Why We Sleep

by Matthew Walker

  SELF-IDENTIFYING SLEEP

  Your lightning-quick judgment of Jessica being asleep was likely correct. And perhaps you accidentally confirmed it by knocking something over and waking her up. Over time, we have all become incredibly good at recognizing a number of signals that suggest that another individual is asleep. So reliable are these signs that there now exists a set of observable features that scientists agree indicate the presence of sleep in humans and other species.

  The Jessica vignette illustrates nearly all of these clues. First, sleeping organisms adopt a stereotypical position. In land animals, this is often horizontal, as was Jessica’s position on the couch. Second, and related, sleeping organisms have lowered muscle tone. This is most evident in the relaxation of postural (antigravity) skeletal muscles—those that keep you upright, preventing you from collapsing to the floor. As these muscles ease their tension in light and then deep sleep, the body will slouch down. A sleeping organism will be draped over whatever supports it underneath, most evident in Jessica’s listing head position. Third, sleeping individuals show no overt displays of communication or responsivity. Jessica showed no signs of orienting to you as you entered the room, as she would have when awake. The fourth defining feature of sleep is that it’s easily reversible, differentiating it from coma, anesthesia, hibernation, and death. Recall that upon knocking the item over in the room, Jessica awoke. Fifth, as we established in the previous chapter, sleep adheres to a reliable timed pattern across twenty-four hours, instructed by the circadian rhythm coming from the brain’s suprachiasmatic nucleus pacemaker. Humans are diurnal, so we have a preference for being awake throughout the day and sleeping at night.

  Now let me ask you a rather different question: How do you, yourself, know that you have slept? You make this self-assessment even more frequently than that of sleep in others. Each morning, with luck, you return to the waking world knowing that you have been asleep.fn1 So sensitive is this self-assessment of sleep that you can go a step further, gauging when you’ve had good- or bad-quality sleep. This is another way of measuring sleep—a first-person phenomenological assessment distinct from signs that you use to determine sleep in another.

  Here, also, there are universal indicators that offer a convincing conclusion of sleep—two, in fact. First is the loss of external awareness—you stop perceiving the outside world. You are no longer conscious of all that surrounds you, at least not explicitly. In actual fact, your ears are still “hearing”; your eyes, though closed, are still capable of “seeing.” This is similarly true for the other sensory organs of the nose (smell), the tongue (taste), and the skin (touch).

  All these signals still flood into the center of your brain, but it is here, in the sensory convergence zone, where that journey ends while you sleep. The signals are blocked by a perceptual barricade set up in a structure called the thalamus (THAL-uh-muhs). A smooth, oval-shaped object just smaller than a lemon, the thalamus is the sensory gate of the brain. The thalamus decides which sensory signals are allowed through its gate, and which are not. Should they gain privileged passage, they are sent up to the cortex at the top of your brain, where they are consciously perceived. By locking its gates shut at the onset of healthy sleep, the thalamus imposes a sensory blackout in the brain, preventing onward travel of those signals up to the cortex. As a result, you are no longer consciously aware of the information broadcasts being transmitted from your outer sense organs. At this moment, your brain has lost waking contact with the outside world that surrounds you. Said another way, you are now asleep.

  The second feature that instructs your own, self-determined judgment of sleep is a sense of time distortion experienced in two contradictory ways. At the most obvious level, you lose your conscious sense of time when you sleep, tantamount to a chronometric void. Consider the last time you fell asleep on an airplane. When you woke up, you probably checked a clock to see how long you had been asleep. Why? Because your explicit tracking of time was ostensibly lost while you slept. It is this feeling of a time cavity that, in waking retrospect, makes you confident you’ve been asleep.

  But while your conscious mapping of time is lost during sleep, at a non-conscious level, time continues to be cataloged by the brain with incredible precision. I’m sure you have had the experience of needing to wake up the next morning at a specific time. Perhaps you had to catch an early-morning flight. Before bed, you diligently set your alarm for 6:00 a.m. Miraculously, however, you woke up at 5:58 a.m., unassisted, right before the alarm. Your brain, it seems, is still capable of logging time with quite remarkable precision while asleep. Like so many other operations occurring within the brain, you simply don’t have explicit access to this accurate time knowledge during sleep. It all flies below the radar of consciousness, surfacing only when needed.

  One last temporal distortion deserves mention here—that of time dilation in dreams, beyond sleep itself. Time isn’t quite time within dreams. It is most often elongated. Consider the last time you hit the snooze button on your alarm, having been woken from a dream. Mercifully, you are giving yourself another delicious five minutes of sleep. You go right back to dreaming. After the allotted five minutes, your alarm clock faithfully sounds again, yet that’s not what it felt like to you. During those five minutes of actual time, you may have felt like you were dreaming for an hour, perhaps more. Unlike the phase of sleep where you are not dreaming, wherein you lose all awareness of time, in dreams, you continue to have a sense of time. It’s simply not particularly accurate—more often than not dream time is stretched out and prolonged relative to real time.

  Although the reasons for such time dilation are not fully understood, recent experimental recordings of brain cells in rats give tantalizing clues. In the experiment, rats were allowed to run around a maze. As the rats learned the spatial layout, the researchers recorded signature patterns of brain-cell firing. The scientists did not stop recording from these memory-imprinting cells when the rats subsequently fell asleep. They continued to eavesdrop on the brain during the different stages of slumber, including rapid eye movement (REM) sleep, the stage in which humans principally dream.

  The first striking result was that the signature pattern of brain-cell firing that occurred as the rats were learning the maze subsequently reappeared during sleep, over and over again. That is, memories were being “replayed” at the level of brain-cell activity as the rats snoozed. The second, more striking finding was the speed of replay. During REM sleep, the memories were being replayed far more slowly: at just half or quarter the speed of that measured when the rats were awake and learning the maze. This slow neural recounting of the day’s events is the best evidence we have to date explaining our own protracted experience of time in human REM sleep. This dramatic deceleration of neural time may be the reason we believe our dream life lasts far longer than our alarm clocks otherwise assert.

  AN INFANT REVELATION—TWO TYPES OF SLEEP

  Though we have all determined that someone is asleep, or that we have been asleep, the gold-standard scientific verification of sleep requires the recording of signals, using electrodes, arising from three different regions: (1) brainwave activity, (2) eye movement activity, and (3) muscle activity. Collectively, these signals are grouped together under the blanket term “polysomnography” (PSG), meaning a readout (graph) of sleep (somnus) that is made up of multiple signals (poly).

  It was using this collection of measures that arguably the most important discovery in all of sleep research was made in 1952 at the University of Chicago by Eugene Aserinsky (then a graduate student) and Professor Nathaniel Kleitman, famed for the Mammoth Cave experiment discussed in chapter 2.

  Aserinsky had been carefully documenting the eye movement patterns of human infants during the day and night. He noticed that there were periods of sleep when the eyes would rapidly dart from side to side underneath their lids. Furthermore, these sleep phases were always accompanied by remarkably active brainwaves, almost identical to those observed from a brai
n that is wide awake. Sandwiching these earnest phases of active sleep were longer swaths of time when the eyes would calm and rest still. During these quiescent time periods, the brainwaves would also become calm, slowly ticking up and down.

  As if that weren’t strange enough, Aserinsky also observed that these two phases of slumber (sleep with eye movements, sleep with no eye movements) would repeat in a somewhat regular pattern throughout the night, over, and over, and over again.

  With classic professorial skepticism, his mentor, Kleitman, wanted to see the results replicated before he would entertain their validity. With his propensity for including his nearest and dearest in his experimentation, he chose his infant daughter, Ester, for this investigation. The findings held up. At that moment Kleitman and Aserinsky realized the profound discovery they had made: humans don’t just sleep, but cycle through two completely different types of sleep. They named these sleep stages based on their defining ocular features: non–rapid eye movement, or NREM, sleep, and rapid eye movement, or REM, sleep.

  Together with the assistance of another graduate student of Kleitman’s at the time, William Dement, Kleitman and Aserinsky further demonstrated that REM sleep, in which brain activity was almost identical to that when we are awake, was intimately connected to the experience we call dreaming, and is often described as dream sleep.

  NREM sleep received further dissection in the years thereafter, being subdivided into four separate stages, unimaginatively named NREM stages 1 to 4 (we sleep researchers are a creative bunch), increasing in their depth. Stages 3 and 4 are therefore the deepest stages of NREM sleep you experience, with “depth” being defined as the increasing difficulty required to wake an individual out of NREM stages 3 and 4, compared with NREM stages 1 or 2.

  THE SLEEP CYCLE

  In the years since Ester’s slumber revelation, we have learned that the two stages of sleep—NREM and REM—play out in a recurring, push-pull battle for brain domination across the night. The cerebral war between the two is won and lost every ninety minutes,fn2 ruled first by NREM sleep, followed by the comeback of REM sleep. No sooner has the battle finished than it starts anew, replaying every ninety minutes. Tracing this remarkable roller-coaster ebb and flow across the night reveals the quite beautiful cycling architecture of sleep, depicted in figure 8.

  On the vertical axis are the different brain states, with Wake at the top, then REM sleep, and then the descending stages of NREM sleep, stages 1 to 4. On the horizontal axis is time of night, starting on the left at about eleven p.m. through until seven a.m. on the right. The technical name for this graphic is a hypnogram (a sleep graph).

  Figure 8: The Architecture of Sleep

  Had I not added the vertical dashed lines demarcating each ninety-minute cycle, you may have protested that you could not see a regularly repeating ninety-minute pattern. At least not the one you were expecting from my description above. The cause is another peculiar feature of sleep: a lopsided profile of sleep stages. While it is true that we flip-flop back and forth between NREM and REM sleep throughout the night every ninety minutes, the ratio of NREM sleep to REM sleep within each ninety-minute cycle changes dramatically across the night. In the first half of the night, the vast majority of our ninety-minute cycles are consumed by deep NREM sleep, and very little REM sleep, as can be seen in cycle 1 of the figure above. But as we transition through into the second half of the night, this seesaw balance shifts, with most of the time dominated by REM sleep, with little, if any, deep NREM sleep. Cycle 5 is a perfect example of this REM-rich type of sleep.

  Why did Mother Nature design this strange, complex equation of unfolding sleep stages? Why cycle between NREM and REM sleep over and over? Why not obtain all of the required NREM sleep first, followed by all of the necessary REM sleep second? Or vice versa? If that’s too much a gamble on the off chance that an animal only obtains a partial night of sleep at some point, then why not keep the ratio within each cycle the same, placing similar proportions of eggs in both baskets, as it were, rather than putting most of them in one early on, and then inverting that imbalance later in the night? Why vary it? It sounds like an exhausting amount of evolutionary hard work to have designed such a convoluted system, and put it into biological action.

  We have no scientific consensus as to why our sleep (and that of all other mammals and birds) cycles in this repeatable but dramatically asymmetric pattern, though a number of theories exist. One theory I have offered is that the uneven back-and-forth interplay between NREM and REM sleep is necessary to elegantly remodel and update our neural circuits at night, and in doing so manage the finite storage space within the brain. Forced by the known storage capacity imposed by a set number of neurons and connections within their memory structures, our brains must find the “sweet spot” between retention of old information and leaving sufficient room for the new. Balancing this storage equation requires identifying which memories are fresh and salient, and which memories that currently exist are overlapping, redundant, or simply no longer relevant.

  As we will discover in chapter 6, a key function of deep NREM sleep, which predominates early in the night, is to do the work of weeding out and removing unnecessary neural connections. In contrast, the dreaming stage of REM sleep, which prevails later in the night, plays a role in strengthening those connections.

  Combine these two, and we have at least one parsimonious explanation for why the two types of sleep cycle across the night, and why those cycles are initially dominated by NREM sleep early on, with REM sleep reigning supreme in the second half of the night. Consider the creation of a piece of sculpture from a block of clay. It starts with placing a large amount of raw material onto a pedestal (that entire mass of stored autobiographical memories, new and old, offered up to sleep each night). Next comes an initial and extensive removal of superfluous matter (long stretches of NREM sleep), after which brief intensification of early details can be made (short REM periods). Following this first session, the culling hands return for a second round of deep excavation (another long NREM-sleep phase), followed by a little more enhancing of some fine-grained structures that have emerged (slightly more REM sleep). After several more cycles of work, the balance of sculptural need has shifted. All core features have been hewn from the original mass of raw material. With only the important clay remaining, the work of the sculptor, and the tools required, must shift toward the goal of strengthening the elements and enhancing features of that which remains (a dominant need for the skills of REM sleep, and little work remaining for NREM sleep).

  In this way, sleep may elegantly manage and solve our memory storage crisis, with the general excavatory force of NREM sleep dominating early, after which the etching hand of REM sleep blends, interconnects, and adds details. Since life’s experience is ever changing, demanding that our memory catalog be updated ad infinitum, our autobiographical sculpture of stored experience is never complete. As a result, the brain always requires a new bout of sleep and its varied stages each night so as to auto-update our memory networks based on the events of the prior day. This account is one reason (of many, I suspect) explaining the cycling nature of NREM and REM sleep, and the imbalance of their distribution across the night.

  A danger resides in this sleep profile wherein NREM dominates early in the night, followed by an REM sleep dominance later in the morning, one of which most of the general public are unaware. Let’s say that you go to bed this evening at midnight. But instead of waking up at eight a.m., getting a full eight hours of sleep, you must wake up at six a.m. because of an early-morning meeting or because you are an athlete whose coach demands early-morning practices. What percent of sleep will you lose? The logical answer is 25 percent, since waking up at six a.m. will lop off two hours of sleep from what would otherwise be a normal eight hours. But that’s not entirely true. Since your brain desires most of its REM sleep in the last part of the night, which is to say the late-morning hours, you will lose 60 to 90 percent of all your REM sleep, even though you a
re losing 25 percent of your total sleep time. It works both ways. If you wake up at eight a.m., but don’t go to bed until two a.m., then you lose a significant amount of deep NREM sleep. Similar to an unbalanced diet in which you only eat carbohydrates and are left malnourished by the absence of protein, short-changing the brain of either NREM or REM sleep—both of which serve critical, though different, brain and body functions—results in a myriad of physical and mental ill health, as we will see in later chapters. When it comes to sleep, there is no such thing as burning the candle at both ends—or even at one end—and getting away with it.

  HOW YOUR BRAIN GENERATES SLEEP

  If I brought you into my sleep laboratory this evening at the University of California, Berkeley, placed electrodes on your head and face, and let you fall asleep, what would your sleeping brainwaves look like? How different would those patterns of brain activity be to those you are experiencing right now, as you read this sentence, awake? How do these different electrical brain changes explain why you are conscious in one state (wake), non-conscious in another (NREM sleep), and delusionally conscious, or dreaming, in the third (REM sleep)?

  Figure 9: The Brainwaves of Wake and Sleep

  Assuming you are a healthy young/midlife adult (we will discuss sleep in childhood, old age, and disease a little later), the three wavy lines in figure 9 reflect the different types of electrical activity I would record from your brain. Each line represents thirty seconds of brainwave activity from these three different states: (1) wakefulness, (2) deep NREM sleep, and (3) REM sleep.

  Prior to bed, your waking brain activity is frenetic, meaning that the brainwaves are cycling (going up and down) perhaps thirty or forty times per second, similar to a very fast drumbeat. This is termed “fast frequency” brain activity. Moreover, there is no reliable pattern to these brainwaves—that is, the drumbeat is not only fast, but also erratic. If I asked you to predict the next few seconds of the activity by tapping along to the beat, based on what came before, you would not be able to do so. The brainwaves are really that asynchronous—their drumbeat has no discernible rhythm. Even if I converted the brainwaves into sound (which I have done in my laboratory in a sonification-of-sleep project, and is eerie to behold), you would find it impossible to dance to. These are the electrical hallmarks of full wakefulness: fast-frequency, chaotic brainwave activity.