In a first step toward helping severely paralyzed people communicate more easily, Utah researchers have shown that it is possible to translate recorded brain waves into words, using a grid of electrodes placed directly on the brain.

Although they have only done it with one person and individual words can only be identified with accuracy in tests 50% of the time, the study reported Tuesday provides a ray of hope for people who can now communicate only by blinking, or wiggling a fingertip.

"This is quite a simple technology … based on devices that have been used in humans for 50 years now," said bioengineer Bradley Greger of the University of Utah, the lead author of a report in the Journal of Neuroengineering.

"We're pretty hopeful that, with a better design, we'll be able to decode more words and, in two or three years get approval for a real feasibility trial in paralyzed patients," he said.

The device could benefit people who have been paralyzed by stroke, Lou Gehrig's disease or trauma and are "locked in" — aware but unable to communicate except, perhaps, by blinking an eyelid or arduously moving a cursor to pick out letters or words from a list.

Researchers have made tremendous strides in developing ways for patients to move a computer cursor or even an artificial arm using electrodes implanted in the brain. But researchers have been reluctant to implant electrodes in the speech centers for fear of causing irreversible damage.

Some researchers have been attempting to "read" speech centers in the brain using electrodes placed on the scalp. But such electrodes "are so far away from the electrical activity that it gets blurred out," Greger said.

He and his colleagues instead use arrays of tiny microelectrodes that are placed in contact with the brain, but not implanted. In the current study, they used two arrays, each with 16 microelectrodes. The arrays were placed directly on the brain of a volunteer patient with epilepsy whose skull had already been opened to measure aberrant electrical signals that trigger seizures.

The team tested 10 words, such as "yes," "no," "hungry" and "thirsty," that a patient might need. The volunteer spoke each word 31 to 96 times while the researchers measured brain waves.

In the best case, the researchers could correctly distinguish between two words, such as "yes" and "no," 90% of the time. But when trying to distinguish among all 10 words, their best accuracy was 48%.

Greger believes the accuracy can be improved by using more electrodes, and the team is now working with grids of 121 sensors.