Assistive technology that helps severely paralyzed people navigate the world and communicate with others often taps into whatever abilities the disabled retain, such as blinking or moving the mouth and tongue. Now, for the first time, researchers have invented a device that allows the paralyzed to write, surf the Web and steer an electronic wheelchair—all by sniffing. Initial tests, described July 26 in Proceedings of the National Academy of Sciences (PNAS ), suggest that many severely paralyzed people can easily master the "sniff controller," which offers certain advantages over other technological aids.

Sniffing is controlled in part by cranial nerves in the soft palate, the tissue lining the back of the roof of the mouth. Because cranial nerves emerge directly from the brain, as opposed to the spinal cord—and because their neural networks are well distributed—many severely paralyzed people retain movements regulated by the cranial nerves, such as blinking, sipping and sniffing. The "sniff controller," developed by researchers at the Weizmann Institute of Science in Israel, fits onto the nose like a cannula—the small plastic tubes used in hospitals to provide patients with oxygen. The device measures pressure inside the nose, translating the intensity and frequency of sniffing into electronic commands for a computer or wheelchair. The researchers also created a passive sniff controller that provides nearly the same level of control to individuals on respirators.

Maysam Ghovanloo, an electrical engineer at the Georgia Institute of Technology who has been developing magnetic tongue-controlled technology for the paralyzed, sees potential in the new approach: "The disabled have to use any remaining ability that they might have. Cranial nerves are some of the last nerves they might lose, although it's different from one patient to another. As far as the control aspect is concerned, it's definitely possible that sniffing would work quite well."

Anton Plotkin, lead author of the PNAS report and an electronics engineer at Weizmann who has also submitted a patent for the new technology, thinks the new study makes a major contribution. "As far as we know, this is the first time someone is looking to sniff power to control movement and communication," Plotkin says. "We figured out that sniffing can control just about anything. [It]…is quite accurate and fast and can be compared to control with a joystick or mouse."

The research team first tested the sniff controller on 36 healthy non-paralyzed volunteers who could play a series of computer games using the device with as much accuracy as a mouse or joystick. The researchers then tested whether healthy participants could use the sniff controller to guide text-writing software. On a screen, a cursor continuously skipped between a block of letters, a block of numbers, and a word-completion box, successively highlighting each in bright green. Participants sniffed to halt the cursor's movement when it highlighted the right block, sniffing again to move between lines or characters within a single block. Volunteers were able to write a 43-letter sentence, spending an average of 6.73 seconds on each letter, or about nine or 10 letters per minute. Although this might seem tedious, the researchers explain that for people with extreme paralysis even sluggish forms of expression and communication are welcome.

After trying the sniff controller on healthy participants, the researchers began trials on patients with locked-in syndrome, who maintain awareness and cognitive abilities but cannot move at all. The team's first patient was a 51-year-old woman who developed locked-in syndrome following a stroke seven months earlier. She could breath independently but could not control her eye movements or properly direct airflow through her nose—at first. After 20 minutes of practice a day for 19 days the woman finally learned to sniff with purpose and immediately began using the text-writing software, composing her first personal message to her family since her stroke. She spent approximately 20 seconds on each letter—writing about three letters a minute—taking more than twice as long as healthy volunteers. The researchers note, however, that this rate exceeds that at which locked-in French journalist and author Jean-Dominique Bauby wrote his novel The Diving Bell and the Butterfly, which he dictated at an average rate of one word every two minutes, using blinks alone.

Encouraged by these results, Plotkin and colleagues tried the sniff controller on a 42-year-old man who had been locked-in for 18 years following a car crash and was only capable of communicating by blinking a single eye. After only 20 minutes of practice, the man could write his name and later wrote that the device was "more comfortable and more easy to use" than eye-tracking technology. A 63-year-old quadriplegic woman who could only speak with extreme difficulty learned to use the sniff controller to precisely direct a cursor, allowing her to write for the first time in 10 years, surf the web and send e-mail. A 64-year-old man locked-in after a stroke four years earlier, however, could not learn to use the sniff controller at all. The researchers are not sure whether this was due to failings of the device, or because the patient was experiencing severe depression and may have lacked the will to learn. "We completely failed with that person and we're don't know whether to attribute it to the technology or an emotional failure," Plotkin says.

The sniff controller enables more than just communication, it offers paralyzed patients who use wheelchairs a new way to navigate the world. The researchers devised a simple code to guide wheelchair movement: two sniffs in means move forward; two sniffs out means go backward; a sniff in followed by a sniff out turns the chair left; reversing the order turns it right. Ten healthy volunteers used the sniff controller to maneuver an electronic wheelchair through a complex 35-meter-long course that presented many 90-degree turns. Although they navigated with ease, improving over time, the patients were still a little more accurate with a joystick than when they relied on sniff steering. One 30-year-old quadriplegic patient paralyzed from the neck down learned to guide the wheelchair by sniffing with just as much precision as healthy participants after only 15 minutes of practice.

"We cannot say that our technology has absolute advantages over other technologies," Plotkin says, "but it's much simpler and less expensive than eye-tracking—and not all locked-in patients can really move their eyes. But most can still control their sniffing via cranial nerves in the soft palate." Plotkin also points out that the sniff controller minimizes interference with gaze and speech, allowing paralyzed people who can speak to continue doing so even when sniffing. Speech is not possible with the sip-and-puff technique, in which a paralyzed person sips in and puffs out onto a straw to control a wheelchair or a computer mouse.

Georgia Tech's Ghovanloo wonders whether the sniff controller is too limiting in its vocabulary of commands. Sip-and-puff technology offers users four basic commands—hard sip, soft sip, hard puff and soft puff—which some users find restricting. With the sniff controller there are really only two basic commands—sniff in or out—which is why wheelchair control requires a code of sequential sniffs.

Ghovanloo also worries about the amount of muscle control needed to sniff. "In order to change pressure in nostrils you need control over your diaphragm. Many individuals with high-level spinal injury lose control over their diaphragm and cannot breathe properly on their own." The sniff controller, however, can measure changes in pressure caused by sniffs regulated by the soft palate, independent of breathing.

Another concern is the possibility of hyperventilation when navigating demanding terrain or that ordinary breathing could disrupt the sniff controller's code and crash the wheelchair. "My real concern would be how normal breathing interferes with this particular type of assisted technology," Ghovanloo says. "With the assisted technologies that track eye movements, if you are looking at something else there is always a chance that the computer processes your irrelevant eye motion as a command. The same thing could happen here with breathing."

Carbon dioxide tests by the researchers indicate that hyperventilation is not likely. And the sniff controller code, which requires patients to sniff multiple times even for a single command, reduces interference of accidental sniffing or ordinary breathing. It is possible, however, that commands and codes of increasing complexity might hinder ease of use. The researchers only tested sniff-powered wheelchair navigation on one paralyzed individual, so further testing will be necessary to tease out any flaws and confirm the technology's overall feasibility.