A review of research and literature in this area was undertaken, focussing on the biological and behavioural studies of the effects of sleep on memory consolidation; the neural structures in which memory processes are stored; the link between sleep stages and the consolidation of different forms of memory, and how the content of dreams reflect this.
The responses from the scientific and psychological communities to these studies were also examined, in particular the criticisms towards the methodological aspects of some of the studies, the inconclusive results from certain sleep deprivation studies, and the presentation of alternative hypotheses were explained. This wide review of literature on the debate from the last three decades ultimately leads to the conclusion that there is definitely a clear link between the processes of memory consolidation and dreaming, namely that dreaming is a by-product of ifferent types of memory consolidation processes in different stages of sleep, but the exact mechanisms by which this occurs needs more research to be fully understood.
Sleep is an extremely interesting phenomenon in which the mind almost completely departs from the usual realm of consciousness (Foulkes, 1999). Many theories have been proposed over the years as to why we need sleep. An essential part of sleep is the surreal world and narrative that is experienced during the state known as dreaming (Combs & Krippner, 1998).
Dreaming has been a source of fascination for many cultures over thousands of years, from primitive cultures to the ancient Greeks. Explanations for the functions of dreams have ranged from communicating with the gods, providing visions of the future, and divining a person’s character (Aristotle, translated by Beare 2007). Dreaming, in particular, has also been of particular interest to psychologists, beginning with Freud and Jung, who believed that dreams were a window into the subconscious and could be used to analyse the character of a person (Gross et al. 000).
Freud’s ideas are and always have been controversial, but they pioneered the topic of dream analysis and psychotherapy. Most of Freud’s theories have become obsolete with modern advances in medical science, but what remains is the concept that dreams must serve some kind of purpose (Gross et al. 2000). Most human behaviours have been shown to serve some kind of evolutionary purpose, so it makes sense that dreaming must have a function.
Among the general public, it is a widely accepted notion that being well rested improves memory capabilities, however in the sleep research community the relationship between sleep and memory is a very nuanced and debated topic (Frank & Benington 2006). With the discovery of rapid eye movement (REM) sleep, and importantly, its association with dreaming, a large body of research has focused on determining the exact functional purpose of REM sleep and dreaming. Amongst these theories is the hypothesis that REM sleep s an essential aspect of memory consolidation, and that dreaming is the conscious manifestation of the brain’s neuronal activity associated with the processes of memory consolidation (Feldman & Dement 1968, Payne & Nadel 2004).
This essay will ask the question: is there a link between dreaming and the processes of memory consolidation? The essay will positively argue this research question, by providing more recent theories about different types of memory, non-REM sleep, connections between different brain structures, and the content of dreams.
Firstly, the essay will examine what types of memory exist in humans, namely semantic and episodic memory, and what the definition of memory consolidation is, according to the standard model of systems consolidation (Frankland & Bontempis 2005). Secondly, it will review what the different stages of sleep are, and look at studies which indicate that episodic memory is consolidated during non-REM sleep, while semantic memory is consolidated during REM sleep. The essay will examine the role of the stress hormone cortisol, the level of which rises during REM sleep, and which impairs interactions between the neocortex and hippocampus.
This is relevant because episodic memory is stored in the neocortex, while semantic memory is stored in the hippocampus. The essay will examine how the content of dreams in all stages fits in with the argument that the neocortex, which stores episodic memory, actively consolidates memory during non-REM sleep, while the hippocampus, which stores semantic memory, actively consolidates memory during REM sleep. For the most part, this large body of research provides a compelling argument for the proposition that sleep plays a role in memory consolidation and that dreaming is a by-product of this process.
However, this view is not universally accepted. Therefore, this essay will also examine the criticisms of this theory, particularly in relation to the apparent lack of positive findings in some studies as well as issues with the methodology used by many of the studies supporting the theory. It will examine an alternative theory, namely that the function of REM sleep is merely to aid recovery from deep sleep, and attempt to relate this to the topic. Finally, the essay will explore the research into how recent brain imaging studies are providing mounting evidence in support for memory consolidation theory.
The essay concludes that sleep definitely does play a role in memory consolidation, but that more research is required to fully understand the role it plays. Memory consolidation Understanding what memory is has been described from behavioural as well as biological perspectives. Tulving (1983) considered memory as consisting of two different types, namely episodic and semantic memory. Episodic memory (which is also called declarative memory) is memory of the past and of specific events, including exactly when and where they occurred (Tulving 1972).
Semantic memory (also known as procedural memory) is memory of specific pieces of information and data, and is not coupled with place or time, for example, what a tree looks like, what the word “big” means, or how to perform a particular task (Tulving 1972). Payne and Nadel (2004) argue that the brain processes and shares episodic and semantic information in parallel to consolidate memory. Many neuroscientists have investigated the structures and functional mechanisms that the brain uses to record and consolidate memories.
McClelland et al. (1992) found that the hippocampus and neocortex brain structures are each suited to different types of memory storage. They argue that the structure of the hippocampus suggests that it is well suited to episodic and long term memory as its circuitry is akin to a partial map, which requires other pieces of information to complete the pattern. For example, memory of walking down a street requires specific information about what buildings or trees look like to complete the picture (McClelland et al. 1992).
On the other hand, the structure of the neocortex suggests it stores information as overlapping elements which are strengthened by similarities between them, and controls sensory perception, motor commands, spatial reasoning, conscious thought and language. This therefore makes the neocortex better suited for semantic memory (Kali & Dayan 2004). In this sense, episodic memory and semantic memory must work together, because episodic memory provides a sequential “template” of events, but lacks in specific details. This detail, such as the appearance buildings or trees, is drawn from semantic knowledge as the events are ecalled (Tulving 1972).
In terms of memory consolidation, the most widely accepted model is the Standard Model of Systems Consolidation developed by Frankland & Bontempis (2005). This model proposes that memories are initially registered in the brain neurons and then stored temporarily in the hippocampus and cortical regions of the brain. These memories are then reinforced (or “consolidated”) by being recalled or replayed in either a conscious or unconscious state. Physiologically, this is explained as the reactivation and strengthening of neural connections encoding these memories.
Stages of sleep and dream distribution In 1953, Aserinsky & Kleitman discovered a sleep stage called Rapid Eye Movement (REM) sleep, in which most of the body completely stopped, moving except for the eyes (which moved as rapidly as in waking life) and other core processes like breathing and heartbeat. Dement & Kleitman (1957) then described five specific sleep stages, one of which was REM, and the other four were collectively called non-REM (nREM). They documented five stages of sleep by examining 33 adults over a period of 126 nights of undisturbed sleep by using EEG scans and other qualitative data.
The authors found that the stages of sleep underwent regular cyclical variations throughout the night, alternating between REM and nREM. Almost all REM sleep occurred in the latter half of the night, while about 80% of nREM sleep occurred within the first half of the night. The periods of REM sleep became longer as the night progressed. Additionally, the researchers found that when subjects were awakened during REM sleep, dreams were often recalled, however no dreams were recalled when participants were woken during nREM sleep (Dement & Kleitman 1957).
Although Dement & Kleitman’s study (1957) was conducted on only 33 adult males, the research was robust and the results of the study have been replicated many times with other ethnicities, genders and age groups, and it is now widely accepted as general scientific knowledge (Gross et al. 2000). When they first wrote their study, they proposed that REM sleep is the only stage of sleep in which dreams occur, and because of this, REM sleep has been the centre of most dream research until very recently. But evidence has emerged to suggest that dreams do occur during nREM sleep.
For instance, in a 1962 study by Foulkes on eight males between the ages of 17 and 27, the participants were awakened during various stages of sleep over 57 non-consecutive nights of sleep. The stages of sleep were determined by EEG and eye movement activity. When awakened, the content of their dreams or thoughts were explored. Foulkes (1962) found that when the subjects were awakened during periods of REM sleep, their comments about their dreams emphasised feelings, moods, strong visual and bodily content (such as running), and did not relate to any specific events in their waking life.
When awakened during nREM sleep, a high percentage of subjects reported that they had been thinking about some event in their daily life. Although Foulkes’ study needs to be regarded carefully as it was only conducted on eight males, it still revealed that during all stages of sleep, the human mind is constantly active, and that cognitive activity in the brain is present in some manner during all stages of sleep. Foulkes’ study has been replicated using larger samples since then by Cavallero et al. in 1992 and Baylor & Cavallero in 2001.
REM and nREM dreams are quite distinctive in their content, as shown in Foulkes’ 1962 study. The responses given by the participants when they were woken up suggest that REM dreams rarely consist of specific episodes of waking life, and are usually quite hard to relate to specific events, while nREM dreams usually reflect recent or even distant events of waking life. This could possibly be explained by the idea that the various stages of sleep are responsible for consolidating different aspects of memory, and indeed, there are research results that support this view.
For example, Plihal & Born (1997) tested 20 healthy men’s episodic and semantic memory after different stages of sleep. They used a paired-associate task (recalling a word presented to them earlier from only the first three letters) to test episodic memory and a mirror-tracing task (trying to trace an image while looking at the hand only through a mirror) to test semantic memory. They found that recall of the paired-associate task improved significantly more after a three-hour period of nREM sleep than REM sleep, while mirror-tracing improved significantly more after a three-hour period of REM sleep than nREM sleep.
A problem with this experiment was that it was only conducted on men and, because it was constrained by time and money, the sample size could not be increased. Additionally, in their article, they define nREM sleep as early sleep (as it occurs earlier in the night), and REM sleep as late sleep, which suggests they might not have been very accurate with when they woke their participants up. However, the results of this experiment do support the view that REM sleep is mostly responsible for consolidating semantic memory, while nREM sleep is primarily responsible for consolidating episodic memory.
Physiological processes during sleep Other researchers have looked at some of the physiological processes occurring during sleep to determine whether these processes have any impact on memory consolidation. For example, the stress hormone cortisol and its impact on the brain during sleep have been investigated (Payne & Nadel 2004). One of the effects of cortisol is that it disrupts the connections between the neocortex and the hippocampus (Kim & Diamond 2002).
During the course of a night’s sleep, cortisol levels begin to slowly rise from about halfway into the sleep cycle, suddenly increase during periods which relate to the onset of REM sleep, and finally peak just before waking (Weitzman et al. 1971). This suggests that during periods of REM sleep, the neocortex and the hippocampus are unable to communicate due to the disruptive effects of cortisol. As episodic memory requires the input of semantic memory from the neocortex, no episodic memory consolidation could occur during REM sleep. Therefore, any memory consolidation during REM would have to be purely semantic (Kim & Diamond 2002).
More direct evidence of neocortical and hippocampal connections being disrupted during REM sleep is to be found in a study by Braun et al. (1997). They used PET scans to record the brain activity of 37 male volunteers during different stages of sleep and analysed the images by comparing different stages of sleep. They found that the activity of the prefrontal cortex significantly decreased during REM sleep. Although this study was only conducted on male volunteers, it is generally accepted to be scientifically sound and able to be generalised as the structure of the prefrontal cortex is identical in males and females.
Disorders in the prefrontal cortex are usually characterised by delusional symptoms which are somewhat similar to the content of REM dreams (Mitchell et al. 2000). Additionally, the hippocampus and prefrontal cortex are thought to work together in combining semantic and episodic memory, so deactivation of the prefrontal cortex could explain how the connections between the hippocampus and neocortex during REM sleep are disrupted (Weitzman et al. 1971). The content of dreams can also be examined in more detail to see if they reflect these specific types of memory consolidation.
In REM dreams, creative and fantastic narratives are generated with all kinds of bizarre content which rarely obey the laws of physics (Stickgold et al. 2001). However, these dreams aren’t often just a series of disassociated images conjured up and presented to the viewer; there is often an understandable narrative which, although nonsensical, still flows realistically (Foulkes 1999). Wagner et al. (2004) argued that sleep restructures memories if information for similar tasks is presented in a different way before different sleep intervals.
They proposed that sleep can be shown to facilitate problem solving in a creative and insightful way by demonstrating that a cognitive task that measures the time when insight into a hidden rule is acquired is twice as fast after sleep than after wakefulness. Jacobs and Nadel (1998) attempted to explain this as an inherent brain process active in both waking life and sleep in which the brain automatically attaches meaning and a narrative to ideas or images presented to it.
This “smoothing” process could indeed be the reason and source of the creativity associated with dreams, as the brain must quickly sort the semantic information presented to it in some logical order. In summary, the theory that dreams are important in memory consolidation is supported by behavioural, physiological, and cognitive research, which together provides a reasonable explanation of how dreams actually function. It is likely that during nREM sleep, the brain’s neocortical and hippocampal connections are active due to low levels of cortisol, so episodic memory consolidation can occur.
However, in REM sleep, high cortisol levels inhibit the neocortical and hippocampal connections, making episodic memory consolidation impossible and leaving dreams to consist entirely of semantic memories. These processes are reflected by the content of dreams, with REM dreams generally being fragmented or bizarre, although still containing a narrative (explained by mental “smoothing” of events), while nREM dreams reflect events of waking life. Arguments against sleep and memory consolidation However, the theory that dreams and sleep are important for memory consolidation is not universally accepted.
This is because some of the studies which have supported the theory have been criticised on methodological grounds, and many studies have failed to come to positive conclusions. For example, there have been many studies on humans and animals which have tried to look at the effect that sleep deprivation on memory consolidation (Lewin & Glaubman 1975; Frank & Benington 2006). Studies which have used total sleep, partial sleep or just REM deprivation have been inconclusive, and there have been almost as many studies in support of the theory as studies which refute it (Stickgold & Walker 2005).
The fact that there have been so many different experimental methodologies and different outcomes on memory consolidation in these sleep deprivation studies has provided opponents of this issue with a reason to reject the sleep-memory consolidation theory. A number of researchers (Horne & McGrath 1984, Vertes & Eastman 2000) argue the impairment of memory due to REM deprivation shown in many studies could be due to the stress caused by the experimental techniques used to interrupt REM sleep, possibly due to stress-induced elevation of cortisol levels.
The pedestal technique is one of the criticised REM interruption techniques in which mice are made to live on a rotating tilted pedestal above a water bath which will wake them up during REM sleep because the loss muscle tone allows them to roll. However, Fishbein and Gutwein (cited in Fishbein 2003) demonstrated that the pedestal technique does not induce stress in mice as long as they are allowed to move unrestricted around the cage. They showed that mice living in these conditions exhibit identical activity scores to mice living in standard shoebox cages (above average activity is an indication of high stress levels).
In addition, a number of other studies by Hairston et al. 2005 and Ruskin et al. 2005 have ruled out stress as a factor. Vertes & Eastman (2000) argue that the three major classes of antidepressant drugs that suppress REM sleep do not disrupt learning or memory, which they regard as evidence against a link between dreaming and memory. However, the memory that they were investigating was episodic, rather than semantic, and so their hypothesis actually supports the present theory as episodic memory would not be consolidated during REM. They provide an alternative theory that REM is not linked to memory consolidation.
Rather, it is a mechanism to provide constant stimulation of the brain throughout sleep, and its main purpose is to aid recovery from deeper stages of sleep. Central to this hypothesis is the idea that the chaotic EEG readings during in REM sleep are simply the result of the brain stem being highly stimulated and active, which randomly activates other components of the brain. Vertes & Eastman (2000) believe that there is very little functional value of this random activation in the consolidation of memory. But once again, Fishbein (2003) argues that this is in fact central to the processes of semantic memory consolidation.
He argues that if semantic memory is locked away in the brain and not stimulated regularly, it will degrade and disappear, so randomly activating chunks of semantic information strengthens them and prevents important knowledge from disappearing. This “maintenance” process would be more important to semantic memory than episodic memory because semantic information is used in everyday life and underpins our knowledge of the world. On the other hand, in the long term, episodic memory would only need to be recalled for a limited number of important specific events.
Even more tellingly, recent brain imaging studies (Hoffman & McNaughton 2002) have provided direct evidence of the replay and reactivation of memory traces during sleep. Hoffman and McNaughton (2002) used brain imaging techniques to record the activity of specific groups of neurons in Macaque monkeys during learning tasks. The pattern and sequence of firings of about 800 neuronal cells during learning tasks were observed to reoccur in a coordinated fashion during rest, indicating that the recently acquired memory traces were being consolidated.
This has also been shown to occur during sleep in rats (Battaglia et al. 2004). In light of these studies, Vertes and Eastman’s (2000) argument against the ‘memory consolidation in dreaming’ hypothesis is seen to be weak, as there is clearly a link, backed up by studies in behavioural, electrophysiological, cellular, and molecular fields to support the hypothesis. Frank & Bennington (2003) argue that the field needs different experiments to be repeated in different laboratories using consistent protocols to provide a more stable baseline to the different findings.
Indeed, the future task is, of course, to conduct further research into these fields to provide a complete and comprehensive understanding of the exact processes and functions that occur during dreaming, and exactly how important dreaming is to memory consolidation. The majority of evidence supports the theory that REM and nREM sleep does indeed play an important role in memory consolidation, and dreams are a by-product of the processes of memory consolidation while the body is offline.
However, there is still some uncertainty about which portions of memory are processed during sleep, and which parts are relevant for human memory. While the evidence is mounting to support this function, there is still much work to be done on the precise role of sleep in memory consolidation.