• • •
The boy’s brain was going haywire but he was fast asleep, out cold. His father called his name: Armond? Armond? No response. Was he pretending? No, it sure didn’t look that way.
It was December 1951, and Eugene Aserinsky, a young graduate student at the University of Chicago, had brought his eight-year-old son, Armond, to his basement lab to perform an experiment on sleep. Aserinsky was studying for a degree in physiology and trying to build his credentials as an experimental scientist; he had little interest in sleep research as a career. He was only here pulling night duty, on orders from his academic advisor, Nathaniel Kleitman, who happened to be the father of modern sleep science. Aserinsky had been tinkering with a machine called an Offner Dynograph to track the sleeping brain. A forerunner to the EEG, the Dynograph registers electrical signals from the brain, through electrodes taped to the skull. Aserinsky was using Armond as his test subject. He’d taped a couple of electrodes to the boy’s head and eyelids (to track their motion) and then tuned the machine from the next room, asking his son to look this way and that, calibrating the dials. Gradually, Armond nodded off and Aserinsky, sipping his coffee, watched as the Dynograph settled, its ink pens tracing smaller, smoother waves, as expected. But after a few hours the waves began to spike—all of them, those coming from Armond’s eyelids as well as his brain—as if the boy was awake and alert. Aserinsky got up from his chair and slipped into the room where his son lay, to make sure his son was asleep and safe.
Armond?… Armond? No answer.
Aserinsky returned to the next room, and watched the Dynograph. Scientists at the time considered sleep a period when the brain essentially shut down, becoming a playground for the unconscious, a canvas for dreams. The Dynograph said differently. Aserinsky paced the lab—“flabbergasted,” he would say later, by the frenzied wave activity—and watched as Armond’s brain waves settled down again, the pens ceasing their chatter. It was late, there was no one else around. Was he seeing things? If so, then reporting the finding would be potentially embarrassing, written off as the misplaced exuberance of an inexperienced researcher. If not, his son’s sleeping brain could be telling him something that no one suspected about unconsciousness.
He brought Armond back into the lab for another session weeks later, to see if his original observation was a fluke. It wasn’t. At various periods during the night, Armond’s brain leapt to life as if he were wide awake. Aserinsky was now confident that this pattern was no mirage. “The question was, what was triggering these eye movements?” he said years later. “What do they mean?”
He didn’t have enough expertise in the field or its experimental techniques to know. He’d have to go to the top—to Kleitman—and ask whether such odd brain activity had been reported in sleep experiments before, and whether it was worth the time to follow up. Kleitman didn’t hesitate. “Study more people,” he told Aserinsky. “You might be on to something.”
By late 1952, Aserinsky had upgraded his equipment and embarked on a study of two dozen adults. Their brain patterns looked just like Armond’s: periods of slow undulations, punctuated by bursts of intense activity. The flare-ups had no precedent in the sleep research literature, so he wasn’t even sure what to call them. He consulted Kleitman again, and the two of them reviewed the data. If they were going to report such an unusual finding and claim it was universal, they’d better be sure of their measurements.
Their report finally appeared in September of 1953 in the journal Science. The paper was all of two pages, but Aserinsky and Kleitman did not undersell the implications of their work. “The fact that these eye movements, this EEG pattern, and autonomic nervous system activity are significantly related and do not occur randomly suggests that these physiological phenomena, and probably dreaming, are very likely all manifestations of a particular level of cortical activity which is encountered normally during sleep,” they concluded. “An eye movement period first appears about three hours after going to sleep, recurs two hours later, and then emerges at somewhat closer intervals a third or fourth time shortly prior to awakening.” They eventually settled on a more scientific-sounding name for the phenomenon: rapid eye movement, or REM, sleep.
“This was really the beginning of modern sleep research, though you wouldn’t have known it at the time,” William Dement, then a medical student in Kleitman’s lab and now a professor of psychiatry and sleep medicine at Stanford University, told me. “It took years for people to realize what we had.”
One reason for the delay was lingering infatuation with an old theory. In the 1950s many brain scientists, particularly in the United States, were still smitten with Freud’s idea that dreams are wish fulfillment, played out in fantasy and symbolic imagery that’s not accessible during waking. Money poured into sleep research but it was used to investigate the content of dreams during REM, not the mechanics or purpose of REM per se—and to little avail. People roused from REM described a tangle of anxieties, fantasies, and nonsense scenes that said nothing consistent about human nature. “It was exciting work to do, but in the end we weren’t able to say anything conclusive,” Dement told me. Still, those dream studies and others confirmed beyond any doubt that REM was universal and occurred periodically through the night, alternating with other states of unconsciousness. In fact, people typically experience four or five bursts of REM during the night—of twenty to thirty minutes in duration—as the brain swims up to the brink of consciousness before diving back down again. By 1960, sleep scientists began to speak of sleep as having at least two dimensions: REM and non-REM, or NREM.
Later, using EEG recordings as well as more specific electrical recordings from the eyes and eyelids, researchers found that NREM has its own distinct stages as well. The definition of these stages is arbitrary, depending mostly on the shape and frequency of the waves. The light sleep that descends shortly after we doze off was called Stage 1; this is when the brain’s jagged waves of conscious awareness begin to soften. In Stage 2, the waves become more regular, resembling a sine wave, or a clean set of rollers moving toward shore on a windless day. In Stages 3 and 4, the waves gradually stretch out, until they undulate gently like a swell over open ocean, a slow-wave pattern that signals the arrival of deep sleep. The brain cycles though its five sleep stages in order: from Stage 1 down to Stage 2, deeper to Stage 3, and bottoming out at Stage 4, before floating back up, through Stages 3 and 2, and then into REM. The cycle then repeats throughout the night, dropping down again to Stage 4 and back up, to REM. These four stages and REM describe what scientists call sleep architecture, which maps easily onto a graph:
The discovery and description of this previously hidden architecture did more than banish the notion, once and for all, that our brains simply “power down” at night, becoming vessels for dreams. It also begged a question: If the brain is so active while we sleep, what’s it up to, exactly? Nature doesn’t waste resources on this scale. With its bursts of REM and intricate, alternating layers of wave patterns, the brain must be up to something during sleep. But what?
“To do science, you have to have an idea, and for years no one had one,” J. Allan Hobson, a psychiatry professor at Harvard, told me. “They saw sleep as nothing but an annihilation of consciousness. Now we know different.”
• • •
One reason that palace intrigue makes for such page-turning fiction or addictive TV is what psychologists call “embedded hierarchy.” The king is the king, the queen the queen, and there are layers of princes, heirs, relatives, ladies-in-waiting, meddling patriarchs, ambitious newcomers, and consigliere types, all scheming to climb to the top. Which alliances are most important? What’s the power hierarchy? Who has leverage over whom? You have no idea until you see the individuals interact. And if you don’t see them square off one-on-one, you play out different scenarios to see if you can judge the players’ relative power. Could Grishilda have Thorian shackled and tossed in the moat if the two clashed? She is a favorite of the king’s, after all. Y
et Thorian might have some connections up his sleeve … wait, who’s his mother again?
Learning scientists like embedded hierarchy problems because they model the sort of reasoning we have to do all the time, to understand work politics as well as math problems. We have to remember individual relationships, which is straight retention. We have to use those to induce logical extensions: if A > B and B > C, then A must be > C. Finally, we need to incorporate those logical steps into a larger framework, to deduce the relationships between people or symbols that are distantly related. When successful, we build a bird’s-eye view, a system to judge the relationship between any two figures in the defined universe, literary or symbolic, that’s invisible to the untrained mind.
In a 2007 study, researchers at Harvard and McGill universities tested college students’ ability to discern an embedded hierarchy in what looked like a simple game. The research team asked the students to study pairs of colored eggs, one pair at a time, on a computer screen. The eggs were ranked one over another. For example:
The students were split into two groups: one studied the eggs in the morning, one studied them in the evening. Both groups memorized the relative ranks of the pairs quickly and aced a test on them just afterward. But twelve hours later, the groups got another test, asking them to rank eggs they’d not seen directly compared. This is the “embedded” Grishilda-Thorian question, and the answer is not so obvious. If aqua trumps rainbow, does that mean it also trumps paisley? And what about coral? Does it rank third, or fourth? The students never got to see the entire ranking of all the eggs while studying, so it was hazy.
It was hazy, that is, until they slept on it.
The group that studied in the evening and took the test the next morning after a night’s sleep—the “sleep group,” as they were called—scored 93 percent on the most distantly related pair, i.e., the hardest question. The group that studied in the morning and took the test in the evening, without having slept—the “wake group”—scored 69 percent. A full twenty-four hours later, each student took the test yet again, and the sleep group’s advantage had increased on the most distantly related pairs. That’s a large difference on the hardest questions—35 percent, separating one kind of student from another—but it’s not unusual in studies of sleep and learning. “We think what’s happening during sleep is that you open the aperture of memory and are able to see this bigger picture,” the study’s senior author, Matthew Walker, told me. “There is evidence, in fact, that REM is this creative memory domain when you build different associations, combine things in different ways and so on.”
In a game like this one, he and his coauthors argue, we are very good at building separate categories of associations (aqua over rainbow, paisley over coral), but the more obscure relationships between those categories are harder to sort out—until we sleep.
The investigation of sleep as consolidator of learning is still a work in progress. After scientists chasing Freud hit a wall in the 1960s, sleep research, like its nocturnal subjects, dropped off the deep end. The money tapered off. The window Eugene Aserinsky had opened, revealing REM sleep, seemed, for a time, to expose little more than another dark room. “You had this great excitement, basically followed by forty years of nothing; it was just horrible,” Robert Stickgold, a neuroscientist at Harvard, told me. But in the past two decades, dozens of studies like Walker’s have brightened the horizon, turning sleep into one of the most promising—and contentious—frontiers of learning science. The preponderance of evidence to date finds that sleep improves retention and comprehension of what was studied the day before, and not just for colored eggs. It works for vocabulary. Word pairs. Logical reasoning, similar to what’s taught in middle school math. Even the presentation you’ll be giving at work, or the exam that’s coming up at school. For all of these, you need to memorize the details of important points and to develop a mental map of how they fit together. The improvements tend to be striking, between 10 and 30 percent, and scientists don’t understand the dynamics of unconscious states well enough yet to explain why.
My own theory is that sleep amplifies many of the techniques we’ve discussed in this book. The spacing effect described in chapter 4, for instance, is especially strong with intervals of a day or two (plus sleep). Philip Ballard’s “reminiscence”—that puzzling improvement in memory of “The Wreck of the Hesperus” poem described in chapter 2—crested in the first day or two. A good night’s sleep could surely loosen the “fixedness” that makes it hard to see a solution to the Pencil Problem, discussed in chapter 6, right away. The brain is likely doing many of the same things with information while asleep as it does while awake—or at least performing complementary functions.
The story hardly ends there, however.
Scientists have begun to study the effects of interrupting particular stages of sleep, like REM, to isolate the impact those stages have on learning specific skills or topics. Remember, sleep has five dimensions that we know of: REM, and the four stages surrounding it. Our brain waves have distinct patterns in each of those periods, suggesting that different mental dynamics are at work in each one. Could it be that each stage is specialized to consolidate a specific kind of skill, whether it’s a geometric proof, a writing assignment, or a tennis serve? Many scientists now suspect so, based on evidence that comes from both animals and humans. These findings have coalesced into a remarkable hypothesis, first described in 1995 by Italian scientists led by Antonio Giuditta at the University of Naples Federico II. The idea has since been fleshed out by others, mostly Robert Stickgold at Harvard and Carlyle Smith of Trent University in Peterborough, Ontario, who have contributed enough experimental heft to make this model of sleep learning a full-grown theory, the most comprehensive explanation yet for how the different stages of sleep consolidate memory.
Technically, I suppose, we should call this idea the Giuditta-Smith-Stickgold Model of Learning Consolidation. I prefer to call it, simply, the Night Shift Theory. The lights go out, and basic maintenance is done. Here’s what the Night Shift Theory says happens overnight, during each stage:
Stage 1: This one is a scratch. It’s impossible to deprive people of Stage 1 light sleep, if they’re going to sleep at all. Its role in consolidating memories is hard to isolate, though it’s often laced with REM-like periods.
REM: These storms of neural firing appear to aid pattern recognition, as in the colored egg experiment, as well as creative problem solving and perceiving relationships that weren’t apparent during the day, as in a difficult calculus problem. It likely plays the largest role, of all the stages, in aiding percolation. People still get these benefits from sleep sans REM—just not to the same degree. REM is also involved in interpreting emotionally charged memories. “We believe that it’s during REM that the brain strips away the visceral feeling experienced at the time an emotional memory is formed,” Matthew Walker, the Berkeley brain scientist who coauthored the colored egg study, told me, “but holds on to the actual information, the details, the where and when of what happened.” That panic you felt the last time you opened a geometry exam? It’s better to have that feeling “stripped”—or at least reduced—so you can recall what the panic-inducing problems actually were. Walker describes REM as “a nighttime therapy session.”
Stage 2: This is the motor memory specialist. In a series of little-known studies, Carlyle Smith trained people in what he calls the “rotor task.” This is a hand-eye coordination exercise in which people have to use their nonwriting hand to chase a moving spotlight across a computer screen using a joystick. It’s easy enough to improve and people generally do—but not as quickly if they’re deprived of Stage 2 sleep. “Stage 2 seems to be the single most critical stage for motor learning,” Smith told me. “When we deprive people of Stage 2, we don’t see that same level of improvement, and we believe the findings extend to all types of motor learning, whether it’s music or athletics and possibly mechanical skills.”
Stages 3 and 4: These two are usually l
umped together in learning research as slow-wave or deep sleep. This is prime retention territory. Starve people of deep slumber, and it doesn’t just dim their beauty; they don’t get the full benefit of sleep-aided recall of newly learned facts, studied vocabulary, names, dates, and formulas. “We have a lot of evidence that slow-wave is important for declarative memory consolidation, and that this doesn’t happen as much in REM,” Stickgold told me.
To put all this in some perspective, let’s dial up the sleep architecture graph once more.
The first thing to note about this diagram is that it traces the architecture for a person who, in this case, goes to sleep at 11 P.M. and wakes up at 7 A.M. The architecture looks roughly the same for everyone, though, no matter what time he or she regularly goes to bed and wakes up. In an important sense, getting the usual doses of all five stages is the meaning of a full night’s sleep. Each stage somehow complements the others’ work. Where it really gets interesting is when we alter our usual sleep schedule to prepare for some performance, whether a speech, a tryout, or an exam.
Notice, for example, that the longest stretch of Stage 2 sleep is just before waking. Cut that short and you miss out on the period when your brain is consolidating a skateboarding move, a difficult piano fingering, or your jump shot. “The implication is that if you are preparing for a performance—a music recital, say—it’s better to stay up late than get up early,” Smith told me. “These coaches that have athletes or other performers up at five o’clock in the morning, I think that’s crazy.”
The same logic applies to REM. The largest dose is in the early morning, between those chunks of Stage 2. If you’re prepping for a math or chemistry test, an exam that’s going to strain your ability to detect patterns, better to stay up late and, if possible, hit the snooze button in the morning. Let the cock crow till he’s hoarse.
How We Learn Page 20