Sleep, Memory, and Forgetting
When we are awake, we form memories; during sleep, these memories are replayed and transformed. The reverberation of memories is subtly present in the psychoanalytic concept of the “day residue,” but Freud’s works do not mention the role of sleep in learning. Carl Jung came closer to this idea, suggesting that dreams prepare the dreamer for the coming day.
The first experimental approach to studying the relationship between sleep and learning did not emerge in Europe—the undisputed center of scientific knowledge in the 19th century—but in the United States, where university traditions were just beginning to form. In the early 1920s, researchers John Jenkins and Karl Dallenbach at Cornell University attempted to replicate a classic German experiment. In the original study, conducted decades earlier, volunteers memorized lists of syllables unfamiliar to their language, and scientists measured how well these syllables were retained over time.
Using this simple procedure, Ebbinghaus, forty years before Dallenbach and Jenkins, established that once-formed memories fade exponentially over time. He developed the “forgetting curve,” which describes memory dynamics across countless biological species. Jenkins and Dallenbach’s innovation was to have participants sleep immediately after learning the syllables.
For comparison, they repeated the experiment but changed the conditions—this time, volunteers were not allowed to sleep. Strangely, after the same period, participants who slept remembered much more than those who stayed awake. The awake group consisted of undergraduate students who went to class after the session. A joke born at the time is still popular among scientists: Jenkins and Dallenbach proved that sleeping is more useful for learning than attending lectures.
On a serious note, we now know that the relevant variable for low retention in the awake group is sensory and cognitive interference. While awake, the brain is constantly bombarded by various stimuli, which greatly hinder the mnemonic process. A clear example is trying to hum one song while listening to another—the effort required is proportional to the volume of the interfering music. This demonstrates how difficult it is for the awake brain to isolate itself from reality and not react to it.
For some reason, Jenkins and Dallenbach’s discovery was not picked up by their contemporaries. It remained in the shadows for decades and had no impact on science. A few minor studies were conducted in the 1940s, but World War II was underway, followed almost immediately by the Cold War. These were pre-internet times—information spread slowly and unpredictably, and often did not reach a wide audience.
Only in the 1950s did the U.S. become the epicenter of REM sleep research and its connection to dreaming, but initially, no one focused on the cognitive aspect of sleep. Jenkins and Dallenbach’s findings, published in 1924, had to wait 40 years before being noticed.
Jouvet and the Flowerpot Method
In the late 1960s, there was a surge of interest in this topic in France and the U.S.: a new generation of scientists, influenced by new ideas, began to focus on the importance of sleep for cognitive function. The common denominator in these experiments was depriving rodents of sleep after a training session.
The “flowerpot method,” proposed by Jouvet, was simple, effective, and inexpensive. It quickly spread among labs studying the biological consequences of sleep deprivation. The method is based on the fact that slow-wave sleep is accompanied by a drop in muscle tone, which decreases even further at the onset of REM sleep.
The test animal was placed on a small platform—an upside-down flowerpot set in a container of water. As soon as the animal fell asleep and lost muscle tone, it would fall into the water and wake up. By adjusting the platform’s diameter, researchers could deprive the animal of all sleep or just REM sleep.
Early experiments using this method showed that rats subjected to training sessions—spatial learning, conditioned fear, trial-and-error learning—remembered less after total sleep deprivation or specifically REM sleep deprivation.
Lack of sleep must be compensated for, especially REM sleep. Its deprivation consistently leads to a rebound, requiring recovery proportional to the amount of missed sleep. However, the reverse is not true: you can significantly increase REM sleep duration at the expense of total sleep time, but you cannot “stock up” on sleep—there will be no reduction in REM sleep the next day.
Emotions have a significant impact on this dynamic. Moderate anxiety reduces total REM sleep duration, but severe stress (such as a life-threatening event) can, after it passes, lead to a significant increase in REM sleep. This highlights the vital role of REM sleep in human cognitive function.
Throughout the 1970s, several scientists convincingly demonstrated that sleep deprivation is harmful to learning. This sparked a surge of interest in the context of international competition and collaboration. REM sleep was the focus, considered the most interesting phase due to its close connection with dreaming. However, over time, a movement grew against the idea that REM sleep had any cognitive value.
Stress or Sleep Deprivation?
The sleep deprivation method received the most criticism from skeptics. The flowerpot, which Michel Jouvet devised as a means of deprivation, itself causes stress. If it’s too small, the animal falls into the water at the first signs of sleep. If it’s slightly larger, the animal falls off the platform when muscle tone is very low. Unexpectedly falling into water after a certain threshold of atonia causes a strong shock. Clearly, the situation in these experiments is unnatural and stressful for the animals.
In addition to abrupt awakenings from falling into cold water, the rats’ movement is severely restricted. After several hours of sleep deprivation, they begin to walk around the flooded cage, remaining wet. As a result, the animals show irritation and general metabolic changes, including the production of glucocorticoids (stress hormones) in the hippocampus, which can negatively affect memory. With so many side effects from the experimental conditions themselves, it would be an overstatement to attribute the observed memory deficits solely to lack of sleep.
This argument is valid. Therefore, newer experiments used less stressful methods of sleep deprivation. Doctoral student William Fishbein and his advisor William Dement took advantage of important behavioral traits of rodents. Unlike heavy adult rats (over 300 grams), small, light mice (30 grams) can hang from cage bars for long periods—even upside down. They feel so comfortable that they can even eat and drink this way.
The “flowerpot method” did not cause excessive stress in small, light mice—they stayed on the platform only when they truly wanted to sleep. Still, sleep deprivation experiments with mice also showed impaired memory, supporting the hypothesis.
Nevertheless, the argument about the stressful method kept resurfacing. As an alternative, researchers began using gentle but effective interventions—waking the animal each time it tried to fall asleep. Obviously, the success of this method depends on the experimenter’s attentiveness: if they get distracted, data quality suffers and interpretation becomes less convincing. Thus, clouds of protest began to gather.
Two American scientists, psychiatrist Jerome Siegel from UCLA and neuroanatomist Robert Vertes from Florida Atlantic University, became known for their strong arguments against the cognitive hypothesis of sleep.
Skeptics vs. the Lone Ranger
Questions continued to multiply. If REM sleep is so important for cognitive function, why do reptiles, birds, and even mammals like the echidna lack it? If it’s used for learning, why don’t intelligent animals like dolphins have it, while others, clearly less intelligent (like armadillos), have it in abundance? Why does REM sleep shorten in people taking antidepressants, yet no learning deficits occur? Why is there no correlation between time spent in REM sleep and a person’s learning ability?
Defenders of the theory countered: we’re not entirely sure dolphins lack REM sleep—their episodes may simply be too brief to record. Also, dolphins evolved from land mammals that moved into the water. In cetaceans, REM sleep may have been reduced or eliminated to prevent total atonia in the aquatic environment, which could lead to drowning. In the context of adaptation to a new environment, the cognitive functions of REM sleep may have been replaced by other, metabolically equivalent processes.
Armadillos spend a lot of time underground. Data from the past 20 years show that, contrary to previous beliefs, echidnas, birds, and even reptiles do have REM sleep.
Treatment with antidepressants increases levels of neurotransmitters—norepinephrine, dopamine, and serotonin—which are important for memory formation. So it’s likely that memory consolidation during wakefulness compensates for the shortened REM phase.
The debates intensified and became more heated in the 1980s. Scientists clearly divided into camps based on their views of sleep’s cognitive properties. For a while, it resembled a dialogue of the deaf. Discouraged by the hostile atmosphere at sleep research conferences and increasingly aggressive anonymous peer reviews, veterans gradually left the field. Over a decade, scientific interest in the connection between sleep and learning declined significantly.
During this turbulent period, the burly Canadian Carlisle Smith, a psychologist from Trent University, almost single-handedly defended the cognitive role of REM sleep. Through a series of rodent experiments, he demonstrated its positive effect at certain intervals after training and identified periods when memory was most vulnerable to sleep deprivation.
But the lone ranger Smith could not sway critics of the cognitive theory of sleep, and the situation remained at an impasse until the early 1990s. Then, a new and unexpected player entered the scene, tipping the scales through experiments on humans.