber with its paw. In many previous encounters, such a paw-press had triggered a little robot arm to swing into reach, bringing a sip of water. Now, for some reason unknown to the rat, the rules had abruptly changed. The bar had stopped working, so that even the most determined pressing failed to bring water. But a thirsty rat is a creative rat, and this rat had quickly learned a trick that would stun even its human handlers. This time the rat merely thought about pressing the bar, and the little robot arm whined to life, swinging about to deliver a welcome drink.

The rat's startling mental feat only brought it water, but to the scientists who devised the experiment, it has brought the extraordinary promise of new ways to give paralyzed people control over their environment. And more broadly, the rat's trick has added important support for a remarkable new theory of how the brain copes with the avalanche of data from the outside world.

Last June, neurobiologists John Chapin of Hahnemann University and Miguel Nicolelis of Duke Medical Center reported that they had used sophisticated computer analyses to distinguish brain waves emanating from the brains of laboratory rats. The particular area of the rat brain they targeted was the "motor cortex" that controls muscle movement. And the particular brain signals they detected were those controlling the sequence of muscle movements the rat used to press a bar activating a motorized robot arm to deliver water.

Once the researchers had distinguished these signals, they rigged the arm to become a brain-wave-activated "neurorobot." Their aim was to determine whether the system could read the rat's intention. And to their utter surprise, when the researchers deactivated the bar, the rats quickly learned to operate the water-giving robot without moving a muscle, but by generating only the brain waves that signaled their intent to press the bar.

"It was really quite shocking to us that the animals learned, and very quickly, that they didn't need to make the movement; they just needed to express the brain waves," says Nicolelis. "It's almost like the rat managed to dissociate the central planning and the output production, and we have no idea how that happens." What rats can do, people can do, which raises the extraordinary possibility that paralyzed humans might someday be able to control neurorobotic appendages with only brain waves.

The scientists' dramatic result is only the latest in many experiments--from operating neurorobots to tweaking rats' whiskers--in which Nicolelis and his colleagues are seeking to understand the near-magical ability of the brain to adjust itself constantly to the world around it. Their discoveries are not only yielding potential practical benefits but also a startling new paradigm for brain function, challenging old notions of the brain as a static, passive computer.

As with most advances that seem like the wildest science fiction, Nicolelis' work--supported by the National Institutes of Health and the Defense Advanced Research Projects Agency--is grounded in careful, painstaking development of a new experimental method. "It began when I was doing my postdoctoral fellowship for John [Chapin], and we decided to develop a technique nobody had yet done --recording the activity of many neurons at once." Such recording is a scientific challenge because the brain is a complex tangle of some 100 billion such neurons, delicate cable-like cells that transmit the electrical waves of nerve impulses, linking to one another to form intricate preferred circuit pathways. The brain lays down these pathways when nerve impulses in one neuron trigger impulses in its neighbors, and those connections strengthen with learning and experience.

Tapping into such brain activity was also technically daunting, since it involved an attempt to insinuate surgically an array of many hair-thin electrodes into the shifting, gelatinous brain tissue. Other researchers had failed, finding to their dismay that the electrodes tended to shift after surgery, losing their signal and slicing dangerously through the pudding-like brain tissue and damaging it.

Chapin and Nicolelis were intent on learning to use implanted electrode arrays to detect such telltale signals because they were convinced that the living brain could teach them new lessons about brain function, improving on the traditional technique of tracing the wiring of neurons in preserved brain slices. They knew that it was the brain's incessant neural activity that somehow gave rise to the near-miracle of thought. Only by recording brain activity (called "action potentials") from many electrodes at once could the neurobiologists hope to get an instant-by-instant picture that would allow them to understand a tiny element of their "thought."

The scientists hoped especially that their insights could help them confirm a rapidly evolving theory that neurons are not hard-wired circuit elements permanently assigned to one computing task, like the microprocessor inside a computer. Rather, the new theory holds that neurons are adaptable, living entities that can participate in many processing tasks at once. Moreover, the theory holds that those tasks may change from millisecond to millisecond. This idea is much like the revolution in understanding the nature of the atom, says Nicolelis. "Before quantum theory, we always thought of an electron like a little ball circling a big ball, just like a planet circling the sun. But quantum theory led us to understand that the electron is a probablistic entity. Quantum theory taught that you cannot measure an electron's exact position and velocity simultaneously, only a probability that it is at a particular point at a particular time. It is the same thing with the activity and properties of neurons. These are probablistic entities. They function in both a spatial and a temporal domain."

So the neurons that control rats' muscle movement in pressing a bar are not narrowly tuned to trigger a particular arm movement. Rather, they are widely tuned--in effect, singing little snatches of many neural tunes at once, and contributing their voices to the multiple choruses of many movements.

Spurred by the potential of such new insights, Chapin and Nicolelis succeeded, after years of development, in implanting the electrode arrays, thanks to both skill and some remarkable luck. "For example, we discovered absolutely by accident that the Teflon that we coated the electrodes with somehow stuck to the cells, so that they moved with the brain," says Nicolelis. "This made the implanted electrodes so stable that we found we could record signals from the behaving animals for weeks or months."

Such electrodes were only receiving antennae for brain signals. Another critically important development was the sophisticated signal-processing techniques that could sift through the constant cascade of brain waves to make sense of them. Nicolelis devised such statistical approaches to analyzing the signals; and to detect the signals he enlisted an artificial brain-like computer called a "neural network" to listen to the real brain. Neural networks are basically interconnected arrays of adaptive elements that function roughly like neurons, adjusting their connections to "learn." The neural network was assigned to sift through the incoming signals from the electrodes to detect particular brain signals that represented the rats' bar-pressing. "The neural network does not make any assumption about what 'code' the brain is using," says Nicolelis. "It is just looking for a pattern, so if there is some sort of statistical signature in a pattern, the network learns to recognize it."

Using these recording and analyzing techniques, the scientists made their remarkable discovery with the neurorobot-operating rats. Seeking immediately to build on that advance, they set out to refine the pattern-recognition techniques and shrink the electronics. In the first experiment, the electrode-implanted rats had to be connected via a cable to the electronics. Now, the neurobiologists are working with Duke biomedical engineers to develop a new fingernail-sized microchip that will transmit signals via telemetry, eliminating the wiring. "We believe with state-of-the-art electrophysiology and microelectronics, we can work toward something that would be clinically useful," says Nicolelis.

To take the next steps toward developing neurorobots for humans, Nicolelis has now launched experiments using owl monkeys and expects to graduate to even larger primates. Such primates, like humans, will present challenging technical problems. "For one thing, large primates' brains are more convoluted than rats' or owl monkeys'," he explains. "We don't know whether placing electrodes will be more complicated. However, because their brains are larger, we believe we can implant close to two hundred electrodes in primates' brains, versus about forty-eight in rodents. So, we can sample from more neurons and obtain better signals. We could use those signals to code for more complex movements, working toward a system that people could actually use." For his monkey experiments, Nicolelis has begun building a more complex neurorobot that moves in three dimensions, more like a human prosthetic might.

In another promising development, Nicolelis has been contacted by a British manufacturer of electric-powered prosthetic limbs. The manufacturer proposes to test whether signals from the monkeys' brains might, indeed, be able to operate such limbs.

Rat whiskers are the focus of another set of Nicolelis' experiments that have yielded dramatic proof of the brain's incredible adaptability. He and his colleagues have been tweaking rat whiskers and measuring the resulting signals generated in the brain region responsible for processing such touch data. The rat's brain devotes large areas to processing contact signals because the animal's facial whiskers are among its most important sensory organs.

In exploring the details of the brain responses, the neurobiologists first pinpointed the "receptive field" of a given neuron--which is the specific whisker or skin area that, when stimulated, leads that neuron to fire off a signal. They next measured whether the receptive field changed as the rat explored its environment using its whiskers. Traditional neural theory held that information from each whisker is represented by a specific pool of neurons responsible for detecting signals from that particular whisker. Nicolelis' whisker-tweaking studies revealed that the receptive fields of cortical neurons can shift their location on the whisker or skin in mere thousandths of a second. "These receptive fields are not stable entities; they're moving around," he says. "As the animal sweeps its whiskers back and forth, the receptive fields in the brain tend to move in the same way." Such discoveries have profound implications for understanding the brain, he says. "This dynamic behavior that we see even at the level of a single neuron may be the reason why as adults we can learn or recover from injury such as a stroke."

The realization of such instant-to-instant brain plasticity will transform our most basic perception of our own intellects, and how we cope with the world around us, says Nicolelis. "Suppose you have a neuron whose receptive field includes a spot on your fingertip. The traditional theory is that these receptive fields could be described purely in spatial terms. It was known that they may shift over time, say to compensate for injury. But, until studies like ours, nobody dreamed that a neuron's receptive field was bouncing around all the time, and that if you didn't define precisely when you were measuring it, you couldn't really tell where it was."

The "phantom pain" experienced by people who lose limbs is a good example of the effects of shifting receptive fields, he says. "We know that 80 percent of amputees report the illusion that the part of the body they lost is still there. We believe that the immediate brain reorganization that begins after amputation may account for such phantom limb sensation." Such effects can be deeply strange. Touching the face of a person who has lost an arm often gives the person the sensation of having touched the missing arm. This phenomenon arises because the brains of people who have lost upper limbs are reorganizing to transfer to the face the allegiance of receptive fields formerly associated with the limb.

Besides revealing the shifting of receptive fields, Nicolelis' experiments have revealed that the brain signals producing a single event, such as a rat's paw press, may be mirrored in many places in the same brain region. It's as if the brain has enlisted neurons from many precincts to "vote" on all of its actions. Such redundancy makes good evolutionary sense, he says. "If you lose one of these areas, you still have all this processing machinery available in other parts of the brain. Also, the brain can let many areas handle one message and check each other for accuracy."

In miguel nicolelis' experiments, rats were trained to press a bar to obtain water, and their neural signals were analyzed for patterns associated with pressing the bar. to the human scientists' surprise, when they programmed the system to "listen" for those patterns, the rats had learned to trigger the bar purely by generating the right brain waves.
Illustration: Jerry Schoendorf

For Nicolelis, the major scientific challenge of his career is to sort out how such a complex, constantly changing, multi-processing brain somehow manages to come up with rational thought. "We learned a long time ago that there is no such thing as a stable brain. The brain is continuously changing through learning. This is in sharp contrast to the traditional view that the brain is pre-wired by early experience during a critical period, and changes little after that. So, we're trying to figure out the principles that allow the brain to be so dynamic and yet so reproducible--to control and produce the same kind of output day after day, even though its own individual elements are constantly changing due to the changes in the environment or in its own growth."

The brain manages to produce stable output because, despite its ever-changing nature, it seems to encode what Nicolelis calls an "internal model" of the world, which it compares with new experiences. "The brain actually has a point of view," he says. "It develops a model and it uses this model continuously to test whether signals coming from the environment can be ignored, or whether they're so novel that they need to be incorporated and used to update the model. This new view says the brain is such a dynamic machine that it never stops sucking out information to update what it has inside." And the modest laboratory rat is helping scientists understand how the brain goes about updating its internal models.

Nicolelis' rat-whisker studies have shown that new input reverberates throughout the brain, altering it at many levels. "As rats grow up, they constantly scan the environment with their whiskers and they somehow create a model of the external tactile environment. Our latest studies show that if we manipulate the cortex--the highest brain center --we also change the way the lower centers respond. It's as if the cortex continuously refines its internal model of the environment, and it has to let the entire brain know what information it is interested in getting."

The studies also reveal that the brain "decides" how to respond to external stimuli based on its own activity at the moment. "When we compared the response to a whisker-touch of a rat busily moving its whiskers and one sitting quietly, the response was totally different," he says. "The response has to do with attention, context, and behavioral relevance of the stimulus. It has to do with the brain's state at that particular moment. So, by no means is a stimulus always looked upon the same way. Every millisecond, the brain sees things differently because every millisecond tells a different story."

Nicolelis has discovered that his studies of the changing brain are, in fact, changing his own mind about his field. "The more I think about it, the more I realize how deep this dynamic-brain theory cuts into the dogma of neuroscience. It really opens up new avenues of research and thinking." He cites some philosophers who assert that this new dynamic-brain theory and brain-wave analytical techniques may represent the first steps toward understanding the very nature of thought. "Until now, people have always believed that a thought was some holistic, even spiritual entity that was impossible to define, that one could never really grasp. But some of us now contend that a thought is actually only a complex pattern of brain activity produced by the work of millions of neurons. Of course, in our experiments, the thought is simple--just a rat planning to move--but, nonetheless, it is a thought."

"Who knows?" he muses. "Maybe one day, decades in the future, we might even be able to record a pattern of human brain activity and actually decipher the thought it represents."

Share your comments

Have an account?

Sign in to comment

No Account?

Email the editor