The ability to better define sleep states could shed light on the various diseases, such as depression and Alzheimer's, that have been linked to sleep disorders. "We know, for example, that kidney disease is associated with sleep disruption," says Sejnowski. "I suspect we will be able to pick that up and maybe diagnose these diseases before they become so serious that they require surgery or dialysis."

Because the algorithm allows automatic analysis, the technology might be useful for studies that requires long-term recording, such as readings of epilepsy patients, or large numbers of patients, which are more labor-intensive to analyze by hand. "It would be ideal for studying drug effects on sleep," says Jerome Siegel, a neuroscientist who studies sleep at the University of California, Los Angeles. (Other automated sleep-analysis methods do exist, but many sleep centers still rely on a human interpretation.)

Researchers say that simpler sleep-analysis technologies would be useful. But they caution that more testing is needed to determine if this particular method will prove more reliable than previous analysis algorithms, as well as to show when it may be most useful. "A computer program might not understand patients with severe disabilities," says Mark Eric Dyken, director of the Sleep Disorders Center at the University of Iowa.

Such studies are in the works. The researchers plan to analyze sleep records from patients with Alzheimer's and narcolepsy. They are also working on preliminary research into the basis of sleep. "Why do we need sleep?" asks Sejnowski. "Why does it have the effect it does? Clearly sleeping does something to the brain to help fix it. We want to find out what that is."