Fifty nine years ago, two doctors at the University of Chicago, Eugene Aserinsky and Nathaniel Kleitman, reported in the journal Science on the “eye motility, and concomitant phenomena” that seem to occur every night when a normal, healthy person sleeps. The researchers noted that their subjects, usually a few hours after descending into slumber, would enter a phase in which their eyes moved about oddly—in a “rapid, jerky, and binocularly symmetrical” manner—as if, behind their closed eyelids, they were seeing things. Electroencephalograms (EEGs) taken at these times resembled those taken when subjects were awake. And indeed the subjects, behind their trembling eyelids, seemed to be having experiences: when roused, they typically reported that they had been dreaming.
Aserinsky and Kleitman had discovered what would later be called rapid-eye-movement (REM) sleep, the sleep phase in which we humans—and most other mammals—seem to do the vast majority of our dreaming. But there was one big question that Aserinsky and Kleitman were unable to answer conclusively: Why? Why do we have REM sleep? Why do we dream? What is it for?
“It’s the state of consciousness that seems to generate the most interest besides the waking state, yet there’s still little agreement about its function,” says Jerome M. Siegel, a psychiatry professor and sleep researcher at the University of California – Los Angeles.
Even before Aserinsky and Kleitman’s paper, some researchers had suspected that sleep helps to strengthen or “consolidate” new memories in the brain. The discovery of REM sleep suggested that dreaming is the subjective experience of this consolidation process—the experience of new memories being replayed.
Evidence for this hypothesis eventually emerged. In the 1980s, researchers reported that rats spend more time in REM sleep after learning new tasks, and show weaker recall of new memories if deprived of REM sleep. Others reported similar findings in cats.
However, the degree of association between REM and memory consolidation seemed to depend significantly on the species tested—even the strain, among rats and mice—as well as on the type of task that the animal learned. When researchers tried to repeat such studies with human subjects, they also found results that were less than straightforward. Reducing REM sleep had no effect on new-memory consolidation in some experiments. In at least one study, it unexpectedly improved memory consolidation. In the past five years or so, studies have shown more consistently that the consolidation of new memories, especially declarative memories that we can verbalize, depends much more on the deepest part of sleep, called slow-wave sleep (SWS).
Many sleep researchers now think that the hippocampus, a short-term memory storage region in the brain, uses the relative neural quiet of SWS to stop recording and start replaying new memories. In so doing, the hippocampus reactivates and strengthens the circuits in the brain’s cortex that correspond to the slower-forming, longer-term versions of the memories. In this view, REM sleep plays a supporting role by igniting a local reactivation of these cortical memory circuits so they integrate better with surrounding cortical circuits—forming new connections that may enhance creativity—and might also seem bizarre when experienced during dreams. For example, “I may have a memory that I consolidated during SWS, of myself sitting in a classroom, but when I get to REM sleep I get this crazy dream that an elephant shows up in the classroom,” says Rebecca Spencer, a psychologist at the University of Massachussetts who researches sleep and memory.
As Spencer’s and others’ research suggests, REM sleep is also a time when the amygdala, a storehouse of emotional memory, is particularly active, and may be strengthening its own connections to long-term cortically-stored memories. This would give the memories emotional connections that—among other things—make them easier or harder to recall.
A virtual-reality training program?
The memory-consolidation hypothesis is still a matter of debate. One reason is that researchers don’t yet agree that waking-state experience and subsequent dream content are closely related. Harvard researcher Robert Stickgold and colleagues published a much-cited study in 2010, in which people who played the Alpine Racer 2 video game for lengthy periods reported similar experiences during dreams the next time they slept. Yet as Spencer points out, the dream content that subjects record during such studies is a small fraction of their overall dreaming, and is often biased towards the night’s earliest periods of REM sleep, when recall seems easiest. “Some people in the sleep research field would ask, well, is that the same as the REM sleep you have later in the night?” she says.
J. Allan Hobson, a professor of psychiatry emeritus at Harvard, and a member of the Dana Alliance of Brain Initiatives, also doubts that REM sleep is principally for new-memory consolidation. “You do Alpine Racer for two hours a day, it’s gonna get in there,” he says. “But most of what happens [in dreams] is not predictable from anything that happened during the daytime.”
To Hobson, the big clue about REM sleep is that it accounts for more sleep-time in children and infants, and in fact is a baby’s principal state of consciousness in the last weeks in the womb, before there are any real-world memories to consolidate. “It’s very tempting to think that [REM sleep] is just a replay of waking experience; but it isn’t,” Hobson says. “It’s a pre-play.”
Hobson’s hypothesis, which he described in a review in Nature Neuroscience three years ago, is that REM sleep is essentially the manifestation of a “virtual reality program,” encoded in our genes to help exercise and prepare our brains for the real world. In other words, although REM dreaming may be influenced by specific new memories from experience in the world, it is more about the ancient fundamentals of human and primate life. “It’s a reinforcement of basic knowledge—knowledge that precedes any waking-state learning: how to be a person, how to be an ego, how to exist in a space, how to move in a space, how to feel,” Hobson says. “It’s not environmental memory; it’s genetic memory.”
Just warming up the engines?
Like humans, non-human mammals typically rely more heavily on REM sleep in the fetal and newborn phases of life, less so in later childhood, and still less so in adulthood. This would seem to buttress the idea that REM sleep evolved to boost development. However, Siegel notes, REM sleep cannot be absolutely necessary for normal development or cognition, because there are relatively “smart” mammals, such as dolphins, that apparently do not experience REM sleep. “It also seems that dolphins and killer whales, and probably whales generally, don’t have any sleep at all for the first few months of life,” he says. “They’re just continuously active, and meanwhile they’re developing the biggest brains on the planet and arguably some of the highest levels of cognition.”
As for humans, although reports in the 1960s suggested that REM sleep deprivation can be emotionally destabilizing, it now appears to have no major harmful effects. In fact, some anti-depressant drugs have the side-effect of reducing REM sleep, which might even contribute to the intended mood-lifting effect. There are also reports of people who sustain brain injuries that leave them with little or no REM sleep, but still seem cognitively normal. Variations in REM sleep time in sample populations also seem to have little to do with IQ or other measures of cognitive ability. “People looked at all these parameters right after the discovery of REM sleep, and there just wasn’t anything there,” Siegel says.
He suspects that the key to the mystery of REM sleep lies within the broader function of sleep overall. In his view, that function is an “adaptive inactivity” that conserves an animal’s energy and lowers its risk of violent death, by keeping it in a low-metabolism state and relatively safe from predators at times when it doesn’t need to be out hunting or foraging. Siegel and others have observed that animals which seem less able to afford sleep—because they are too large to hide from predators, and/or have low-calorie food sources that require them to be awake and feeding for long periods—do less of it, and tend to sleep less deeply too. Giraffes, for example, sleep for only about five hours per day in total, and seldom lie down for more than a half-hour at a time. “If giraffes slept deeply for two hours a day, there soon wouldn’t be any giraffes; they would be eaten,” Siegel says. Other animals may dispense with sleep altogether during special circumstances, such as mating season.
A big drawback of energy-conserving deep sleep is that an animal will be groggy and unable to function, if awakened suddenly from that state. REM sleep is much closer to the waking state, and we normally emerge from it when we do wake up, if left to wake naturally without an alarm clock. REM sleep’s principal function, Siegel suggests, may thus be to nudge the brain into a ready-for-action state, to better prepare it for actual waking. “The brain has higher metabolism in REM sleep and physically warms up, so this may be a way of reversing the loss of alertness that comes with non-REM sleep,” says Siegel.
In other words, REM sleep may be no more than a warming of our neural engines, and its associated dream content may be less meaningful than has often been assumed. Like other hypotheses for REM sleep, Siegel’s remains to be established as an accepted scientific fact. But he feels certain that REM sleep evolved for a reason. “It’s costly in terms of energy, so it’s hard to believe that it has no function,” he says.