The Sleep Elaboration–Awake Pruning (SEAP) theory of memory: Long term memories grow in complexity during sleep and undergo selection while awake. Clinical, psychopharmacological and creative implications
Medical Hypotheses; 73: 1-4
Bruce G. Charlton and Peter Andras
bruce.charlton@buckingham.ac.uk
***
Summary
Long term memory (LTM) systems need to be adaptive such that they enhance an organism’s reproductive fitness and self-reproducing in order to maintain their complexity of communications over time in the face of entropic loss of information. Traditional ‘representation–consolidation’ accounts conceptualize memory adaptiveness as due to memories being ‘representations’ of the environment, and the longevity of memories as due to ‘consolidation’ processes. The assumption is that memory representations are formed while an animal is awake and interacting with the environment, and these memories are consolidated mainly while the animal is asleep. So the traditional view of memory is ‘instructionist’ and assumes that information is transferred from the environment into the brain. By contrast, we see memories as arising endogenously within the brain’s LTM system mainly during sleep, to create complex but probably maladaptive memories which are then simplified (‘pruned’) and selected during the awake period. When awake the LTM system is brought into a more intense interaction with past and present experience. Ours is therefore a ‘selectionist’ account of memory, and could be termed the Sleep Elaboration–Awake Pruning (or SEAP) theory. The SEAP theory explains the longevity of memories in the face of entropy by the tendency for memories to grow in complexity during sleep; and explains the adaptiveness of memory by selection for consistency with perceptions and previous memories during the awake state. Sleep is therefore that behavioural state during which most of the internal processing of the system of LTM occurs; and the reason sleep remains poorly understood is that its primary activity is the expansion of long term memories. By re-conceptualizing the relationship between memory, sleep and the environment; SEAP provides a radically new framework for memory research, with implications for the measurement of memory and the design of empirical investigations in clinical, psychopharmacological and creative domains. For example, it would be predicted that states of insufficient alertness such as delirium would produce errors of commission (memory distortion and false memories, as with psychotic delusions), while sleep deprivation would produce errors of memory omission (memory loss). Ultimately, the main argument in favour of SEAP is that long term memory must be a complex adaptive system, and complex systems arise, are selected and sustained according to the principles of systems theory; and therefore LTM cannot be functioning in the way assumed by ‘representation–consolidation’ theories.
***
The nature of long term memory: instructionist or selectionist?
What follows is an ‘in principle’ argument about the basic nature of human long term memory. Although the details of real human memory may differ; if the premises concerning the nature of complex systems are correct, then memory ‘must’ work in something like the way we describe [1].
Human long term memory is typically described as a brain system for the storage of information about what has happened to an organism, so that the organism will be able to use this information in the future in order better to survive and reproduce (i.e., to increase its ‘fitness’). The kind of thing which is ‘stored’ in the long term memory system includes external stimuli (perceived via the five senses) and internal body states (perceived via the autonomic nervous system and messenger molecules such as hormones) [2]. Long term memories (LTMs) are typically conceptualized in terms of changes to brain circuitry [3], for example changes in the pattern of synaptic sensitivities [4].
The vast capacity of human long term memory implies that memory must be an extremely complex system, and all complex systems share basic formal properties [1], [5] and [6].
The usual description has memories as ‘representations’ of environmental entities being formed while an animal is awake and alert; after which these memories are edited, sorted, combined, selected or pruned (i.e., ‘consolidated’) while the animal is asleep. This could be termed an awake elaboration–sleep simplification theory of memory, or an awake representation–sleep consolidation theory.
In contrast, we propose almost the opposite idea: that memories are elaborated mostly during sleep (when the brain is more-or-less cut-off from interaction with its environment) and these memories are then selected or ‘pruned’ by interaction with other brain communications when awake. Our theory could be termed the Sleep Elaboration–Awake Pruning (or SEAP) theory of memory.
The traditional ‘representation–consolidation’ view of memory is ‘instructionist’ because the environment is seen as instructing the brain during awake periods. In other words, the complexity of memory is ‘exogenous’ because it originates in the environment; and complexity is transferred from the environment into the brain such that brain complexity ‘represents’, or in-effect mirrors, environmental complexity; then subsequently this complex information in the brain is summarized and thereby simplified mainly during sleep (i.e., the ‘consolidation’ phase).
By contrast, we see the complexity of memories as having an endogenous origin: i.e., originating within the brain. We suggest that complexity of memories is generated within the brain during sleep, to create elaborate memories which are then simplified (‘pruned’) mainly during the awake period of alertness, when the brain is brought into a more intense interaction with both its internal and external environment.
So the SEAP theory sees complexity as arising in the brain (mostly during sleep), and the editing of this complexity as a secondary consequence of the brain interacting with the environment when the animal is awake and behaving. Instead of the complexity of memories deriving from the environment, the complexity of memories is reduced by interaction with the environment – such that (for instance) in response to experience some memories are lost, others are abbreviated, while other connected memories are separated.
The SEAP theory is therefore akin to ‘selectionist’ accounts of neurobiology such as those provided by Edelman [7] and Gazzaniga [8]. SEAP is a selectionist theory of memory in which random variation and combination generates non-adaptive complexity; and in which complexity is reduced and adaptiveness emerges by a competitive selection mechanism of differential growth and extinction of complex systems. The complex systems which survive selection and grow are those which are more adaptive in that particular selection environment.
Like Edelman and Gazzaniga we regard memories as a consequence of the generation of diversity and selection among variants. But we also believe that only systems of communications [5] undergo selection, and therefore it is not necessarily or usually the brain’s physical units (such as neurons) which are selected [1] and [6].
SEAP is therefore based on the axiom derived from Luhmann’s theory of complex systems [5] that systems of communications are primary, and the communicating components of these systems (such as neurons) are secondary [6]. So that long term memories should properly be conceptualized as abstract systems of communications between neurons; and not the particular pattern of anatomical entities such as neurons or synapses.
Growth of memory systems
Memory systems need to be self-reproducing in order to maintain their complexity of communications over time in the face of the universal tendency for entropic loss of organised complexity: i.e., loss of information. In other words, the intrinsic tendency is for memories to be lost [4], and memory systems need a mechanism whereby complexity can be generated and information can be maintained despite this entropic tendency.
According to the SEAP theory, self-reproduction of memories generates surplus memory communications and a tendency for expansion in complexity of the memory system. So that self-reproduction of the memory system randomly generates memory complexity, and growth of new memories and combinations of memories will in turn create competition between the newly-generated memories.
This competition between new memories leads to differential survival and extinction of memories and is the basis of the process of selection among memories in a manner precisely analogous to natural selection in biology, or the selection processes of the immune system. All types of selection share basic formal properties [6] and [9]. Natural selection and immune system selection both lead to adaptation to the environment, but by the generation of random complexity being subjected to pruning of ‘maladaptive’ (or, more exactly, less-adaptive) variants via interaction with the environment.
Hence memory systems must tend to grow to ensure their own survival, and growth in this systemic sense entails growth in complexity. So the conclusion is that memory systems must have a tendency to grow in complexity; and superimposed on this tendency to growth are selection mechanisms that enforce adaptiveness.
Selection of endogenously-generated memories occurs by interaction with other memories within the long term memory system, and also from the memory system’s interaction with other brain systems. Interaction between memories and their environment probably happens at the level of neurons – which are the units generating and receiving communications in the memory system. Presumably, individual long term memory (LTM) neurons will typically participate in ‘coding’ (communicating) more than one memory; and some LTM neurons will also participate in other neural systems.
Selection of memory systems
Selection is a consequence of these interactions at the neuron level between more than one memory sharing a particular neuron, and between memories and other brain functions in which that neuron participates. For example, a cortical neuron may participate in several memories relating to an individual person, and also in the awake processing relating to visual perception [4]. Some of these networks of communication will be compatible, and memories then may be combined and thus grow to generate more complex memories by including more neurons into the communications network, or by triggering more frequent communications from neurons already in the network.
This may be conceptualized as memories being selected by their compatibility with ongoing personal experience. ‘False’ memories are therefore those contradicted by perceptions, while ‘true’ memories are those compatible with perceptions.
And memories will also conflict and interfere, such that one memory system may suppress neuronal participation in another memory system; and such memories cannot be combined and cannot grow in complexity. Such memory systems are more likely to become extinct, and these memories be lost.
The rationale by which memories either grow and increase in complexity, or interfere and reduce in complexity, are the structural and functional properties of the long term memory system – which are currently poorly understood. Presumably the sketchy current knowledge of how and why memory associations form, or how and why some memories rapidly disappear, are preliminary evidence concerning the structural and functional properties of the long term memory system.
Memories are therefore subject to continual selection and reshaping by the organism’s ongoing waking experience, and each night memories will be elaborated and combined, so that the interaction between nocturnal memory growth and diurnal pruning means that memories will tend to evolve over time. Most memories will become extinct, but those which are not ‘contradicted’ by awake experience will continue to increase in complexity (mainly during sleep) until such a point that they do eventually lead to contradiction after which the erroneous memories will be pruned-back.
Cyclical nocturnal growth of complexity and diurnal competitive pruning by the perceptual system is therefore the process by which long term memories on the one hand overcome the continuous tendency to loss of information by random entropic processes, and on the other hand maintain their adaptive relevance such that the long term memories (on average, and in the environment where they evolved) will tend to be fitness-increasing.
The function of sleep in memory
While sleep is advantageous to reproductive fitness in most (although not all [4] and [10]) animals, nonetheless understanding the ‘function of sleep’ has proved elusive [11]. While sleep very probably has to do with the editing and maintenance of long term memory [4], the specifics of this have proved hard to pin-down (e.g. [12] and [13]).
The reason sleep remains poorly understood, we suggest, is that sleep does not really have ‘a function’ in terms of the organism as a whole. Rather, according to SEAP theory, sleep is the behavioural state during which most of the internal processing of the system of long term memory (LTM) occurs. The primary ‘function’ of sleep is therefore maintenance and increase of LTM complexity. Or, the function of sleep is the expansion of long term memories.
This implies that serving as an adaptive ‘memory’ system for the organism is merely a secondary function of memory; and that sleep does not exist to improve the ‘accuracy’ (or adaptive relevance) of memories but instead to generate the complexity of memories.
The main requirement for LTM is among complex animals living in complex and changing environments – i.e., situations in which organisms have a repertoire of potential behaviours, where each day generates different challenges, and when therefore animals stand potentially to benefit from memories of their previous experiences [4]. In such animals (including humans) LTM often has a vast information capacity, and therefore necessarily memory is vastly complex.
The complexity of a system can be defined in terms of its having a much greater density of internal communication than its interactions with the non-system environment: in principle, the quantitative differential between internal communications and external interactions is a measure of system complexity [5] and [6]. Such internal complexity appears to an external observer as memory activity ‘autonomous’ from the rest of the organism, and with little or no communication between the LTM and its environment.
In other words, the memory system (like any complex system) needs to be relatively cut-off from environmental interactions (especially the computationally-heavy load of visual stimulation). The long term memory system likewise needs to be all – but disengaged from initiating ‘action’ – therefore not engaged in purposive movement, with the organism either temporarily inert or merely performing repetitive and stereotyped motor behaviour. This set of conditions is closely approximated by the state of sleep [3] and [4].
Sleep may therefore be considered as the cycle during which memory systems are most engaged in their primary activity of internal processing. There is a great deal of evidence to suggest that sleep is important for memory functions [14] – but the perspective of abstract communication systems goes considerably further than this.
From the perspective of the long term memory (LTM) system, sleep processing is its main activity; sleep allows its maintenance, self-reproduction and increase in complexity, and the ‘memory function’ of the LTM system is a subordinate activity which has evolved to enable the LTM system to emerge, survive and thrive in the context of the rest of the brain.
In a metaphorical sense, the ‘memory function’ is merely rent paid by the LTM system to the organism.
Clinical and behavioural implications of the SEAP theory
Sleep
Sleep disturbances – reduced amount or quality of sleep – are an extremely common aspect of clinical practice. Lack of alertness is another common clinical problem. According to the SEAP theory, both sleep disturbance and impaired alertness would both be expected to impair memory – but in different ways.
Insufficient or too-often-interrupted sleep would presumably result in a reduction of complexity of communication in LTM: that is, a reduction in informational capacity of LTM. In summary, after sleep deprivation memories would be accurate and correct, but there would be a loss of content. The consequences of reduced complexity might include a reduction in potential total memory capacity of LTM, simplification of memories (less informational content, less combination of individual memories to form scenarios), and a greater probability of loss and extinction of memories. All of these predicted effects would be in principle measurable by properly designed memory tests.
Because the SEAP theory predicts that the accuracy of memories is mainly a consequence of selection processes during the awake and alert period; so that a major consequence of reduced alertness would be reduced accuracy of memories. So long as sleep was un-impaired; lack of alertness would be expected to produce inaccurate or false memories but not to cause memory losses. There would be plenty of memories, and memories would not be lost to entropy, but memory information would be inaccurate, unreliable, maladaptive. This inaccuracy would happen because memories had not undergone effective selection by interaction with perceptual systems and other pre-existing memories which had themselves undergone selection. So memories might be incompatible with direct experience and also with previous knowledge.
This situation of distorted and incoherent memories resembles the bizarre delusions which occur in psychotic states; and many psychotic states are associated with impaired alertness or ‘delirium’ [15]. Of course, sleep deprivation can itself be a cause of reduced alertness by causing increased sleepiness/impaired consciousness [16].
So specific states of insufficient alertness would be expected to produce errors of commission (memory distortion, false memories, memory inaccuracy), while sleep deprivation would produce errors of memory omission (memory loss). These predictions are testable, given the development of specific psychological measurement instruments to distinguish these types of memory error.
However, the specific consequences of sleep deprivation may be hard to predict without knowledge of the principles (or contingencies) of internal organization of the LTM. Furthermore, there may be various combinations of sleep loss and lack of alertness. One confusing factor is that sleep loss can itself produce drowsiness/delirium/lack of alertness (see below). These factors might explain the difficulties that sleep and memory researchers have experienced in precisely defining the function of sleep. For instance, the effects of ‘pure’ sleep deficiency on memory would be expected to be seen in terms of impairing the complexity of memories – but not necessarily on reducing the accuracy of memories.
Psychopharmacology
The SEAP theory implies that there is likely to be a trade-off and a phasic effect in the memory effects of some psychotropic drugs.
There are many sedative drugs (e.g., benzodiazepines, sedative antihistamines) that improve sleep; and also several psychostimulant drugs (e.g., dexamphetamine, methylphenidate) that improve alertness. However, these mainstream drugs lack specificity of action, because sedatives tend to have hangover effects of drowsiness after wakening, while stimulant drugs tend to have ‘hangover’ effects of insomnia and other types of sleep disturbance [17].
Therefore, the expected memory effect of sedatives might (assuming that sleep really is improved) be first to improve the recall of memories; but the secondary effect would be to impair the accuracy of memories (due to hangover and reduced alertness). The effect of psycho-stimulants might be the opposite: firstly to improve the accuracy of memories, then (when sleep disturbance became a problem) secondarily to impair sleep and increase the problem of memory losses.
Perhaps no single drug would therefore be expected to improve memory, and the most likely possibility for pharmacological enhancement of memory would be alternate and sequential circadian dosages of short-acting stimulants and short-acting sedatives.
Another possibility for memory enhancement might be methods for direct and specific brain stimulation – if it became possible technologically to initiate at will both restorative sleep and an awake and alert state of consciousness, and to impose these states alternately and sequentially.
Creative trance
It can be seen that the elaborative phase of long term memory bears considerable resemblance with the process of creativity as we usually understand it [18]. This suggests that most creative activity is likely to occur during sleep – indeed the characteristic ‘wide associative field’ of creative thinking has long been recognized as similar to the mental processes of dreaming. However, some creative people seem able to engage in associative thinking while relatively awake and alert – especially in a ‘trance’ state of altered consciousness of a kind traditionally associated with religious, spiritual, artistic and scientific breakthroughs and ‘eureka’ moments [19].
Since the process of elaborative memory is prone to maladaptive errors of commission, the SEAP theory emphasises that the creative trance is also likely to suffer from the same errors of commission as occur during other states of impaired alertness – and the products of a creative trance state therefore typically require pruning or ‘editing’ by the alert mind (or by other people) in order to eliminate this type of error.
Indeed, this two stage procedure of generation of raw material in a trance state of ‘impaired consciousness’ followed by the period of revision of critique in a state of alertness and clear consciousness is frequently seen in accounts of creativity. For example, the English poet and novelist Robert Graves described his writing procedure in precisely these terms: as firstly a self-induced trance state which generated the primary ‘raw material’, then a stage of making multiple revisions and re-shapings to the raw material when in a ‘normal’ state of alertness and concentration [20]. And the same stages are also observed in some examples of scientific creativity – the ‘breakthrough’ coming in a visionary state of actual sleep, sleepiness or some other altered state of consciousness – followed by a period of checking and validating [19].
This model may also explain the role of alcohol in creativity, since a high proportion of creative geniuses (especially in the arts) also ‘abuse’ alcohol [18] and [22]. A very intelligent and knowledgeable person may find their creativity limited, and use the sedative (alertness-reducing) properties of alcohol to enable the associations which form the basic raw material of their creativity (so long as the dose of alcohol is not so great as to lead to inertia). The alcohol-fuelled raw material is then selected, pruned and revised when sober.
Furthermore, creative geniuses may exhibit a phasic pattern of asocial versus social behaviour: a phase of solitude when they are cut-off from interaction with others so that their ideas (memories) may increase in complexity; followed by social engagement when these ideas are selected by interaction with the peer group.
Autodidacts, who lack interaction with a peer group; are often very creative and original but their ideas often also tend to be wrong or ‘crazy’ because they have lacked the selection process of peer interaction. They have too much solitary introspective brooding, and not enough interaction. However, professionals working in institutions tend to generate ideas that are sensible and correct but which tend to be dull and unoriginal – merely incremental extrapolations from existing knowledge. They exhibit too much peer interaction, and not enough solitary brooding.
The SEAP theory may therefore explain why most creative people are introverted [18], but that intermittent periods of peer interaction are also usually necessary.
Conclusion
The Sleep Elaboration–Awake Pruning theory of memory is not merely a reversal of the mainstream instructionist theory of memory since the putative memory processes are quite distinct. In particular, SEAP regards the complexity of memories as being endogenously-derived rather than ‘representing’ environmental complexity; and SEAP replaces the concept of ‘consolidation’ during sleep with interactional pruning while awake. Ultimately, the main argument in favour of SEAP (or something similar) is that long term memory must be a complex adaptive system, and that complex systems arise and are sustained along the lines we have described, and not in the way assumed by ‘representation–consolidation’ theories of memory.
Therefore, by re-conceptualizing the relationship between memory, sleep and the environment; SEAP provides a radically new framework for memory research, with implications for the measurement of memory and the design of empirical investigations in clinical, psychopharmacological and creative domains.
References
[1] Charlton BG, Andras P. Complex biological memory conceptualized as an abstract communication system–human long term memories grow in complexity during sleep and undergo selection while awake. In: Perlovsky LI, Kozma R, editors. Understanding complex systems (series): neurodynamics of cognition and consciousness, www.cs.ncl.ac.uk/publications/books/papers/315.pdf; 2007 [accessed 12 March 2009].
[2] A.R. Damasio, Descartes error: emotion, reason and the human brain, Putnam, New York (1994).
[3] J.A. Hobson, The dream drugstore, MIT Press, Cambridge, MA, USA (2002).
[4] J.L. Kavanau, Memory, sleep and dynamic stabilization of neural circuitry: evolutionary perspectives, Neurosci Biobehav Rev 20 (1996), pp. 289–311.
[5] N. Luhmann, Social systems, Stanford University Press, Palo Alto, CA (1996).
[6] B.G. Charlton and P. Andras, The modernization imperative, Imprint Academic, Exeter, UK (2003).
[7] G.M. Edelman, Neural Darwinism: the theory of neuronal group selection, Basic Books, New York (1987).
[8] M. Gazzaniga, Nature’s mind: biological roots of thinking, emotions, sexuality and intelligence, Penguin, London (1994).
[9] D.L. Hull, Science and selection, Cambridge University Press, Cambridge, UK (2001).
[10] J.L. Kavanau, Is sleep’s ’supreme mystery’ unraveling? An evolutionary analysis of sleep encounters no mystery; nor does life’s earliest sleep, recently discovered in jellyfish, Med Hypotheses 66 (2006), pp. 3–9.
[11] J.M. Siegel, Why we sleep, Sci Am (November) (2003), pp. 92–97. View Record in Scopus | Cited By in Scopus (18)
[12] P. Maquet, The role of sleep in learning and memory, Science 294 (2001), pp. 1048–1052.
[13] J.M. Siegel, The REM sleep–memory consolidation hypothesis, Science 294 (2001), pp. 1058–1063.
[14] M.P. Walker and R. Stickgold, Sleep, memory, and plasticity, Ann Rev Psychol 57 (2006), pp. 139–166.
[15] B.G. Charlton, Theory of mind delusions and bizarre delusions in an evolutionary perspective: psychiatry and the social brain. In: M. Brune, H. Ribbert and W. Schiefenhovel, Editors, The social brain – evolution and pathology, John Wiley & Sons, Chichester (2003), pp. 315–338.
[16] B.G. Charlton and J.L. Kavanau, Delirium and psychotic symptoms – an integrative model, Med Hypotheses 58 (2002), pp. 24–27.
[17] D. Healy, Psychiatric drugs explained (4th ed.), Churchill Livingstone, Oxford (2004).
[18] H.J. Eysenck, Genius: the natural history of creativity, Cambridge University Press, Cambridge, UK (1995).
[19] B.G. Charlton, Alienation, recovered animism and altered states of consciousness, Med Hypotheses 68 (2007), pp. 727–731.
[20] M. Seymour-Smith, Robert Graves: his life and work, Paladin, London (1982) p. 99–100 [1987 edition].
[22] B.G. Charlton, Scientific discovery, peak experiences and the Col-oh-nell Flastratus! phenomenon, Med Hypotheses 69 (2007), pp. 475–477.
