As Philip R. Kennedy, physician and inventor, prepares a paralyzed man to operate a computer with his thoughts, it briefly seems possible a historic scene is unfolding in this hospital room, and that Kennedy might be a new Alexander Graham Bell. Of course, it's also possible he's a new Johann Philipp Reis. Reis built a telephone out of a violin case, sausage skin and pieces of a beer barrel in Germany in 1860. That was 16 years before Bell unveiled his telephone, which revolutionized human communication. Reis's invention worked, too, but not well enough to be useful. It proved a point, but did little more. "Johnny, are you comfortable?" Kennedy asks as he steers three carts covered with electronic gear over to the hospital bed and begins to unpack wires. "Is your back hanging in there?" The man in bed is middle-aged, red-faced and motionless, except for the muscles of his face. He blinks twice, his sign for "yes." The room, in a veterans hospital, is supposed to hold four patients, but Kennedy's equipment takes up the space where one would normally lie. Two other men occupy beds across the room from the paralyzed man. They pay no attention as the doctor and his assistant plug in wires, calibrate an oscilloscope, tape electrodes onto the patient's skin, crank up the head of his bed and position a video monitor in front of his face. It takes about 20 minutes before Kennedy opens a cloth-bound ledger, the official record of his audacious experiment. "Day 259," he says aloud, starting a new entry. "Day 259 after implantation." The man in bed tracks him with his eyes. Nearly nine months earlier, two small electrodes were placed in John L. Ray's brain. They record the firing of his brain cells -- which is to say, his thoughts. Now, he's learning to use those thoughts, carried out of his head through tiny wires, to move the cursor on a computer screen. He can't do it very well. But he does it better than he did a few months ago. If someday Ray learns to control the cursor as smoothly and effortlessly as he once controlled his limbs, all sorts of possibilities will open up. He'll be able to write a sentence, surf the Web, turn on the lights, drive a wheelchair. In theory, that is. On the other hand, it may turn out that all he, and the patients who may follow him, will ever be able to do is clumsily and exhaustingly bounce the cursor around. In which case this work will be a historical curiosity -- something that proves a point, but does little else. "Johnny, you ready to start?" Kennedy asks. Philip Kennedy's work is based on a simple fact: Thoughts are not ethereal things. They obey the laws of physics, and can be used as physical tools. Until recently, only in science fiction, magic shows and certain forms of religious devotion did people try to move objects with thought, and without the intervening action of human muscle. Now, a few people are trying to do it in medicine. The purpose of "cortical control devices" is to capture the physical essence of cognition and, with the help of lots of intervening machinery, translate it into action. Kennedy is the first scientist to oversee the implantation of such a device into a person's brain. Under normal circumstances, voluntary movement begins when a nerve cell in the brain -- a neuron -- fires. This sends a low-voltage pulse of electricity through the cell's microscopic tendrils, which contact dozens of other cells. The pulse causes some of those cells to fire, others to stop firing, and in all triggers an immensely complex chain of events that ends with an eye blinking, a finger pointing, a toe tapping. Disease or accident can break this chain of transmission, rending the filamentous path traced from neuron to neuron to muscle cell. Of course, a person may still be able to think about moving a hand or a foot. But in a literal sense, the thought will have nowhere to go. This is what happened to Johnny Ray. He's a 52-year-old man from Carrollton, Ga., a drywall contractor, a three-tour veteran of the Vietnam War, an amateur country music guitarist. About 14 months ago, a blood vessel burst in his brainstem, which connects the spinal cord to the brain. This is about the worst place for a stroke. No bigger than two fingers lying side by side, the brainstem consists (mostly) of millions of nerve fibers carrying signals from the brain out to the body, and from the body to the brain. It's the main trunk line of the human nervous system. The clot that formed in Ray's brainstem compressed nearly all the fibers going from brain to muscle, and some of the fibers carrying sensation from the skin to the brain, as well. Today, he can produce flickers of movement in his left big toe and left thumb. He can move his tongue, the muscles of his face, his eyeballs. Otherwise, he can't move. He can't breathe on his own. He is attached to a mechanical ventilator through an opening in his throat called a tracheostomy. He can't talk. He can, however, feel much of his body. Pain and temperature sensation below his waist is diminished, but otherwise his sense of physical presence is largely unchanged. He is fully conscious. He can see and hear, think and desire. The name for this condition is worthy of Poe. It's called "locked-in syndrome." For several months after the stroke, it appeared Ray was in a light coma, not locked-in. But it turns out his diminished consciousness was the result of repeated infections and other complications of the paralysis, not from brain damage. Surgery on his sinuses and removal of his teeth brought these under control, and by last March his mind was clear. He agreed to have the two electrodes implanted in his head. For more than a decade, Kennedy had been working on a strategy he believed would let paralyzed people use their own neural activity to manipulate the environment. It's not too much to say the idea possessed him. A native of County Limerick in Ireland, Kennedy trained as a surgeon in Dublin. In 1976, he immigrated to Canada to study neurosurgery. After a few years, he decided he really preferred research. He moved to Chicago, where he got a PhD in neuroanatomy at Northwestern University, and then to Atlanta, where he had various post-doctoral positions at Emory University and the Georgia Institute of Technology. In 1986, he began working on a cortical control device. He designed an electrode that would not move around in the brain, which was the major problem with many ones previously tried by researchers. He tried his in rats and monkeys, doing the surgery himself. It worked. It was stable, undamaging, and capable of carrying a signal out of the brain for months on end. He got a few small grants, but was unable to get one from the National Institutes of Health, which spends little on research on cortical control devices. (This year it's roughly $900,000 out of a budget of $15 billion.) Georgia Tech didn't want to spend the money to patent his device, so Kennedy found a lawyer willing to work for a reduced fee, and patented it himself. By 1992, when he did his last monkey implant, he was certain he had what was needed for human trials. He was also divorced, and for a brief time on unemployment. "Nobody would fund it," he says of the research. "I was going broke." In 1993, he started his career a third time. He did a four-year residency in neurology. Now 51, he's in private practice, shoehorning his research into a day and a half each week, subsidizing it out of his own practice. Recently, he got his first NIH grant -- about $100,000 divided over three years. The Food and Drug Administration, which regulates medical devices, has given Kennedy and Roy A.E. Bakay, an Emory neurosurgeon, permission to try the device in three patients. Their first was a woman with amyotrophic lateral sclerosis -- Lou Gehrig's disease. She was profoundly incapacitated, near the end stage of the disease, and survived only 77 days after the surgery. It was long enough to prove the electrodes were working and that she could exert some control over the neural signals they picked up. It wasn't long enough, however, for her to learn to do anything with them. That's what John Ray is trying to do. The idea that one could profitably attach electronic hardware to neuronal wetware only makes sense because the brain is finely organized on the basis of function. This is untrue of most organs. A sample of tissue taken from one place in the liver does the same work as a sample taken three inches away. But the work of brain cells changes profoundly over distances as small as a millimeter. Neuroscientists mapped the brain's "motor cortex" more than 50 years ago, identifying the areas where cells controlling various groups of muscles lie. Those locations, however, vary somewhat between individuals, and in some circumstances can shift during life. So the first thing Kennedy and Bakay had to do was determine precisely where Ray was doing his thinking when he was thinking about moving muscles he could no longer move. They did this with magnetic resonance imaging (MRI), which can identify the increased blood flow that occurs when a region of the brain is in active use. The researchers asked Ray to think of moving his left hand. They watched where the MRI scan lit up. Ultimately, Bakay implanted two electrodes side by side in that part of the brain. Kennedy's electrode is a hollow glass cone less than one-tenth of an inch long. It's open at both ends and attached to two tiny gold wires. After implantation, the microscopic tentacles, or "neurites," of nearby neurons grow into the cone over several months. The neurites don't make direct contact with the electrode wires. Instead, they go about their business trafficking in electrical impulses. The wires simply listen in and carry the signal out of the skull, where it's amplified. It's pure chance which neuron sends a neurite into the cone. But as the patient thinks about moving one muscle or another -- activating hundreds of cells with each thought -- eventually he'll use one of the ones that's grown into the cone. The person knows this because the signal coming out of his skull is connected to a loudspeaker and an oscilloscope, which provides a visual display of his brain signals, among many other pieces of machinery. He can hear his thought crackle and see its electrical spikes. With that feedback, he can learn to turn it on and off, make it weak or strong. Kennedy found that monkeys could produce two or three distinct signals, as measured by frequency and wavelength, from a single electrode when they moved a muscle. (Unlike the patients, the monkeys were not paralyzed.) This meant that, theoretically at least, each signal could perform a different task. One could move a cursor horizontally, another vertically, and a third could click. Most people control a computer with their hands. But for people with total quadriplegia, the mouse is hopelessly out of reach. Cortical control devices attempt to move it closer. "We've moved the mouse back up here," says Bakay, tapping on the side of his head. In this field test, things are better, worse and not as they are supposed to be. None of this surprises Kennedy, Bakay or the other people working with Johnny Ray. They know that when it comes to the brain, things are rarely simple or easy. One of the two cones in Ray's motor cortex works so poorly it's not usable. Something is interrupting the transmission. Furthermore, the cone that does work is no longer recording cells activated by Ray's thoughts of moving his hand. Now, it's activated when he thinks of moving part of his face. This second unexpected event isn't a problem with the cone. Instead, it's the result of Ray's brain remodeling itself. This is not entirely unexpected. The brain's "functional architecture" sometimes changes, even in adulthood, so it can devote more space to parts of the body that have become exceptionally important. (Blind readers of Braille, for example, have large areas of brain devoted to processing sensation from the fingertips.) Ray can still use the muscles of his face. The part of his motor cortex devoted to that task appears to be growing. The most important unexpected event is that in recent months Ray has recovered movement in a toe and a thumb. The movement isn't enough to be mechanically useful. But the signal given off by the contracting muscle is enough to drive a computer cursor once it's amplified, transformed and sent through the appropriate electronics. Now, in practice sessions, Kennedy attaches a surface electrode to those appendages. When Ray moves his toe, the cursor moves down. When he moves his thumb, the mouse clicks. The only thing his thoughts do is move the cursor to the right. Nevertheless, even under these unexpectedly simplified conditions, the work is very hard. "Johnny, don't worry," Kennedy says. "Don't be too careful. Just go for it." Ray stares at the computer monitor, whose screen is divided into blocks representing the characters on a typewriter keyboard. This is his fourth session trying to control the cursor well enough to spell a word. Today, the word is John. He starts the cursor rightward across the number row, with his thoughts. Then he moves it down, with his toe. Then to the right again. It stops on a block and he clicks. It's H, one character short of the J he is trying for. Kennedy clears the screen and adjusts the speed at which the cursor moves. Ray starts again. He thinks the cursor to right, but this time overshoots the J. He then has to move it further to the right until it disappears off the edge and wraps, appearing on the left edge. He moves it across and clicks. M. It's the character right below the J. He's thought the cursor to the right spot, but one row too low. He tries again. After many minutes, he has MJOJ}. "You feel tired, Johnny?" Kennedy asks. Ray blinks once for "No." Kennedy says, out of his earshot: "When he gets tired, the signals drop off completely. And when he's more active and joking around, the signals pick up." Kennedy and Bakay have gotten a lot of help custom-fitting the computer software to Ray's capabilities. To adjust for the varying strength of Ray's signal, for example, Melody Moore, a computer engineer at Georgia State University, rewrote the program so that when he's active and jumpy, the baseline "neural force" needed to move the cursor increases. When he's tired, the baseline falls, so it takes less concentrated thought to move it. A Georgia Tech senior, Gregory Montgomery, has written a program that allows Ray to move the cursor onto icons that deliver preprogrammed messages. An Atlanta Braves cap -- Ray is a baseball fan -- triggers the message: "Hi, my name is Johnny Ray." A blue ice cube produces: "I feel too cold." In a minute, Ray tries again. Bakay, dressed in a white medical coat, has joined the group around the bed. "A bit nervous?" Kennedy says puckishly to Ray. Ray gets the J. The cursor moves slowly across the screen in search of the O. "Get mad at it," Kennedy says. The oscilloscope crackles and the cursor starts to fly. "But not too mad at it." He adjusts a dial so that a lighter twitch of Ray's left thumb will trigger the enter function. Ray's having an increasingly hard time clicking the letters as the session progresses. He gets the O, and then arduously moves the cursor across the line with the H in it, letter by letter. He clicks. "Try the foot to bring it down now," Kennedy says. The cursor drops, and almost immediately he gets the N. "All right!" says Kim Adams, a young research engineer who recently began working with Ray. "Awesome," Kennedy says. Then to the assembled group: "He spelled his name without error in a reasonable amount of time. In about six minutes." The doctor gets a washcloth and wipes off Ray's forehead and temples. He flips the cloth to find a dry section for the eyelids, which in the low light of the room are tan pools in a field of pink flush. "How was the effort that last time?" Kennedy asks. "One out of five?" Ray doesn't blink, and the doctor continues until he reaches "five out of five." Ray blinks twice. "Yes." Ray's wife died unexpectedly last summer, which set him back badly. His spirits are better now, says a 31-year-old stepson, James Christopher Mitchell. One day every weekend, Mitchell and his family drive 60 miles from Mount Zion, Ga., to spend a few hours at his bedside. "We sit with him and we talk with him. The whole point of it is don't treat him like he's an invalid. Treat him like he was at home," Mitchell says. "As far as the computer, I think Johnny was thinking that it would be a lot quicker. But he's working with it. He hasn't given up." Recently, Ray developed two pressure sores. They aren't the first since his stroke. The combined effects of infection and painkillers have made it impossible for him to practice for several weeks. Before this happened, Ray spelled out "Phil Kenedy" and "Melody" and "Kim." "It's a big disappointment to him and to me," Kennedy said last week. "Because he was doing better, and he knows it." Capturing Thoughts A paralyzed stroke victim named Johnny Ray has learned to use thoughts to control a computer cursor via electrodes implanted in his brain. 1. The electrodes (two gold wires attached to a hollow glass tube about the size of a ballpoint pen's tip) were implanted in the motor cortex, where cells controlling various muscle groups lie. Normally, movement begins when neurons in the motor cortex fire. 2. The microscopic tentacles of neurons, called neurites, have wrapped in and around the electrodes. When Ray imagines certain kinds of movement, the electrodes pick up the signals emitted from firing neurons. 3. The signals are amplified and transmitted to the computer, which translates them into cursor movements. SOURCES: Philip Kennedy and Ray Bakay CAPTION: At veterans hospital in Atlanta, John L. Ray attempts to move cursor on computer through brain signals transmitted by implanted electrodes. ec CAPTION: Philip Kennedy, left, and Roy Bakay work with patient John Ray at VA Medical Center in Atlanta. Electrode implants transmit Ray's neural signals to computer. ec