Cover Story

Researchers have increasingly recognized, in recent years, that sleep serves many functions, including providing an opportunity for the body to rest, facilitating metabolic and endocrine function, and enabling "offline" memory processing. Still, the mechanisms underlying these functions have remained remarkably mysterious.

"Of all the biological drives--such as thirst, hunger and sexual desire--sleep is the only one whose function wasn't obvious 2,000 years ago," observes Robert Stickgold, PhD, a cognitive neuroscientist at the Harvard Medical School and the Massachusetts Mental Health Center.

In the case of sleep's effect on learning and memory, for example, studies have produced conflicting results over the years, with some linking REM sleep, in particular, to improved memory, and others failing to find such effects. But recent investigations have yielded new insight into sleep's role in memory and learning. The research confirms what some sleep experts have long theorized: that sleep is critical for firming up the learning that took place during the day--a process known as memory consolidation.

In two recent studies, Stickgold and his colleagues tested people's learning on a standard test of perceptual memory known as a visual discrimination task. They first trained study participants to perform the discrimination task as quickly as possible, then measured how well participants had learned the procedure by testing their improvement on the task hours later.

In an article published in the Journal of Cognitive Neuroscience (Vol. 12, No. 2) in March 2000, the researchers reported that participants showed no improvement when they were tested on the same day on which they had been trained. But when people had slept for at least six hours after training and before testing, they showed consistent improvement on the discrimination task.

In addition, the study showed that the amount of improvement that participants demonstrated was directly proportional to the amount of slow-wave sleep that they obtained early in the night and the amount of REM sleep that they obtained late in the night. In fact, Stickgold's group found that the amount of sleep that participants got during these two phases accounted for a striking 80 percent of the difference among participants in improvement on the visual discrimination task.

"That was startling to us," says Stickgold. "I've never seen such a strong correlation between the quality of people's sleep and any behavioral measures, let alone measures of memory and learning."

A further investigation, published in the journal Nature Neuroscience (Vol. 3, No. 12) in December 2000, found that a single night of sleep deprivation permanently short-circuited the memory-consolidation process: Participants who had been kept awake for 30 hours after an initial training session showed no evidence of improved performance on the visual discrimination test, even after two nights of restorative sleep.

In related research published in the journal Science (Vol. 290) in October 2000, Stickgold and colleagues had participants play the computer game "Tetris" for seven hours over the course of three days. They found that when research participants were awakened just after they fell asleep, 75 percent reported experiencing visual images of the game, suggesting that they were continuing to work on the problem during sleep.

"This is what 'sleeping on a problem' is all about," says Stickgold. He speculates that slow-wave and REM sleep activate a sort of "primal calculation mechanism" that helps the brain extract meaning from jumbled information.

Further support for that view comes from recent research on memory consolidation in rats by Massachusetts Institute of Technology neuroscientist Matthew A. Wilson, PhD, and colleagues. Two studies, published in the journal Neuron in November 1998 and January 2001 (Vol. 21, No. 5 and Vol. 29, No. 1, respectively), revealed the same patterns of brain activation when rats were learning to navigate in a maze and during subsequent slow-wave and REM sleep.

Such evidence of recently learned material being "replayed" during sleep is consistent with recent research in humans, led by Pierre Maquet, MD, a neuroscientist at the University of Li├Ęge, Belgium, and published in the journal Nature Neuroscience (Vol. 3, No. 8) in August 2000. Using positron emission tomography to trace brain activity, Maquet's group found that the brain areas that were activated when participants were learning a serial reaction time task were also activated during subsequent REM sleep.

Like Wilson's research, those findings appear to confirm that "what one learned earlier is being replayed and rehearsed, and the [memory] circuits are becoming stronger," says Carlyle Smith, PhD, a biopsychologist at Trent University in Peterborough, Ontario, and one of Maquet's co-authors.

Together, agrees Wilson, the recent findings call for "a changing view of memory, not just as storage but as a way of learning from experience, reorganizing thinking and building models of the world.

"If we can think of sleep as a creative learning process that allows us to extrapolate and look for novel juxtapositions of events, clearly that would be valuable in terms of our use of memory."