Support the news

"Sleep, Memory"

Scientists have known for 50 years that the brain handles short- and long-term memories in different ways.

Now researchers in Boston are contributing to a growing understanding of the role of sleep in transforming memory. WBUR's Health and Science reporter Allan Coukell explains.

The audio for this story will be available on WBUR's web site after 10 a.m. on Friday.

ALLAN COUKELL: We take it for granted that we know what memory is — like it's some sort of mental backpack that we can put stuff in and take out when we need it again. But Matt Wilson says much about memory is still a mystery.

MATT WILSON: We tend to think of memory simply as storage and retrieval. I show you a face, tomorrow you recognize the face. I give you a phone number, five minutes from now you are able to write down that phone number again. But ultimately, I think memory is a component of broader system of learning.

Wilson, a professor of neurobiology, keeps a symbol of memory — a glass statue of a sea horse — on a table in his office at MIT.

The Latin for sea horse is "hippocampus." And hippocampus is the sea horse-shaped part of the brain that Wilson studies. It is crucial for making new memories.


To see memory in action, Wilson and his post doc researcher Miguel Remondes have been running rats through mazes.

MIGUEL REMONDES: In this maze, basically the rat has to go around the circle and choose which entry to take...

The maze is shaped like a wagon wheel. If the rat enters the spokes in the right order, he's rewarded with a sip of chocolate syrup.

REMONDES: He's drinking the reward right now...

As the rat runs, electrodes monitor the activity of cells in his hippocampus.

REMONDES: He didn't go through number one, he went through number two.

COUKELL: Clever rat!

REMONDES: Yeah, he is clever (laughs).

Converted to sound, you can hear the crackling of about a hundred brain cells. There's a distinct pattern that corresponds to the rat's physical movements — each twist and turn of the maze. Matt Wilson has found that when the rat sleeps his brain plays back the same pattern of activity.

WILSON: They are replaying in an almost literal sense. You might think of replaying a movie that you just watched.

And it's not just the hippocampus. The "higher" visual cortex of the brain is also replaying its activity. It's as if the sleeping rat is re-living his experience. Wilson believes this coordinated replay may be a sign of memory being transformed, or consolidated, in the brain.

And evidence from human studies shows a similar effect.

MATT WALKER: The way we go about testing these questions is we'll usually train subjects on memory tasks.

Matt Walker is a sleep and memory researcher at Beth Israel Deaconess Medical Center.

WALKER: So subjects would sit at a computer keyboard and using their left hand they would type out a particular sequence, such as 4-1-3-2-4.

Walker's subjects practice the task over and over. Then, after a 12-hour break, they're re-tested for speed and accuracy. The performance of subjects who stay awake doesn't change. But those who sleep during their break improve by twenty to thirty percent.

WALKER: So in other words, following practice, your brain will continue to learn in the absence of any further training. So it is really quite magical. However, that delayed learning occurs exclusively across periods of sleep.

Walker has also shown that depriving someone of sleep for one full night and then giving them catch-up sleep on later nights produces no improvement. As Walker puts it, "if you don't snooze, you lose."

Some scientists remain skeptical of the link between sleep and memory. Jerry Siegel, a sleep researcher at UCLA, points out that many people do not experience REM, or "dreaming", sleep at all.

JERRY SIEGEL: People who don't have REM sleep either through brain injury or taking drugs have no cognitive deficit whatsoever. And there are people who have gone without REM sleep for years.

Matt Walker, of Beth Israel Deaconess, counters that the different phases of sleep seem to serve different functions. He says REM sleep appears most important for reinforcing emotionally charged memories — precisely the area where brain injury patients and people on psychotropic drugs often have trouble.

But if memory isn't a backpack where facts and experiences are preserved, what is it? And what is the role of sleep? Matt Wilson of MIT believes we process memories so we can re-evaluate them:

WILSON: Either for the purpose of selecting and strengthening certain memories. Or I think more importantly for going back and comparing different experiences that may have occurred at different times. And the reason for comparing is so that one can learn from them.

A decade ago, there was little scientific evidence to explain why we sleep. The emerging science of memory provides at least some explanation of what's going on during that third of our lives. For WBUR, I'm Allan Coukell.Sidebar: Meat versus Machine

Matthew Wilson believes that the brain processes memory to develop "wisdom"--which he defines as the ability to make the right decision based on partial information. That is also one goal of research in artificial intelligence, and Professor Wilson (who trained as an engineer) finds an interesting correlation between how rats replay their experiences and how some computers make decisions.

It seems that when the rats are asleep, they normally play their experiences forward. But when they are just resting — daydreaming, Wilson calls it — the replay patterns run backwards.

Rich Sutton, an artificial intelligence researcher at the University of Alberta, says a number of computer algorithms, so-called reinforcement learning algorithms, also work backwards. "These are algorithms designed for whenever you have a sequence of decisions to make. It's been applied to autopilots for helicopters, computer games; many, many robots, financial applications, logistics."

Professor Sutton, who studied psychology before switching to computers, says maybe it isn't surprising that a rat learning to remember a maze would start with the reward and work backward to figure out how he got it. The same principals of efficient processing could apply to animals and computers.

This program aired on March 30, 2007. The audio for this program is not available.

+Join the discussion

Support the news