Sleep and memory: Dr Jerome Siegel and Dr Robert Vertes
- 1 Is sleep a neural optimizer?
- 2 Robert Vertes and Jerome Siegel
- 2.1 1. Sleep does not serve a role in declarative memory?
- 2.2 2. REM sleep deprivation does not lead to cognitive impairment?
- 2.3 3. Sleep-dependent enhancement of procedural learning has not been proven?
- 2.4 4. Learning in waking is far more significant than overnight enhancements?
- 2.5 5. Sleep models should be simple
- 2.6 How can random impulsations in REM make a sense in dreams?
- 2.7 Dr Siegel's theory of sleep
- 2.8 My personal bias
- 2.9 Olive branch
- 2.10 More reading for skeptics
- 3 References
Is sleep a neural optimizer?
November 2, 2001 issue of Science included opposing articles on the physiology of sleep that provided more fuel to the fire of a scholastic dispute on the function of sleep. At the time when the effort to disentangle details of complex neural optimization process called sleep seemed close to bringing major breakthroughs, there were reputable researchers throwing a spanner in the works. Prominent narcolepsy expert, Dr Jerome M. Siegel from the Center for Sleep Research of the Department of Veterans Affairs analyzed the evidence for the role of sleep and figured out that... the evidence is scant! An array of well-organized evidence against the role of sleep in memory consolidation was rolled out in a journal that for years has stood as the bastion of highest quality science (Siegel 2001). Siegel acknowledged the vital evolutionary function of sleep but, at that time, did not present a viable alternatives to the position taken by Stickgold, Hobson and Maquet in the same issue of Science (Stickgold et al. 2001, Maquet 2001).
In his later publications, Siegel proposed his own theory of sleep saying: "sleep is merely a good way to increase an animal's efficiency, by hunting only when the hunting is good, for example, and minimize its risk". I have mentioned earlier that this theory fails to meet the shutdown test, and it would be enough to keep the animal conscious and atonic to accomplish Siegel's criteria for preserving sleep in the course of evolution. For more on Siegel's theory see: Robert Vertes and Jerome Siegel.
Behind all opposing positions in science, there is a root belief or model which is only later enhanced or muddled by existing clutter of evidence. The root of the role of sleep in learning is in the evolutionary explanation of the need to turn off the conscious mind for a third of our lives. Those who oppose the role of sleep in memory consolidation start off with the difficulty in reconciling with the fact that unconscious brain might be involved in learning which seems to require a conscious mind. Some researchers dismiss findings showing the role of sleep in learning as a side effect of stressful methods used in sleep deprivation. Others claim that sleep helps learning by nothing more than "refreshing" the mind and restoring energy reserves. In the welter of data confusing the roles of NREM and REM sleep, procedural and declarative learning, memory consolidation vs. memory optimization, short-term vs. long-term memory, spacing effect vs. forgetting vs. interference, animal and human studies, stressful vs. stressless research methods, there will always be room for the voice of dissent. It will take strong computational and unequivocal experimental models to leave less scope for hair splitting over the role of sleep in learning. Until that happens, you will hear many confusing voices.
Robert Vertes and Jerome Siegel
Jerome Siegel is not the only dissenting voice in the role of sleep in memory. Robert Vertes is also a highly reputable sleep researcher. Siegel and Vertes come from two opposite coasts of the US (respectively: Center for Complex Systems and Brain Sciences, Florida Atlantic University, Boca Raton and Department of Psychiatry, School of Medicine, UCLA, Los Angeles). They also come from different fields of neuroscience and sleep research, and yet have been united in their lone battle with the highly established views on the involvement of sleep in processing memories. They have been disappointed with Dr Stickgold's making an impression that "the door is closed" in the debate on the involvement of sleep in learning. As I hope to be open to new ideas, I would like to address the concerns of Drs Vertes and Siegel.
All scientists, including the brightest ones, have their prejudices and intuitions. Actually, nothing produces a stronger prejudice than rich knowledge that has been filtered or selected for a specific purpose. Prejudiced and selective knowledge characterizes every expert. This comes from the simple fact that it is pretty hard to accumulate more than a few hundred thousands of pieces of information in one's memory over a lifetime. At the same time it is often said that to qualify for the title of an expert, one needs some 50,000 pieces of information from a given field. No wonder then that all experts can truly encompass only one field that will permeate all their thinking, even when they double their expertise on general knowledge with 50,000 pieces of information in other fields. Even if the numbers don't sound convincing, judging by the definition of an expert, it is very hard to become one in one field, let alone in two or more.
Only expertly prejudice can explain why top scientists in the field disagree with the rest of the field that includes equally bright individuals. The problem of prejudice affects everyone, and this paragraph is certainly also a reflection of my own prejudice. My prejudice comes from years of involvement in memory and learning, from my background in computing sciences, as well as from numerous events in my own life that left a long-lasting imprint on my thinking about sleep (see later in this chapter).
Vertes and Siegel have opposed the idea of memory consolidation in sleep since the 1990s. As their voices are pretty solitary, they have united their efforts. After years of the battle with "the establishment", in 2005, they issued the following appeal in the journal Sleep: Time for the Sleep Community to Take a Critical Look at the Purported Role of Sleep in Memory Processing (Vertes and Siegel 2005). Here are some points raised by Vertes and Siegel:
1. Sleep does not serve a role in declarative memory?
Indeed, research seems to be pretty contradictory in this respect. The recall of vocabulary learned on one day does not seem to be much affected by lack of sleep on the following night (except on the days that follow sleep deprivation). However, all research papers should be sifted for two important observations:
- It is hard to know what is happening inside a black box by watching the relationship between the input and the output. The transition function may remain unchanged, while the internal workings of the automaton may change. Metaphorically, the brain after sleep could be compared to a computer after a disk defragmentation. The transition function is the same, and yet the operations are different.
- If sleep is considered a neural optimizer, and the dual network theory is right, we should look at long-term effects of missed sleep. If the purpose of sleep is to avoid catastrophic interference, then it is not enough to test the status of the memory after sleep or lack thereof. We also need to allow of a sufficiently long waking period that involves further learning. A simple mental experiment should illustrate that sleep is vital for declarative memories: can anyone expect any meaningful learning after a sleepless night or two? It is not just a matter of tiredness, even a few minutes of deep sleep seems to make a world of difference in our ability to soak in new knowledge. Data obtained with SuperMemo shows unequivocally that sleep is a remarkable tool for unplugging human memory for new input!
2. REM sleep deprivation does not lead to cognitive impairment?
Vertes and Siegel write (shortened):
Several lines of evidence indicate that REM sleep is not involved in memory processing/consolidation. Perhaps the strongest argument for this is the demonstration that the marked suppression or elimination of REM sleep in individuals with brainstem lesions or on antidepressant drugs has no detrimental effect on cognition. A classic case is that of an Israeli man who at the age of 19 suffered damage to the brainstem, and when examined at the age of 33 he showed no REM sleep. The man, now 55, is a lawyer, a painter and interestingly the editor of a puzzle column for an Israeli magazine. Recently commenting on his ‘famous’ patient, Peretz Lavie stated that "he is probably the most normal person I know and one of the most successful ones
REM sleep is younger than sleep overall in terms of its evolutionary origin and might indeed be just a mental extra for smarter animals. REM sleep deprivation results in much later death in rats as compared with sleep deprivation (months rather than weeks). Suppression of REM with drugs does not lead to a drastic cognitive decline. However, this cannot be used to diminish the importance of REM. After a sleepless night, humans can also function remarkably well. After a sleepless night, world leaders often take decisions affecting billions of people. This does not undermine the importance of sleep. Data obtained with SuperMemo shows that sleep deprivation has a dramatic effect on learning.
As for the Isreaeli no-REM man, we would need to know exactly what portion of the REM control system has been damaged. After all, the expression of REM is complex and multithreaded, and we do not know exactly which components of the complex process are most essential for what memory and creativity aspects, and to what degree these are expressed phenomenologically as interpreted by the EEG, or muscle atonia, or other REM constituents that can become dissociated in various REM disorders, or even in healthy people (e.g. blind people's REM). We could be dealing here with a subset of REM that is sufficient for retaining reasonable cognitive powers. Nowadays, neuroscientists tend to shift from the link between REM and dreaming to a more anatomical approach (e.g. the link between the pontine brainstem activity and dreaming). Leading a "normal life" is a very weak test for cognitive assessment, creativity, mnemonic capacity, etc. After all, millions of people destroy their REM-rich sleep with an alarm clock and still lead a "normal" life. Even people with damage to the hippocampus, which destroys their short-memory capacity, can be deceptively "normal" in a conversation. Brain's compensatory capacity is enormous. According to Buzsaki, REM sleep might be the chief optimizer that would help load the hippocampal circuitry with patterns acquired from the neocortex for a memory rewrite in the successive NREM sleep. If that interpretation was taken into consideration, REM deprivation would allow of "normal" life, however, one's creative insight would be diminished significantly. We also need to distinguish here between a reasoning creativity, and an artistic creativity of a painter. Artistic creativity may actually benefit from a degree of information chaos (TMS experiments show how diminished abstract thinking helps a creative artistic expression in a clear neural trade-off). Anecdotal evidence coming from the interpretation of SleepChart graphs seems to indicate that misalignment of sleep that would produce an expected drop in the proportion of REM sleep does indeed affect cognitive performance and perhaps, even more importantly, the creative insight that should characterize a healthy awakening. The story is not different in cases of alcohol or cannabis consumption before sleep. As much as we cannot ignore a single point of data in the case of REM non-expression, we cannot ignore multiple points of anecdotal data coming from SleepChart reports.
3. Sleep-dependent enhancement of procedural learning has not been proven?
Theoretically, due to its stochastic nature, procedural learning should benefit from sleep even more than declarative learning. It simply provides more room for neural optimization. This is why pre- and post-sleep difference should be more pronounced. However, I cannot verify this with SuperMemo as it is generally not employed in procedural learning. I can only rely on reports from other sources. Dr Matthew Walker research, for example, is pretty convincing. I am a bit confused here. In contrast to his stance in the discussed paper, Siegel himself apparently told WebMD that "Studies of procedural learning show clear evidence that you don't get improvement until you sleep". And later: "Studies show that when learning muscle-related tasks -- like dancing, piano playing, gymnastics -- there is significant improvement after a night's sleep". Perhaps the answer comes from the fact that the interview was conducted in 2001.
4. Learning in waking is far more significant than overnight enhancements?
Vertes and Siegel write: "Although we find the notion of enhancement to be counterintuitive (improvements in learning with the mere passage of time), the question arises regarding the contribution of ‘enhancement’, compared to practice on tasks, to overall improvements on them. [...] My guess is that sleep contributes very little".
To me, overnight improvements to the effects of learning are not more counterintuitive than muscle growth in lazy sleep following a heavy workout. The intuition saying that learning is done best in waking is the basis of short-term learning strategies popular amongst (not-so-good) students, where an all-nighter before an exam seems like a good idea. After all, more can be done when cramming before an exam than one might expect from overnight "consolidation". However, if one tries to stretch the idea to just two days, it becomes apparent that without sleep there is no learning. And this is not just a question of rest. If we employ Walker's metaphor of an overflowing Inbox, we can see that missing one night of "enhancement" won't do much damage. Everyone can handle his or her mail ok for a day or two without keeping one's Inbox clean. Many people keep a mountain of mail in the Inbox for weeks or months. However, in the long-run, the bulging Inbox will result in some important mail never getting answered. The same applies to learning, incl. procedural learning. In the long run the additive power of chaos will render further progress hardly possible. It is true then that a single night might not matter much, but without the neural optimization taking place during sleep, our capacity for creative thinking or ability to acquire new skill will, at some point, become saturated and the learning progress will grind to a halt.
Comparing the value and contribution of the learning process in waking with the value of overnight optimizations is not much different than quantifying the contributions of the brain and the heart to our survival. It does not matter that it is the brain that provides the spectacular and the salient. Both organs are vital and indispensable. Similarly, waking and sleep are both vital for learning and memory.
5. Sleep models should be simple
As always lex parsimoniae is a helpful guidance in science. The concept of sleep as a neural optimizer seems to meet the criteria of convincing simplicity. Studying PGO waves, theta rhythms or sleep spindles is important, too, to provide raw data for simple models. After all, without the EEG we would not separate NREM from REM in the first place. And those two sleep modes seem to provide the basis of the essential optimization interplay. Moreover, brain waves are the expression of the behavior of cell populations that, in the synchronous mode, can be an indication of important physiological functions underway. I am here a bit confused again as Vertes once wrote: "We propose that theta serves to gate and/or enhance the storage of information reaching the hippocampus simultaneous with theta. In the awake state, the 'information arriving with theta' is governed by the situation; i.e., the sum of internal and external events time-locked to theta" (Vertes 1995). Is hypothesizing on the function of brain waves ok for waking states and not ok for sleep? Sleep will improve memory not only because of neural optimization. Non-interference or regeneration do also have their role to play. These, however, cannot explain all the effects of sleep on learning. We need to keep the evolution in mind (see the shutdown test) and the neural aspects of sleep (e.g. why 5 min. nap makes a huge difference to one's learning capacity). Siegel's own theory: "the main function of sleep is to increase an animal's efficiency and minimize its risk by controlling how a species behaves with regards to its surroundings" is adequately simple. Neural optimization in sleep is doing just that: increasing an animal's efficiency.
How can random impulsations in REM make a sense in dreams?
As much as the early opponents of the evolution wondered how random mutations could lead to the development of the marvel of the human species, so much some sleep researchers wonder how could random REM impulsation be possibly beneficial in learning or otherwise. Robert P. Vertes says: "The problem seems to be, that unlike waking, in which the nature of the information reaching the hippocampus is essentially dictated by behavioral circumstances, there appears to be no mechanism for the selection and orderly transfer of information to the hippocampus in REM sleep. If this is the case, and the transfer process in REM is not orderly or is basically random (as one might conclude from even the most casual inspection of dreams), then the process would appear to have no functional value for the animal; there would appear to be no benefit in encoding or 'remembering' a randomly ordered, non-related series of events. Instead, a hypothesis is advanced that "the primary function of rapid eye movement (REM) sleep is to provide periodic endogenous stimulation to the brain which serves to maintain minimum requisite levels of CNS activity throughout sleep" (Vertes 1986), which naturally raises the question "why sleep in the first place", which then on the basis of research and evolutionary elimination leads us back to the need for sleep for the effective function of memory, and back to REM as the primary network optimizer. See also Vertes, R.P. and Eastman, K.E. "The case against memory consolidation in REM sleep" (Vertes and Eastman 2000). The role of randomness and pseudopatterns in transferring information between neural networks has been mentioned in Neural optimization in sleep. Examples of self-organization in the natural world are too numerous to list: growth of crystals, folding of proteins, flocking of birds, or the life itself. Complex systems may arise from chaos by an application of a simple set of rules. The brain can convert random shapes in the clouds into vivid images and associations. The visual system can do miracles when juxtaposing contradictory data in stereoscopic vision. Neural networks can recognize faces with just a tiny fraction of data available. Unlike computers, the brain excels in extracting models, producing remote associations and resolving contradictions. Those abilities are essential for reorganizing networks in sleep and for exercising hypothetical survival scenarios.
Dr Siegel's theory of sleep
Dr Siegel proposed his own theory of sleep. Here are a couple of statements by Dr Siegel that do not square well with that I know about sleep physiology and evolution:
- "Sleep is merely a good way to increase an animal's efficiency, by hunting only when the hunting is good, for example, and minimize its risk". For that particular formulation, only a tiny subset of the complex sleep control machinery would be sufficient to meet Dr Siegel's proposed function of sleep. For example, at opportune "sleep" times, it would be enough to keep the animal awake and paralyzed as in REM sleep. All the rest of sleep control would be a simple waste of time and resources. Why would we need a homeostatic control of sleep associated with slow-wave activity? Circadian control would be enough. Or even a simpler solution: a homeostatic reaction to the availability to light. Animals would just drop paralyzed as soon as light was unavailable for long enough (in case of diurnal species).
- "Smaller animals have higher metabolic rates, and this results in the generation of higher levels of small, highly reactive molecules such as free radicals. Studies on rats show that when they are deprived of sleep, their oxidative stress increases". Oxidative stress indeed increases with activity, but turning off the consciousness does not help in recovery. The brain works as hard in sleep as in waking. That work is not directed at reducing oxidative stress, but at making sure that all networks are ready, among others, to combat the effects of oxidative stress via homeostatic mechanisms.
- "My theory doesn't mean that significant things don't happen in sleep — it just means there is no vital universal function for sleep". "A whole spectrum of things are accomplished better in sleep than awake in some animals, such as digesting food, but it doesn't mean the function of sleep is digestion". Evolution teaches us that if a function is conserved over millions of years, it is a near-guarantee for there is an underlying vital component. All aggregates of lesser adaptive functions disintegrate, get molded, mutate, modify, get erased or amplified, split and join. Digestive or respiratory functions might go in sync in one set of circumstances, and part ways in different evolutionary settings. It is the neural optimization that retains sleep as an unavoidable part of animal life for millions of years.
- "Sleep has a lot in common with inactive states observed in a wide range of species. This includes plants and simple microbes, which in many cases do not have nervous systems - a challenge to the idea that sleep is for the brain". We all know that sleep was hooked to circadian rhythms. That means that the circadian function is older than the neural function. That implies that there are primitive organisms that express circadian cycles without sleeping. However, for Siegel's logic to hold, he would have to point to organisms with a complex nervous system that do not sleep.
- "Sleep is lying on a continuum that ranges from dormant states like torpor and hibernation, on to periods of continuous activity without any sleep, such as during migration". Sleep is part of the continuum of dormant states and occupies its most advanced and extreme position. Sleep is the acme of evolutionary success that underlies the creative component of our civilization.
My personal bias
I mentioned my personal bias.
Mathew Walker made the following prediction in 2009:
We will soon have a new taxonomy of sleep-dependent memory processing, and one that will supersede the polarized all-or-none views of the past (Stickgold and Walker 2005; Vertes and Siegel 2005). With such findings, we can come to a revised appreciation of how both wake and sleep unite in a symbiotic alliance to coordinate the encoding, consolidation and integration of our memories, the ultimate aim of which maybe to create a generalized catalogue of stored knowledge that does not rely on the verbose retention of all previously learned facts" (Walker 2009)
More reading for skeptics
If you are interested in the points of view by Drs Vertes and Siegel see:
- a boatload of interesting publications by Dr Siegel: http://www.semel.ucla.edu/sleepresearch
- Dr Robert Vertes 2000: "We believe that the cumulative evidence indicates that REM sleep serves no role in the processing or consolidation of memory"
- Siegel, J.M., "The REM Sleep-Memory Consolidation Hypothesis," Science / Volume 294 / Issue 5544 (November 2, 2001): 1058-1063, doi: 10.1126/science.1063049
- Stickgold R., Hobson J.A., Fosse R., and Fosse M., "Sleep, Learning, and Dreams: Off-line Memory Reprocessing," Science / Volume 294 / Issue 5544 (November 2, 2001): 1052-1057, doi: 10.1126/science.1063530
- Maquet P., "The Role of Sleep in Learning and Memory," Science / Volume 294 / Issue 5544 (November 2, 2001): 1048-1052, doi: 10.1126/science.1062856
- Vertes R.P. and Siegel J.M., "Time for the Sleep Community to Take a Critical Look at the Purported Role of Sleep in Memory Processing," Sleep / Volume 28 / Issue 10 (October 2005): 1228-1229
- Vertes R.P., "Memory consolidation in REM sleep: Dream on," Sleep Research Society Bulletin / Volume 1 / Issue 2 (April 1995): 27-32
- Vertes R.P., "A life-sustaining function for REM sleep: A theory," Neuroscience & Biobehavioral Reviews / Volume 10 / Issue 4 (Winter 1986): 371-376, doi: 10.1016/0149-7634(86)90002-3
- Vertes R.P. and Eastman K.E., "The case against memory consolidation in REM sleep," Behavioral and Brain Sciences / Volume 23 / Issue 6 (December 1, 2000): 867-876
- Wagner U., Gais S., Haider H., Verleger R. and Born J., "Sleep inspires insight," Nature / Volume 427 (January 22, 2004): 352-355, doi: 10.1038/nature02223
- Stickgold R., PhD and Walker M.P., PhD, "Sleep and Memory: The Ongoing Debate," Sleep / Volume 28 / Issue 10 (October 2005): 1225-1227
- Walker M.P., "The Role of Sleep in Cognition and Emotion," Annals of the New York Academy of Sciences / Volume 1156 (March 2009): 168-197