2004 LECTURE SERIES

Sleep, Memory and Dreams: What are they good for?

Dr. Robert Stickgold
Harvard Medical School
Tuesday, March 16, 2004

Thank you, Jim, for those very kind words. It’s a pleasure to be here, not just because there’s a blizzard in Boston, as I speak. You people know what--yeah, you do. Okay. But also to have this opportunity to speak in a public lecture series and bring the new information that we’re discovering to a wider audience than just our colleagues and our students.

What I want to talk about tonight is--there it is. Sleep, dreams, and memory and trying to understand how they relate to each other. We all know that we sleep--well, no, we don’t. We pretend that we sleep about eight hours a night. Let me get out of the light. How many of the high school students in the audience average eight hours of sleep a night? Can I see a show of hands? Are there any high school students in the audience? There’s about two dozen hands that go up and the rest of them get less. And maybe I will scare some of them by the end of this lecture, when we look at the effects of shortened sleep on memory consolidation.

We grew up, most of us, in my generation thinking of sleep as a time when the body rested and the brain shut off and nothing much of interest happened. And I think a lot of the reason that there hasn’t been much research on these subjects over the years is because the idea was that the brain shut off and nothing is happening. And, in fact, nothing can be further from the truth. My first main point for this evening is that the brain activity changes dramatically across the night.

I’m going to start out with a few slides, which are sort of techie-science slides, but the point is just to show you the amount of variation in brain activity across the night. Now, a night of sleep is not a continuous period of sleep, but one that goes along in a cycle that lasts about 90 minutes all night long. If you’re young or if you’re tired, it starts out with a period of very light sleep when you’re just falling asleep. It takes about ten minutes on average for a person who’s not feeling anxious to fall asleep. And then over the next 45 minutes or so, you descend into deeper and deeper stages of sleep. We in the sleep research field are very clever people so we come up with these names like stage one and stage two and stage three and four. So you descend down into this deep, deep sleep here and then after about an hour or so, surprisingly, your sleep starts to lighten up. And then you come up to this period where you have rapid eye movement sleep or REM sleep. For the high school students, this is not named after the band. Actually, the band is named after the sleep stage. And it’s the only sleep stage that has a successful band named after it. There’s no band called any of these others.

So you start out with these 90-minute cycles that are pretty continuous all night long. You’ll see that most of the deep sleep--and we call it slow-wave sleep because of the EEG patterns--most of that deep slow-wave sleep happens in the early part of the night. Most of the REM sleep actually happens near the end of the night. And then sort of stuffed in-between all the way along is this light, non-REM sleep. And in the course of the evening, I’ll talk about possible functions for each of these types of sleep, in terms of memory processing. And one concept that I want to present is that possibly the reason that these complex variations evolved in mammals and even lower than mammals is because there are different types of memory processing that needs to be done offline, after we have the actual experience and that these different phases of sleep are tuned to those different demands. They differ in a bunch of ways. This is just showing the variation in EEG patterns that you get, different stages. And this is that slow-wave sleep and you can see these big, slow waves that are recorded from electrodes of the scalp, reflecting underlying brain activity.

And then there are eye movements, and these are the rapid eye movements that REM sleep is named after. You get these bursts of eye movement where the eyes just jiggle back and forth in the sockets. The function is still debated. The mechanism is pretty well understood. You also get eye movements right as you’re falling asleep. The eyes literally just sort of wander back and forth in the socket at about this speed. We have no idea why. We don’t know the mechanism and we don’t know the function. But all of these things have evolved and they’re very consistent across most of the mammalian range. So these are important functions that we’re still struggling to learn.

This is just a comment on some neurochemicals that are released in the brain and these are chemicals like acetylcholine, norepinephrine, and serotonin, all of which are involved in controlling the state of the mind. You might know them better not by the chemicals themselves, but the drugs that are used interact with the systems. So the SSRIs, like Zoloft and Prozac are served by increasing the function of serotonin. Belladonna, atropine interact with acetylcholine. And the MAOIs, the monamine oxidase inhibitors that have been used in the past as anti-depressants act by inhibiting the breakdown of these chemicals. So these are all chemicals that control the state of the mind.

And if you look what happens across the day, in active wake, they’re all quite strongly released in the brain. As you’re lying in bed starting to fall asleep, the quantities go down. In slow-wave sleep, the acetylcholine release almost stops. And then when you go into REM sleep, the whole system flips upside-down. It’s as if the whole program that’s running the brain is being altered from sleep state to sleep state, in this case, by chemical neuro [inaudible]. It says if you’re taking drugs when you’re in REM sleep. It’s changing the activity of the brain as much as those drugs will do.

If you go into the brain, you can see the different regions of the brain are activated in different sleep states and these regions, which are involved in emotions, tend to be cranked up to higher activity during REM sleep. This structure up here, which is involved in logical reasoning is cranked down in REM sleep. So it looks like as you go into REM sleep, the brain is tuning itself to look at its information, to look at memories, from a more emotional perspective and from a less logical and reasoned perspective. And we can even see changes in the flow of information between different memory systems across the stages. And I’ll talk more about that later.

So that’s sort of the physiology part. And now I’m going to talk about experiments where we actually looked at memory. And I’m going to start off by trying to convince you that, in fact, sleep enhances memories, that you can wake up smarter than you went to bed the night before and that these are active processes occurring in the brain.

Now, when we talk about memories, we have a bit of a problem because there are lots of different types of memories that are affected differently during sleep. So one type of memory is called episodic memory. That’s memories of specific events in your life. So if someone says, can you remember when, you’re trying to find an episodic memory. That’s different from semantic memories, which refer to general knowledge. So if I were to ask you, who was the first President of the United States, you don’t remember when you learned that piece of information, you can’t even remember who you learned it from, but you have that information. And that’s semantic knowledge and semantic memories, what kind of examples.

So your memories of September 11th are episodic memories. You can remember where you were, you can remember how you found out about it. If you think about it, a whole event seems to reply in your mind. Whereas, there’s probably few in the audience who remember Pearl Harbor and for the rest of you, it’s just information you have and you don’t remember where you got it. What was that? So similarly, the Challenger explosion versus the Graf Hindenburg explosion, two explosions and this big, popular press coverage, but we tend to know this in an episodic way and this just in a semantic way.

And in less, you know, exciting realms, what you had for dinner last night is something that you recall as an episodic memory, but if I asked you what your favorite dinner was, that’s more of a semantic memory. If I ask you, where did you park your car tonight when you drove here--you all drove, right? This is Southern California. Anybody--if I asked you where you parked your car tonight, that’s an episodic memory, but if I ask where are good places to park in this part of town, that’s more a semantic memory. So these are the two major categories or memories, in terms of what we can report that we know.

Separate from that is another whole class of memories, called procedural memories. These are how to do things so how to ride a bicycle, how to play a piano. These are things that you, in fact, don’t remember in the same way that you remember either episodic or semantic memories. And what I’m going to tell you is that sleep is really good at consolidating these procedural learnings, such as perceptual skills.

So the first studies that we did are looking at a visual detection task and it sort of looks like this. If you were in this study, you would get to see something like this about 1,250 times. At the center of the screen, there will be either a T or an L and then down here, there’s going to be three diagonal bars. And they might be next to each other in a row or they might be one above each other in a column. And we’re going to flash it kind of quickly and then we’ll flash a messy screen to confuse you. And then the question is, can you remember what you saw here and what you saw there. And the more time we give you between flashing the first screen and the second, the better you’ll be at it. So everybody ready? Okay. There’s the fixation point. Keep your eyes there. There’s going to be a letter here, there’s going to be some diagonal bars there. You ready? Okay. Everybody got it?

I got to tell you, Harvard students have no problem with this. After the first 50 or so. Let’s do it one more time, okay? Keep your eyes at the fixation point. Okay. How many saw a T at fixation? A few hands. How many saw an L at fixation? Oh, sorry Ts. Okay. How many saw three diagonal bars in a horizontal row? I got a hand or two out there. How many saw them in a vertical column? Oh, good for you. Let’s see what we had. Okay. You can all come to Harvard. It’s okay. Okay.

So we do this and by varying the amount of time between the first and second screen, we can figure out how fast you are at detecting it. That was really slow. That was 500 milliseconds between the two. If I actually took you and ran you through the full thousand, you’d get down to about 30 milliseconds, almost 20 times faster than that. And then what we can do is we can bring you back later and have you do it again and see if that threshold has changed. Are you faster? And we can bring you back various times later. And if we bring you back, say, 3, 6, 9, 12 hours later, you show absolutely no improvement at all. In fact, we can see no evidence that you learned anything at all.

But if instead, instead of bringing you back at one of these times, we wait and bring you back after a night of sleep then you show a large and dramatic and consistent improvement. These are all different groups of subjects so these people only were tested out here at 16 hours. So if you’ve had a night of sleep after we trained you then you will perform better. But if you haven’t had a night of sleep, you won’t show any improvement at all. And in fact, if we wait more than one night before we test you, if we test you after two nights or three nights or four nights, you’ll do even better than if we had tested you on the first night, let alone tested you the same day.

So something very unintuitive has happened. Normally, we think about when you study something, you learn something, you’re at your peak performance right after you study and then it’s sort of all dribbling downhill from there. But for this type of learning, it actually improves in the absence of any practice over several days and actually, in fact, over several nights.

Now, for those high school students in the audience, here’s my favorite group. These subjects, like the subjects in this purple bar behind, were trained on day zero and tested 72 hours later. The only difference between the two groups, except for the fact that these don’t show any improvement, is that they stayed up the night after we trained them, okay? The second night, they got all the sleep they wanted. The third night, they got all the sleep they wanted. They come in on this day of testing and they say they’re fully recovered from that first night of sleep deprivation, but they have permanently and unalterably lost the benefits of the training on the first day. So sleep the night after you learn might be as important or more important than sleep, for example, on the night before you train, something that’s very counterintuitive.

What part of the night of sleep do you need? Well, it turns out that you need that slow-wave sleep in the first couple of hours of the night and you need the REM sleep in the last couple of hours of the night. In fact, in our studies, when subjects only, by their own choice in this case, got six hours of sleep a night, they showed no overnight improvement. So think about that before you pull those short nights. You seem to need to have REM sleep in the late part of the night and here, you can actually see it. That’s the correlation, how important slow-wave sleep early in the night is versus the second two and third two and last two hours of the night. And you can see REM sleep following the exact opposite pattern.

Now, as a parent, this is scary because if the kids are not getting more than six hours of sleep, they’re not getting that period where they need REM sleep in order to show improvement. But as a scientist, it suggests that the brain is doing some very complicated work. It looks like it’s doing something here early in the night during slow-wave sleep and something else here late in the night, in REM sleep. So it looks like what we’ve got is a three-step process where you need training and then you need slow-wave sleep in that first quarter of the night and then REM sleep in the last quarter of the night. And if you don’t get all three of those then you don’t show improvement. And as a mathematical way, it suggests that the product of these two numbers will predict how much improvement you get and that’s, in fact, what we get with a group of subjects.

This is how much slow-wave, early, and REM sleep late they got and how much improvement. Notice that if they’re missing either the early slow-wave or the late REM sleep, there’s no improvement. But the more of them they get, the better they perform. And this number up here, which is a statistical geek’s number basically says that 80 percent of the difference that we see between subjects can be predicted just by the quality of their sleep. So I tell my Harvard students, it doesn’t matter where you went to prep school, it doesn’t matter what your SAT scores were, it doesn’t matter how much your dad makes--or your mom nowadays tends to make more--it depends on how well you slept the night after we trained you. About 20 percent of the variation can be explained by all those other factors, but 80 percent just depends on the quality of your sleep.

Now, one of the things we try to do is to see whether, if we pushed kids hard enough, they could actually show improvement without sleep. And what we found was the exact opposite, that when we pushed kids harder to learn, they, in fact got worse at re-testing, as opposed to better. And, in fact, in our original study, we had subjects come back and take this test four times. And the first time, they were pretty okay about it and the second time was sort of okay. By the fourth time, they had to come in through metal detectors because they were really unhappy with it. And their performance would get worse and worse.

What we found is two things. Remember when we showed those examples, those diagonal bars were in the lower left part of the screen. If we brought them back at the end of the day and just flipped them over to the lower right side then they could do it again. It wasn’t a generalized fatigue. It was a fatigue for targets exactly in the area where they had been studying it the hardest, but you could do it by flipping it to the other side or you could do it by giving them a nap.

So if you gave them a nap--and in this particular study, they either got 60 minutes or 90 minutes at two o’clock in the afternoon then we brought them back at 7:00 p.m., hours later, and retested them and that deterioration was completely gone. They showed no significant difference. They were as good as they were in the morning. So that sense of feeling burnt out that you get after working too hard was gone after this short nap.

And interestingly, this is the case if the nap had slow-wave sleep in it, but if it had both slow-wave sleep and REM, then they actually improve in the evening. And they improve statistically as much as they would from a night of sleep. So it looks like napping, in some way, although it’s much shorter than a night’s sleep, can provide a lot more benefit. And it’s a place where we really still are scratching our heads to try to understand how that works.

So again, on the naps, it looks like slow-wave sleep and REM sleep are required, although now in a much compressed period of time. And because the slow-wave sleep seems to be enough to prevent the deterioration, but not the improvement, it looks like we might be using that slow-wave sleep to stabilize the memory after the initial training and it’s the REM sleep that actually leads to that enhancement and improvement. And you need both of these functions working if you’re going to actually get better.

So that’s the perceptual task. Where else can we find it? Well, we can find it with motor learning too. And this is one of those places where I have the honor and the privilege of proving that my mother was right. It’s a very frustrating sort of a career. We do that repeatedly. Many of you will have the experience of learning a musical instrument and practicing some passage in a musical instrument and practicing and practicing and just be unable to get it. And finally, you just said, forget it, I can’t do it and you walk away from the piano, you put the violin down. You come back the next day, you pick it up and the first time through, you can do it perfectly. It’s sort of like magic.

Well, it turns out it’s not magic, it’s sleep-dependent consolidation of motor learning. And my post-doc, now colleague Matt Walker was the one who came up with this task. Rather than trying to teach someone the piano, we do a much more simple task. We have them sit down at the typewriter, the computer, and just type the sequence 41324, over and over again as fast as they can. And if you do this for 30 seconds and then give them a 30-second rest and do it another 30 seconds and do that across 12 times, they’ll get about 60 percent faster and then they kind of plateau here. They’re still a little wise maybe, but basically, they’ve learned as much as they can learn in this session. And, in fact, we can give them a ten-minute break and bring them back and they don’t get any better out there either. We can do it in the morning, we can do it in the evening. It all looks the same.

If though, instead of bringing them back 10 minutes later, we want to look later, we do the following. We take the last two trials and we say, okay, when they finish, this is how good they were, about 22, little more than 22 of those sequences in 30 seconds. Bring them back 12 hours later and how are they doing? Well, if we train them in the evening and bring them back the next morning, they’re about 20 percent faster. Nothing’s been done in-between, no practice. We’ve even made them wear mittens in-between so that they can’t practice. We had some exceptions in the 12 hours, they could take the mittens off. We won’t talk about that.

If we give them another 12 hours during which they’re awake, nothing much happens at all. Now, that could be sleep or that could be time, but if we train them in the morning and bring them back 12 hours later in the evening, we don’t see any robust improvement. Now, we let them go to sleep after 12 hours and now you see the big improvement.

So it turns out that it’s sleep rather than some specific amount of time that’s necessary to produce this large and dramatic improvement. This is the improvement in speed. They actually also make about 30 to 40 percent fewer errors. All of this free of charge, just get a good night’s sleep. Thanks, Mom.

Now, in this case, we were surprised to find that slow-wave sleep and REM sleep didn’t correlate with the improvement at all. In this case, it’s that light non-REM stage to sleep that seems to correlate with improvement. And again, motor tasks have different characteristics. It’s a sequence of movements you have to make over time, you need a temporal integration that might be better served by this different stage of sleep.

Now, just as a sidelight, we’ve started looking at some chemical populations and we see some fascinating things. This is that curve you saw before, morning and evening training. And I’ve now just put in those two trials the next morning that show the 20 percent improvement. We asked the question, what would happen with patients with schizophrenia. And when you train them, their training follows pretty much the same curve. They don’t do as well. They’re slower and they’re slower on almost everything, but you can see they’re following the same time course and, in fact, they show about a 70 or 80 percent improvement during training, but then it levels off. And the question is, okay, send them home, give them a night of sleep, bring them back the next morning and where are they. And they’re right there.

They show in those first two trials no improvement at all. And a guy said to me, said, well, just give them a little more time. So we gave them another ten trials and this is where they ended up. So schizophrenics, for some reason that we don’t understand--and these are medicated schizophrenics so we haven’t yet sorted that piece out--but they can learn while we train them, but they’re unable to develop the benefits of that night of sleep. And if this sleep is not only enhancing the speed, but making it more accurate and more automatic, we believe, this could contribute a lot to the sort of cognitive deficits we see in these patients who clinicians will often say they can learn, but they never really learn. And so that might turn out to be a failure of a sleep-dependent process, which would be quite surprising.

Now, how smart is sleep? I mean, what sorts of things can you learn in sleep and how clever is sleep in what you learn? We talked about this improvement you see overnight in typing this sequence 41324, but we can get a little more detailed and say where in that sequence does the improvement occur? So for example, here’s one subject who, when she was trained, was pretty fast. This is the time it took her to type the 1 after the 4 and then the 3 and the 2 and the 4. She was pretty fast on the first two, typing the 1 and the 3, but then she was quite slow typing the 2 and only a little bit better on the 4. So she had problems on one place in this test.

Now, here’s another subject who has a totally different profile. And when you look across a number of subjects, some people have problems in one place, some people have problems in another. And the question that my post-doc Kanichi Nanamura asked was where do you see improvement. Does everything just get faster? And it turns out that they get faster where they were slowest. It’s as if the brain is able to identify where you’re having problems during the day and focus its effort and focus its improvement on those particular places where you’re having problems.

If you look in large numbers of subjects, we tried 41324, we tried 5 digits on 2 hands, we tried 9 digits on 1 hand, we tried 9 digits on 2 hands. They all show the sleep-dependent improvement. This is how much improvement they show on the fastest number that they can type for each subject averaging their fastest number. And, you see, they don’t get any better at what they were best at, but when we take them and look at the slowest points in that sequence where they were having the most trouble typing, all four groups show a 20 to 25 percent improvement. So your brain is smart enough when you’re sleeping to know how to focus on the places that you’re having the greatest problems and help you sort of even out your learning.

Now, we’re still though just talking about these procedural tasks. A group in Germany, Wagner and Borne, recently published a very exciting paper, looking at the development of insight and creativity. This is for the high school students, okay? Here’s what you have to do in this particular test. You take the numbers two at a time and you follow these rules. If the two numbers are the same then write down that number. So the 1 and the 1 are the same, you put down 1. You notice that there’s only three digits here, 1s, 4s, and 9s. If the two digits are different, like the 1 here that the 4, if they’re different, put down the third digit. So what’s the third digit? Nine. Okay. Take the 9 and the 4. What do you put down?

Audience: One.

Stickgold: Nah.

Audience: Four.

Stickgold: One. Because these two are different, you put down the third one. So the one and the nine are different, you put…

Male: Four.

Stickgold: There we go. These guys are the same. Four. The four and the nine.

Audience: One.

Stickgold: One. And one and the four. Nine. And nine’s the answer. You got to type in nine. And then they give you another one and you get about 100 of these, one after another, okay? We then send them--they--I wish it had been me. They then sent them away, brought them back later and had them do another 100. And the rule was, do it the best way you can, okay?

Now, here’s the gimmick. If you look, you’ll notice that the last three digits here, 419, are the same as the three digits before it except backwards, okay? Now, what that means is that the last digit here is always going to be the same as the second one here. So that’s your answer. So if you figure this out, you only have to go this far and you’re done and you can get out of there about a half hour faster, okay? So they weren’t told that there was a faster way to do it, but they were told, do it the best way you can. So Borne and Wagner sent these people away, they trained them in the morning, they brought them back in the evening. About 25 percent of the subjects, over the course of the next 100 trials, figured out the shortcut, okay? He took another group, he trained them in the evening, kept them up all night, tested them in the morning. You see about 25 percent again figure it out. But if he trains them in the evening and lets them sleep through the night and tests them in the morning, 60 percent of them figure it out.

This is mathematical insight. As far as I’m concerned, this is about the most sophisticated and erudite type of creative logic and reasoning and thinking that people do. And astonishingly, getting a night of sleep, even when you don’t know that there’s a problem to solve, more than doubles the likelihood that you’ll come up with the answer the next day. So sleep is pretty smart.

How am I doing? Okay. So what I’ve been talking about so far is training people when they’re awake, letting them sleep a whole night, testing them again and looking for differences. We wanted to try to get into the brain and know more about what’s happening as the brain is sleeping. And what we wanted to do was test people on various cognitive tests while they were asleep, but they never cooperated, for some reason. So the best we could do was wake them up out of different sleep stages and run them through a very fast test.

And the test I want to talk to you about now is one that tests for associative memory. So if I were to ask you to group these into different categories or different pairs, it would make a lot of sense to put the Civil War and the Graf Hindenburg together. Those are historical events, they’re tragedies in one way or another. You might put these two together because whenever you go out to get your favorite dinner, you have to find a place to park. Those make sense.

But you would not be likely to put these two together, right? But you should. Can anybody figure out why? We got 750 people. Anybody know why those two go together? What--where did the Graf Hindenburg explode?

Audience: New Jersey.

Stickgold: But where? At its parking spot.

Audience: Oh, right.

Stickgold: Right? Now, that might sound goofy and silly, but isn’t that what your dreams are like? That we take these things that fit together in a crazy, goofy sort of a way that you would never think of doing if you were thinking logically. So what we wanted to do is wake people up out of different stages of sleep--and REM sleep is where you have your strongest and most intense dreaming--and see if we saw differences in these sorts of associations. And that’s what we’re going to find.

The test we actually use, you flash a word on the screen and then we flash another one that might or might not be a word. You press one of two buttons, yes, it’s a word, no, it’s not a word. And it turns out that if the words are strongly related to each other, even though you’re just looking at the second word, you’re faster identifying it than if they’re unrelated. And if they’re just kind of related, you get a speed in-between. And that makes sense because that’s how memory works. Whenever you hear a word or a phrase, your brain automatically starts to think about related words. So I say right, you think wrong, I think--I say tall, you think short, I think--I say fat, you think me. All these thoughts go through your mind sort of automatically.

So the test goes like this. You’re going to see one word. Read it as it goes by, look at the second one and decide is it a word or is it not a word, okay? Nice and easy, okay? And here’s another one. Okay. It’s a word, but it’s not related. Here’s another one. Not a word unless you live in Wister, Massachusetts. And there’s another one. Okay. And each time, you just press yes and no as fast as you can. And if we do this with a group of subjects with a long list of words that are strongly related or unrelated or weakly related, we find that if they’re weakly related, they’re about ten milliseconds faster. If they’re strongly related, they’re about 25 to 30 milliseconds faster. So we’re seeing bigger advantages with strongly-related word pairs like hot, cold. If we wake people up out of that light stage of non-REM sleep, the whole thing just seems to be down a bit. This is not significantly different from zero, probably is the same as zero. So you still get the strong priming. You don’t seem to get the week priming anymore. We wake people up from REM sleep and the whole thing turns upside-down on its head. It’s as if when you are in REM sleep, your brain can’t think of what it normally would, but only thinks of these weaker associations, these things that don’t fit together so well.

Now, you might wonder what use this can be. Well, first of all, it can explain again that feature of dream bizarreness, but it also can explain creativity because creativity is nothing more than the taking of two memories that you have, two pieces of information you have, and putting them together in a way that you’ve never thought of before. That’s what creativity is. And it might be that your brain, in REM sleep, is particularly tuned to try to search out and find those kinds of new associations. And that might be while dreams fill both bizarre and meaningful at the same time.

Well, I started to talk about dreams so let’s get to the fun part of the talk. For 100 years, since Freud, researchers have been trying to find ways to look at what causes dreams.

Forty-two percent of all the reports we got on the first night had images of skiing. And this is quite impressive. I mean, this is up near the levels that we see in things like post-traumatic stress disorder where patients have these recurring intrusive images that actually keep them from--make them fearful of going to sleep because their replays of the trauma. And we’ll help that with just this very simple manipulation with a game that they actually enjoy. We had three subjects that we just thought of this at the end who sat in chairs and watched somebody else play the game. And all three of them had images, which are kind of cute, but it also tells us more. It means that you don’t actually have to be playing and learning the game in order to trigger this process whereby your brain replays images of it as you’re falling asleep.

We also had the opportunity now to not just look at visual images, but the sensations, the feelings of it. And if you are--you’re all downhill skiers, of course, it’s California--so you’ve probably had the experience, many of you, of going skiing the first day out in the season and you get in bed at night and you’re back in the curves and you’re back on the slopes. And that’s exactly what we’re seeing with this game. So 11 of the players, about 70 percent, had those sleep onset images of the sensation, the physical sensation of the movement. And even one of the controls had physical sensations, even though he had never been on the game itself.

What do they see? I keep seeing all the places where I fall, I hit the walls, it’s kind of annoying. Then my legs fly up in the air, just like it was in the game itself. But here’s someone who can sort of feel the motions of the game, but more not really seeing it. And half of the subjects, half of those 16 were experienced downhill skiers. And half of that group gave us reports like, I envisioned myself skiing and for a second there, it felt like I was skiing backwards, something I used to attempt when I was younger. He’s not going to live very long, is he? But half of them reported images of memories of skiing from their past in actual situations. So again, the brain is going out and finding these tightly-related memories and replaying them as the person falls asleep.

Now, what is this good for? There’s an MTV theory of dreaming, which basically says before we had MTV, the only entertainment we had at night was dreams and it’s basically there just to entertain us. But I would like to argue that, in fact, this is part of those processes of processing memories to find new ways to put information together. For example in these cases, older skiing experiences with newer skiing experiences that you can be learning from the past as well as the present. Now, if that’s true then things should change over time.

So here’s the protocol that I just described to you. We would have subjects do skiing during the day and then when they went to bed, as they were falling asleep, whenever they started falling asleep, within three minutes of their falling asleep, we would wake them up and ask them what was going on and get a report. We did that for about 45 minutes. What we did with another group is we let them go to sleep normally and sleep all they wanted, just sleep solid through the first two hours of the night, which gets you through that whole first REM cycle and then we’d wake them up and they’d do a few math problems so we were sure they were awake and then we’d tell them to go back to sleep. And then for the next 45 minutes, as they were falling asleep, we would once again wake them up whenever they started to fall asleep, within the first three minutes. And the question we wanted to ask is, are the reports different now that they’ve had a couple of hours to sleep on it or are they going to look just like they did here at the start of the night?

Well, the most startling finding of the study was that with 12 subjects, we didn’t get a single report of skiing in the delayed onset. Whatever it was going on in their brain as they were first falling asleep that made these images, these memories so powerful and intrusive that they appeared in 42 percent of those sleep onset awakenings, after two hours of sleep, they were in some way diffused, weakened, made less intrusive so that you no longer got them. But what was more exciting than that is what we got instead. So on the average subject, we managed to get about 7 reports over these 45 minutes, about one 7 minutes.

And here’s the fourth report we got from one of these delayed onset subjects. I felt as though I was falling downhill and I was dreaming about, like, instructions to a young king or something. We asked him afterwards when he--the next day, no snow, no skis, just falling downhill. And you’ll notice this was his fourth report of the night. The first three, we couldn’t find any relationship. The next one, the fifth one, had no obvious relationship. Sixth time we wake him up, he’s right back there. I felt like I was sort of sliding downhill again and there were instructions on a person I don’t know. I mean, give him a break, it’s two o’clock in the morning, all right?

But once again, the brain has gone back to this same image, but it’s not an image of skiing. It’s an image of what it’s like to ski. So I don’t know if you noticed, in the clip I showed you Debbie was doing a slalom course, she was going down between flags stuck in the snow. Half the time, they do slalom runs and half the time they just do straight downhill skiing. And downhill skiing is a lot like sliding down a hill or falling down a hill.

But this is a report from another subject. I was having a rather vivid image as though I was moving forward through some kind of a forest, right, between vertical wooden trees. I was moving forward very stiffly and my entire upper body was incredibly straight. It felt almost as though I was moving forward on a conveyor belt without my legs actually moving at all. This is a beautiful description of the pose you hold as you downhill ski.

So what it looks like the brain is doing, now that it’s had the chance to sleep on it for a couple of hours, is ask the question, what is this like? What is it like to ski? Skiing is like rolling down a hill. Skiing is like zooming through a forest between trees. And I would argue that this is the process of creating meaning out of the events of your life. Creating meaning is again, one of those most sophisticated things that the human mind does, to understand things in the context of your life. And that’s what the brain seems to be trying to do during REM sleep.

So how can we think about this in a larger picture of the processing of events that happen during the day? Well, a sleep onset that brings physiology and chemistry seems to access neocortical elements. Those are the regions of the brain outside of the central core that’s required for episodic memories, where the pieces of episodic memories are stored. It seems to access those from recent episodic events. That’s what we’re seeing with the skiing. And again, the skiing, as with the Tetris, they don’t see themselves on the machine. They just see the characters skiing down. They don’t even see the big screen. They just see the character skiing down.

Maybe what’s happening here is that this replay of recent memories is tagging emotionally significant events from the day, sort of saying, pay attention to this later on. And by the way, you’ll notice, those of you who might once or twice in your life have had trouble falling asleep, it’s very often because emotionally-charged events from the day are starting to replay in your mind and winding you up again. They don’t have to be negatives things. Visions of sugarplums danced in their heads, right? I mean, the inability to fall asleep on Christmas Eve of kids is because they start to get these images of sugarplums dancing in their head and it cranks them up, right, and when you crank up, you’ve all had the experience, something starts to float into your mind and then it just grabs you emotionally and you know it’s going to be ten minutes before you calm down enough. Insomnia, in large part, might be the system not quite playing out properly so that you get aroused again before you fall asleep.

Now, during REM sleep, the physiology and chemistry of the brain changes completely. And I would argue that this change puts the brain in a state that facilitates the integration of these memories into what are known as cortical association networks, putting together of things so that it makes sense and so that things fit together into your life.

Remember in REM sleep, we only have access to these neocortical memories. The episodic memories seem to be shut off. And there’s actually data I didn’t show from rat studies that information doesn’t seem to flow out of those episodic memory systems during REM sleep. So you only have those pieces, those disjoint elements of the episodic memory from which to create the dream. So you can’t focus on the episodic memories. Instead, your brain is forced to analyze the meaning of the memories. And so we have to reorganize and integrate our memories of the event so we start to put the events of the day into a context.

But remember in REM sleep, the chemistry has changed away from these transmitters, which are normally involved in focus thinking so you don’t think in the clear way you would normally think. Remember that you’re now specifically activating weaker associates, not stronger associations. So again, your brain is going out and looking for new ways to think about events from the day. The regions of the brain I mentioned that are activated during REM sleep seem to facilitate emotional processing of memories. And Jim McGaugh’s book that you can buy out there in the lobby afterwards, I believe titled Memories and Emotions--or if not, close enough. So REM sleep is a period where that type of processing may specifically be enhanced. And again, information during REM sleep is not flowing out of those brain regions that control episodic memories.

So I want to leave you with just some summary statements, that the physiology and chemistry of the brain change dramatically across the night, that sleep consolidates and enhances memory, that it changes the way that associative memories are processed as you’re doing this work during the night, and that dreams are part of the story.

So if I can finish, I want to paraphrase Robert Frost who almost said two roads diverged in the brain. REM sleep takes the one less traveled by and that makes all the difference. Thank you.