Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

5.1 Introduction

Sleep and dreaming have been addressed separately as independent phenomena instead of in a holistic perspective. Sleep was studied as part of the circadian waking-sleeping cycle. The introduction of the electroencephalogram, by Berger in 1927, and more recent imaging techniques have boosted the mostly empirical and observational studies to fully fledged scientific research. Multiple approaches have been used, from neurological, physiological and metabolic point of views, looking for biological rhythms and perturbations of different disorders. The task of distinguishing sleep from the other phenomena is now almost complete. However, most research considers sleep as a specific state without much room for cognitive activities, even less for brain control. Dreaming, as a phenomenon that happens only during sleep, has been regarded mostly from a psychological point of view. This gap has only very recently been partially filled with highly controversial theories of dreaming, such as the cognitive approach leading to a more objective and quantitative analysis of dreams distinct from the traditional psychological and subjective analysis, and more recently the proposed neurocognitive theory of dreaming, which adds the neurological aspects of dreams, and especially the REM–NREM controversies over dreaming.

A more medically oriented technique which has developed strongly since the dawn of electro cortical physiology is that of so-called evoked potentials, where electromagnetic brain patterns are recorded and studied during both the awake and sleeping states in response to somatosensory stimuli. Stimuli in afferent pathways and cortical connections have been found to be so consistent that a normative database has been established and used for diagnosis purposes.

Brain control (BC) research attempts to change brain waves or produce specific brain patterns. Two main streams have been pursued intensively, with specific applications in mind. First, brain-computer interface (BCI) work uses specific patterns of brain activity for activating or controlling certain computer mediated devices. Second, neuro-feedback (NF) uses feedback to mediate voluntary control of brain patterns. As we describe in detail in this chapter, specific brain patterns may have strong influences on brain functions, like memory and attention, but more importantly can be used as alternative therapy for many neurological, psychiatric or psychological disorders. The main advantages of this technique are that it is free of side effects and, more interestingly, that the training results are long lasting, with reported durations of more than 2 years after the end of the training sessions.

The focal point of this work is a holistic approach to working memory in its multiple manifestations, from simple sequence rehearsal up to artistic creativity. Many standard and well established tests are available for memory performance, but the quantification of creativity in general is still an open and controversial problem. Our working approach proposes a differential quantification of creativity based on an objective classification of dream contents.

Why do we dream, what is the function of dreams and what they can tell us? Our work covers a very specific aspect of the wider question, which is huge and still mostly unaddressed. The following sections describe relevant aspects of the different contributions, summarize the groundwork for the proposed research, and derive the integrative perspective. The final sections sketches a research proposal for exploring further the integrative approach.

5.2 Neurophysiology of Sleep

5.2.1 EEG During REM and NREM Sleep Stages and Correlations with the ANS

Figure 5.1 shows a hypnogram based on EEG recordings during REM and NREM sleep stages over the course of one night.

Fig. 5.1
figure 1

Hypnogram: Epochs for every 30 s throughout one night are classified into stages for REM and three depths of NREM: N1, N2 and N3

Full size image

Whereas in active wakefulness low amplitude high frequency waves are seen in the EEG, in slow wave sleep there is the opposite – high voltage and low frequency (delta) waves (Coenen 1998). In the waking state, thalamocortical neurons show low synchronization in their firing pattern. Their membranes are depolarized,Footnote 1 leading to sustained high activity that allows information to pass easily through the neurons. The correlation dimension is a quantitative estimation of the degrees of freedom of a time series and it is used to distinguish between deterministic chaos and random noise (Grassberger and Procaccia 1983). When used in studies for the analysis of the brain electrical activity, it is usually related positively to the amount of cognitive or mental processing (Lamberts et al. 2000) so, during wakefulness, this value is high. As cells undergo moderate hyperpolarization,Footnote 2 the consciousness level drops and afferent information reduces, by synaptic inhibition (Steriade et al. 1993), until slow wave sleep is reached. Here sensory inputs are largely blocked at thalamic level by the rhythmic burst firing of the thalamic network. Although the peripheral sensory organs keep sending impulses resulting from sensory stimuli, as the impulses reach the thalamic network, a stereotyped oscillation is produced that masks the stimuli. During REM sleep, the EEG pattern resembles wakefulness and thalamocortical neurons are also depolarized. The correlation dimension of REM EEG is also higher than for slow wave sleep (SWS) and sometimes can reach waking levels. The fact that this indicator varies during REM is in tune with findings in evoked potential studies that suggest different amounts of information processing in consecutive episodes of REM (Coenen 1998).

Sleep spindles have been recorded in human subjects during stage 2 NREM and during general SWS in animals. Their frequency ranges from 7 to 14 Hz and they are generated in the thalamus as a result of synaptic interactions (GABAergic inhibition) between neurons of the reticular thalamic nucleus (nRT), thalamocortical cells and cortical pyramidal neurons. Spindles can last from 1 to 3 s and recur every 3–10 s (Steriade et al. 1993). During this time, thalamocortical relay cells are moderately hyperpolarized with membrane potentials around −60 mV.

In late sleep stages, along with further hyperpolarization of thalamocortical cells, spindles reduce and lower frequency/higher voltage thalamocortical oscillations occur as delta (1–4 Hz) and slow oscillation (<1 Hz). The higher amplitude of delta oscillations implies that during this activity neurons fire synchronously. Delta oscillations during sleep might be responsible for the reorganization of cortical networks – synaptic connectivity patterns – and the regulation of biochemical activity between single neurons (Steriade et al. 1993).

Ponto-geniculo-occipital (PGO) waves occur during REM, have their origin in the pontine reticular formation and travel to the cortex. These waves are said to be the pacemakers of the thalamus and extended cortical areas as their neurons become depolarized and start firing in a tonic mode similar to wakefulness (Steriade et al. 1993). Meanwhile, neurons from the peripheral motor system are deeply hyperpolarized, resulting in the relaxation of almost every muscle except for those responsible for the movements of the eyes and extremities.

To summarize, during wakefulness there is the highest information flow through relay neurons, the EEG correlation dimension is also the highest and the characteristic EEG pattern is beta activity. When drowsiness starts, the membrane potential of thalamocortical cells drops, the EEG can show predominant alpha activity and spindles (complexity decreases), and the cortical network starts working in an oscillatory mode. During slow wave sleep, membrane potentials drop further along with the EEG complexity and external information processing. Delta activity is predominant in the EEG as neurons fire in burst mode. During REM sleep, neurons fire continuously and EEG complexity is variable, sometimes reaching values similar to those in wakefulness.

Heart rate variability (HRV) is used as a measure of cardiovascular autonomic regulation. More specifically, HRV measures the interaction between the sympathetic and parasympathetic activity in autonomic functioning. This measure has several approaches but for all of them the HRV time series must consist in a sequence of values that represent the interval in time between each R-wave peak in the electrocardiogram (ECG) (Figs. 5.2 and 5.3). Two of these approaches cover the time and the frequency domains. For the time domain, the standard deviation of the time series for each interval is used as a measure of HRV. For the frequency domain, a frequency analysis of the time series is performed (Figs. 5.3 and 5.4). With the frequency analysis, the high frequency (HF) – related to parasympathetic activity – and low frequency (LF) oscillations – sympathetic activity – are observed and their ratio (LF/HF) calculated in order to establish a balance between them (Herbert and Gaudiano 2001). Subjects with poor cardiac health associated with a slower reaction mechanism show a less erratic HRV time series compared to young and healthy subjects whose heart are more robust to interactive factors. Low HRV can be related to autonomic dysfunction and to an increased risk of future cardiac events (Manis et al. 2007).

Fig. 5.2
figure 2

HRV detection: compressed view of 5 min section of a sleep recording. The top trace (red) is the ECG (electrocardiogram), the second curve is the R wave interval detection, followed by the corresponding beat per minute curve (bpm), and the bottom curve is the periodogram power spectrum from 0 to 1 Hz (equivalent) of the R–R Intervals

Full size image
Fig. 5.3
figure 3

HRV bands calculation: screen view of FFT based and AR model based power spectrum of the RRI. The HRV is divided into VLF, LF and HF bands

Full size image
Fig. 5.4
figure 4

HF and LF power bands for every 5 min along a full night of sleep recording

Full size image

The HRV is used to analyze the activity of the autonomic nervous system (ANS) during the different sleep stages. For healthy subjects, the HRV time series shows dominant frequency components with different occurrences during sleep: the HF oscillations are predominant during NREM stages while LF oscillations increase during REM (Ako et al. 2003). In a study realized with seven test subjects, a significant negative correlation was found between delta EEG power and LF oscillations and LF/HF on the first to third NREM period while beta EEG power correlated negatively with HF in the fifth NREM period. Alpha EEG activity is also negatively correlated with heart rate and LF/HF ratio (Ehrhart et al. 2000). Very low frequency oscillations on HRV were found to be closely related to the occurrence of each sleep stage as well as to the SEF90 (spectral edge frequency below which 90% of power of the spectrum resides) (Zhuang et al. 2005). Another contrast between EEG and HRV is that changes in HRV can precede those in the EEG from 10 s to 5 min (Zhuang et al. 2005). The low and high frequencies during one night can be seen in Fig. 5.4.

5.2.2 Neurophysiological Interpretations

The two main probes of brain activity have been functional magnetic resonance imaging (fMRI) and electromagnetic scalp surface recordings. Both have advantages and disadvantages and are mostly complementary, therefore the current research trend is to use both methodologies, but so far the added value of such approaches have been rather limited and very specific. Another drawback of a combined approach is the complexity of the procedure and its cost: fMRI and MEG are far from affordable for routine use. The time resolution of fMRI is also rather limited: only recent high-power machines can achieve near real-time acquisition.

So the method of choice is still the old EEG, where long studies continued over day and night, multiple electrodes and channels, and high sampling frequency recordings are affordable, simple and robust.

The neurophysiology of the EEG is also well known and has been part of routine clinical practice for many years. There are six main rhythms:

  • The occipital alpha rhythm (from 8 to 12 Hz) plays a central role in the waking state and has known reactivity to visual processing (attenuation of alpha amplitude) and mental operations, with an increase in the alpha amplitude in a relaxed state or with eyes closed.

  • The theta rhythm (from 4 to 8 Hz) is associated with drowsiness and hypnagogic states and occurs mainly during the transition from awake to asleep.

  • The delta rhythm (from 0.1 to 4 Hz) is associated with sleep states, in particular of NREM sleep, and gives an indication of sleep depth and long term memorization.

  • The sigma rhythm (from 12 to 16 Hz) is also associated with sleep states and is attributed a sensory blocking function.

  • The beta rhythm (from 16 to 20 Hz) has been associated with intensive cognitive processing in the brain, both during sleeping and waking states.

  • The gamma rhythm is recorded during the waking state with specific procedures, is related to high functional capabilities, and can appear in both sleeping and waking states.

The rhythms are grouped as tonic activity, characterized by slowly varying characteristics and continuously present. By contrast, brief activities that appear suddenly in the EEG trace are known as phasic events and regarded as the system response to external or internal stimuli. In the waking state, phasic events are of low amplitude, with negative signal to noise ratio, and averaging is necessary to extract them. During sleep, the phasic events are usually discernible and easily spotted from the tonic EEG.

The boundaries of the different frequency bands are standardized by averages of the normative population. As a consequence, each individual measure suffers from this standardization. More representative or indicative results would be obtained if it were possible to determine individualized frequency bands. We therefore put the utmost emphasis on the importance of tailoring the analysis by the individual characteristic frequencies in the EEG rhythms.

5.3 Studies of Dreams, Content and Objective Outcomes

5.3.1 Introduction to Dream Theories

There have been several theories and assumptions about the process of dreaming. All these theories need to be reconsidered in view of new findings that appear every year in the field of neuro-imaging, medicine and psychology. To establish a relation with the physiology of sleep and dream content and to determine the role of dreaming in the regulation of the human metabolism and mental states, one needs to consider evidence from dream content analysis, brain functional activation, electroencephalographic data from the various sleep states and wakefulness, the influence of neuromodulators in the sleep states, the limitations and characteristics of dreams in patients with brain lesions and illness, and also in research with animals.

While most of the dream characteristics find explanation, or at least support, in findings in these fields, there are still some doubts about the functions of sleeping and dreaming. There is strong evidence in support of its role in the restitution of the brain, but evidence about how it affects memory, despite its growing volume, is still considered by some as insufficient. Anyway, several theories about sleep and memory exist and most of them have empirical support for some of their claims.

Here we discuss some of these theories along with the other findings from dream content related to functional activation, sleep states and neuromodulators.

5.3.2 Neurocognitive Theories of Dreaming

Dreams have well defined characteristics common to almost any healthy subject. Functional activation evidence from several studies (Dang-Vu et al. 2005) aims to establish a connection between dream characteristics and neurophysiologic aspects. For example, visual imagery during dreams is always present while auditory components are present in 40–60% of dreams and movement and tactile sensations on 15–30%, according to a study by Strauch (Strauch et al. 1996). The activation of temporo-occipital cortices in REM seen in neuro-imaging studies along with evidence that there is a cessation of visual imagery in dreams from some patients with lesions in these cortices (Ako et al. 2003) strongly relates them to these sensory components. Deactivation in the prefrontal areas of the cortex during REM sleep explains dream characteristics like lack of orientation, alteration in time perception and belief that one is awake, as their activation in waking states is responsible for these perceptions (Hobson et al. 1998, 2000).

Dreams in the REM stage are usually presented in a narrative manner although they suffer from some discontinuity with abrupt scene changes. These discontinuities were one of the aspects of dreams that were explained by the activation-synthesis hypothesis of Hobson and McCarley (1977) and were later found in 34% of 200 dream reports analyzed by Rittenhouse et al. (1994). However, these discontinuities, along with bizarre features, were overemphasized and it has been shown that during waking cognition (some experiences were in a darkened room but the subject was still awake) the level of discontinuity and spontaneity in thoughts is not so different from the dreaming experience (Domhoff 2005). Other similarities between dream and waking cognition are shown by Domhoff (1999a), e.g., the deficits present in the waking thoughts are also present in sleeping mentation and speech patterns in dreams have the same grammatical correctness as in waking life. Domhoff also attributes importance to Calvin Hall’s studies with dream content analysis, which showed significant continuity between dream content, waking conceptions and emotional concerns. Dreams that show more aggression are related to people with whom the subject has conflicts in the waking life. Dreams also have components related to recent waking activity in more than half the cases (Fosse et al. 2003). In a study by Maquet et al. (2000), areas involved in learning prior to sleep are reactivated, during REM, in subjects previously trained on motor learning tasks, as opposed to subjects who weren’t trained and showed no activation. These findings suggest reprocessing, during REM, of procedural memory acquired during previous waking states. Both amygdala and hippocampus, responsible for emotional conditioning and declarative knowledge in waking cognition (Bechara et al. 1995), are active during REM sleep and can participate in the processing of memory traces during this state (Ako et al. 2003). The amygdala is also responsible for activating the medial prefrontal cortical structures, associated with the highest order regulation of emotions, in REM sleep (Stickgold et al. 2001). Results from a study by Sterpenich et al. (2009) that compared memory consolidation of emotional and neutral pictures between sleep deprived subjects with well slept controls suggested that sleep during the first post-encoding night influences the long-term consolidation of emotional memory. Roffwarg and Walker’s work is shown as evidence of this characteristic, where the first dreams in the night include memory fragments from recent experiences and later dreams include fragments increasingly farther back in the past (Paller and Voss 2004).

During REM there is activation in the extra-striate cortex along with a deactivation of the striate cortex (primary visual cortex). During wakefulness, their activity is usually positively correlated and this pattern of functional connectivity suggests that REM sleep allows internal information processing in a closed system disconnected from the external world (via striate cortex) (Ako et al. 2003). Although some information can reach the cortical levels, the thresholds for awakening are lower for relevant stimuli than for irrelevant ones, suggesting a subconscious evaluation during sleep. In REM there is also no coordinated motor behavior (muscular atonia), but in subjects suffering from REM sleep disorders, coordinated behavior is observed that is usually somehow related to the dream narrative provided when woken up (Schenck et al. 1986). In contrast to “acting the dream” is the “controlling the dream” characteristic of lucid dreams. The presence of activity in the alpha band on the sleep EEG during REM can indicate an ongoing lucid dream. This dreaming state is characterized by more realistic content in the dreams as well as an improved consciousness of the self (Domhoff 2001a).

A neurocognitive theory of dreams that takes into account all these findings can be constructed. Domhoff has done much work in this direction by reviewing previous theories, such as activation-synthesis and psychoanalysis, and by proposing new elements for a cognitive theory (Domhoff 1999b, 2001b; 2005).

  • Dreaming is a gradual cognitive achievement that correlates with the development of visuospatial abilities. After REM awakenings with children with ages from 5 to 8, only 20% of them reported dream content and it was poor both in content and in visual complexity. The possible inhibiting effect of the laboratory in these dream recalls was shown to be irrelevant by Foulkes, by finding the same results in dreams collected at home (Weisz and Foulkes 1970).

  • Dream content and characteristics can suffer from specific neural defects or lesions. The location and nature of each defect can be compared to the changes in dream content.

  • Dreams express concepts about self and others.

  • It is most likely that dreams are the accidental by-product of sleep and consciousness.

  • Dream content is impermeable to pre-sleep and concurrent stimuli that have been used to influence it (Foulkes and Rechtschaffen 1964).

  • A new understanding in cognitive linguistics needs to be established for the neurocognition of dreams, in order to understand their frequent figurative aspects – metaphors, resemblance metaphors, metonymies and conceptual blends.

  • Some dreams may use the same system of figurative thinking used in waking cognition.

In light of these findings, the activation-synthesis and psychoanalytic theories are found to be incorrect in many aspects, some of them pointed out by Domhoff. First, the existence of dream reports after NREM awakenings invalidates the possibility that dreams only occur during REM. Dream content seems to be much more related to the waking life than to spontaneous image activations, which depend on the area of the brain stimulated, or to the representation of one’s unconscious needs, fears or fantasies. Developmental changes between the ages of 3 and 8 in dream content and its subsequent continuity also deviate from what is claimed in both of these theories. Figures 5.5 and 5.6 show the most common characteristics and elements present in the dreaming context.

Fig. 5.5
figure 5

Characteristics of dreams

Full size image
Fig. 5.6
figure 6

Elements in dreams

Full size image

5.4 Neurophysiological Correlates of Dreaming

5.4.1 REM and NREM Sleep and Dream Collection

The REM sleep state was described in 1953 by Eugene Aserinsky and Nathaniel Kleitman. Along with the rapid eye movement, EEG patterns and autonomic nervous system changes were also observed, which led to the conclusion that these phenomena were all manifestation of a particular level of cortical activity encountered normally during sleep (Aserinsky and Kleitman 1953). For a long period of time, there was a strong belief that dreams only occurred during REM sleep so most of the dreams were collected after this sleep state. In fact, sleep mentation also occurs during NREM but, as reported by Nielsen, with a lower recall after waking. In his review, there was an 80% average of REM dream recall rate and a 50% average of NREM dream recall rate (Nielsen 2000).

These dream recalls are reports from dreams collected in various ways. The most efficient and systematic one is in sleep laboratories. Four or five dream narratives can be acquired in a single night by a subject when this subject is awakened during REM periods or NREM periods to maximize the probability of recall (Foulkes 1979). Dream collection can also rely on dream diaries requested by a dream researcher, personal dream journals and dream discussions recorded in psychotherapy (Domhoff and Schneider 1998). A more recent method of dream collection, adopted because of the lack of funding for dream studies, is the Most Recent Dream method. The dream report is collected in an interview fashion. The interviews last 15–20 min and are done in convention halls, waiting rooms and classrooms. The samples are representative if at least 100–125 reports are collected for each age group (Avila-White et al. 1999). Foulkes states that the reports collected by this method do not differ from those collected in sleep laboratories when controls are introduced (Lamberts et al. 2000) but there are different opinions about the reliability of this method.

To analyze the content of these dream reports, several methods are applied, such as asking the dreamer for the meaning of each element in the report (without censoring) or finding metaphoric meanings for the dreaming narrative and searching for repeated themes in series of dreams. For a quantitative analysis to be reliable, it is necessary to define clear categories of dream content that lead to the same results when used by different investigators so that their findings can relate quantitatively to others. This is achieved by content analysis. The scales used can be based on rating each element or can be nominal, reporting just the presence of an element or characteristic. Hall and Van de Castle developed a set of nominal categories that are widely used in content analysis (Domhoff 1999b).These categories include characters, social interactions, activities, striving, misfortunes and good fortunes, emotions, physical surroundings and descriptive elements. Each category is further subdivided into subcategories. For example, social interactions are subdivided into friendly, aggressive and sexual (Manis et al. 2007). These categories allow the use of indicators that consist of the ratios between groups of categories. For example, the “animal percent” is the ratio between characters that are animals by the total number of characters (Manis et al. 2007). Finally, the indicators of one group are compared with the norm to search for statistically significant differences.

5.4.2 Memory and Learning During Sleep

Several insights about dream content, especially during REM sleep, have been presented. Memory reactivation is one the characteristics of the dreaming experience. We shall now analyze it in more detail.

Declarative (or explicit) memory is referred as the ability to recall and recognize episodes and complex facts and has two subclasses: the episodic memory, which consists in personal facts and experiences, and semantic memory for impersonal facts (Vertes 2004). According to Paller, after an analysis of several studies (Paller and Voss 2004), this aspect of memory depends on multiple neocortical regions and requires “cross-cortical storage”. This means that the memory fragments by themselves are not sufficient to comprise the memory but need to be linked together for it to exist. So the cognitive characteristics of declarative memory probably include a reliance on relational representations (Eichenbaum 2001). Episodic memory, one type of declarative memory, is encoded by the hippocampal networks as sequences of events and places where these traces were created (Paller and Voss 2004). Cross-cortical storage is in tune with other models that suggest a gradual formation of the representation of memories in the cerebral cortex. According to this model, declarative memories exist in a fragile state after initial learning. The hippocampus has a major role in establishing these new connections between separate cortical representations, and these quick links are temporary. After sufficient cross-cortical consolidation, the hippocampal neurons are no longer necessary to unite this set of distinct cortical networks and the declarative memory endures. This process depends on the nature and frequency of memory access from the initial acquisition to the retrieval. These individual memories can change after their initial learning period. So two things can occur when declarative memory is accessed: the fragments of episodic of factual representation become more strongly bound together or they become more strongly linked to new representations of that fact or to related stored information. These connections to stored information provide additional routes for subsequent retrieval and this gives the stored information connections with aspects disjoint in time. The declarative memory network changes with post-acquisition processing, and if this processing provides cross-cortical consolidations these memories can be retrieved in the absence of the hippocampus. The cross-cortical storage links different components of memory together. The memory of an event can include individual links to each item of the event as well as to the context when this learning occurred and emotional experiences created by this event. The conscious experience of remembering declarative memories also depend on such aspects as attention, cognitive control and working memory.

Duda relates two types of amnesiac patients and their memory limitations (Dudai 2004). Amnesiacs who have difficulties recalling recent events but whose remote memory is spared (temporally graded retrograde amnesia for declarative memory) usually show damage in the medial temporal lobe, hippocampus formation and associated cortices. Patients whose amnesia extends over many decades usually show additional damage in the neocortex in the lateral and anterior temporal lobe. This suggests that memory storage is related to the neocortex while the hippocampus formation is related to a time-limited retrieval of intact declarative long-term memory.

Dudai also defines two types of memory consolidation: synaptic consolidation and system consolidation (Dudai 2004). The synaptic consolidation model consists in the consolidation of short-term and labile memories to long-term and stable ones by means of long-term modification of synaptic proteins and in the remodeling and growth of these synaptic connections. The activation of this process occurs in a limited time window during and immediately after the information acquisition and can be disrupted by various types of agents such as inhibitors of protein synthesis. System consolidation explains how older memories are secured by shifting their storage from the hippocampus to the neocortex over a period of weeks.

There are three different views of consolidation of a memory trace, differing in how they account for reconsolidation. The “weak” view states that, when retrieved, a memory trace only has its new parts updated. The existing parts are not changed because consolidation only happens once. The “intermediate” view claims that after retrieval, memory traces become modifiable and labile, and so need to undergo a new process of stabilization. If an appropriate interference occurs, parts of a trace can become corrupted. The “strong” view differs from the intermediate view by including the new information to be processed in the list of parts that can become corrupted.

Procedural memory is a type of non-declarative memory that involves unconscious acquisition and utilization of perceptual and motor skills (Vertes 2004). Despite this difference of role, procedural memory traces are also reconsolidated when accessed.

While it is now evident that declarative memory is accessed during sleep, there still is a void about their consolidation and the role of dreams and sleep stages in this process. Evidence in these areas exists, but it is fragmented and sometimes contradictory. Fortunately, this has stimulated several studies and theories. Because most of the content in dream reports show that sleep mentation is flooded with fragments of recent events and knowledge, Paller assumes that during sleep cross-cortical consolidation of declarative memory traces also occurs (Paller and Voss 2004). He also hypothesizes that these connections between memories can be central for problem solving and that connections between fragments, if they occur on larger scales, can connect episodic memories with behavioral strategies (these behavioral strategies are taken as cognitive fragments based on past experiences and, as they consolidate, become aspects of the subject’s personality). A similar view is presented by Rasch and Born where newly acquired memory traces are gradually redistributed, from fast learning regions of the temporal lobe (hippocampus mainly), to neocortical regions by strengthening cortico–cortical connections during off-line periods (Rasch and Born 2007). Repeated reactivation of the new information is essential for this process and it is accompanied by reactivations of related older memory traces. This pattern of reactivation is thought to gradually integrate the new memories to the pre-existing network of long term memories and may produce qualitative changes in the respective memory representation. The authors present several studies that show the influence of slow wave sleep in declarative memory reactivation in animals. During post-learning SWS periods, firing patterns in the hippocampus similar to the ones present during learning were observed and it is suggested that the learned information is repeated during sleep. Furthermore, it was observed that in rats these reactivations in the hippocampus alternate with reactivations in the visual cortex, an aspect that can be seen as the gradual transfer of the replayed information from the hippocampus to the cortical regions (Ji and Wilson 2007). These results led to the replication of the studies in human subjects, where the same reactivation of hippocampal regions that were activated during a spatial navigation task (declarative memory) was observed, along with evidence for a positive relation between the performance of memory retrieval the next day and the amount of activity in the hippocampus during SWS (Peigneux et al. 2004). Still, slow wave sleep is characterized by a reduced ability to induce synaptic long term potentiation (LTP) while the REM stage seems more appropriate for that effect due to the presence of more neurotransmitters and modulators responsible for synaptic plastic changes. Gais and Born attribute importance to the sequence of SWS and REM for memory consolidation as the hippocampal reactivation serves to tag the synapses that are then strengthened in REM (Gais and Born 2004b), or in wakefulness if REM is not possible (Morris 2006).

Other studies by Gais and Born (2004a) suggest that slow wave sleep is responsible for declarative memory consolidation and that this consolidation cannot be achieved without SWS’s characteristic low levels of the neurotransmitter acetylcholine in the hippocampus. Still, SWS length does not increase after increased declarative learning and declarative memory performance does not increase with the amount of SWS. According to Gais and Borne, high levels of cholinergic activity in SWS, induced by post-trial administration of physostigmine (cholinesterase inhibitor), completely eliminated the consolidating effect of sleep on hippocampus-dependent declarative memory.Footnote 3 These were results from a word pair test compared to a control population that was given a placebo instead of physostigmine and also to a set of experimental and control populations that stayed awake. Hippocampus-independent memory was also tested with a mirror tracing task and no relation was found between subjects administered physostigmine and control subjects. Further studies from the same authors assign acetylcholine the role of a switch between modes of acquisition and consolidation of memory traces, as its low levels potentiate consolidation in SWS and high levels during wakefulness support encoding (Rasch et al. 2006).

Another study showed that it was possible to externally induce reactivation during slow wave sleep by presenting the sleeping subject with cues that were presented during the previous learning period (Rasch et al. 2007) and that this led to increased consolidation. An odor cue was used because it is the least sleep disturbing stimulus. It was observed that this cueing affected neither sleep architecture nor sleep EEG and subjects did not recall the odor treatment during sleep. In the experimental procedure, this stimulus was presented while subjects performed learned object locations in a two-dimensional object location memory task (declarative memory) in the evening before sleep and represented during the first two periods of subsequent SWS. Alternative trials included presenting the odor only during SWS, only during learning and REM and only during learning and waking. The control trial did not include an odor cue in any condition. Improvements in location recall compared to the control trial were only seen in the first experimental procedure accompanied with activation in the hippocampus. To test the odor cueing effect on procedural memory, a finger tapping speed test, with the same procedures as before, was also done but did not show any significant improvements.

Transcranial direct current stimulation (tDCS) also changes the process of consolidation in declarative memory when applied during SWS. Declarative memory consolidation was improved (word pairs) relative to a control group whereas procedural memory was not.

Although memory consolidation is improved, the accuracy of the recalled memories after sleep may not be perfect and false memories can be originated. In a study that used the Deese–Roediger–McDermont (DRM) paradigm for memory recall, an increase was observed in both veridical and false memories in subjects retested after sleep compared to subjects who were retested after the same period of daytime wakefulness (Payne et al. 2009). The DRM paradigm consists on a list of words learning task and the recalled words are interpreted as veridical, false and critical. Veridical words are the words presented on the list, semantically associated words are interpreted as critical and unrelated words are interpreted as false. This study shows that there is an increase in the recall of critical words over learned words after a night of sleep compared to equivalent periods of wakefulness (nevertheless learned words recall also increased compared to equivalent periods of wakefulness). After a daytime nap the recall of critical words was 50% greater than the control recall while there was no significant increase in the recall of veridical words. This at least suggests that sleep enhances recall not only of the exact words but also of the surrounding network of semantically related words, thus pointing to an effect of strengthening and creating connections between memory traces.

Although procedural memory is not improved in the previous experiences, it is a robust and evident finding that it benefits from sleep (Walker and Stickgold 2006). It can be divided into several functional domains, such as perceptual – visual and auditory – and motor skills.

Motor adaptation (like learning to use a computer mouse) and motor sequence learning (like learning a piano scale) are two forms of qualifying motor skills. Some of the following studies presented were reviewed and related to each other by Walker (Walker and Stickgold 2006).

It has been demonstrated that a night of sleep after learning a sequential finger tapping task can produce significant improvements in speed and accuracy while the same amount of wake time provides no significant benefit (Walker et al. 2002). In this study the improvements were positively correlated to stage 2 NREM sleep, but this relation is not consistent between experiments. In other study by Kuriyama et al. the speed of each step in a motor sequence, before and after sleep, was taken into account. It was found that sleep-related improvements were not equal among steps (Kuriyama et al. 2004). In each sequence, some steps were more problematic than others and this was reflected by decreased speed compared to the easier ones. Again, significant improvements were only verified on subjects who slept between learning and testing, and these improvements were greater in those slower steps. Walker interprets these results as a characteristic of sleep-dependent consolidation where a single memory element, comprised of smaller motor memory units, is selectively improved, giving priority to the more problematic units. This suggests that sleep not only increases the strength of existing memory representations but also does it in a selective manner. The need to sleep in the first night or day following the learning, for improvements in a finger tapping task to be noticed, is also confirmed by Fischer (Fischer et al. 2002). In this study, the improvements seem to be more related to the amount of REM sleep than to stage 2 NREM. This finger-tapping test was different from the others because it consisted of finger-to-thumb movements rather than keyboard typing like the previous. Walker hypothesizes that the REM correlation is due to the novelty in the procedure, because keyboard typing is a simple variant of a well learned task while finger-to-thumb movements introduce new tasks (Walker and Stickgold 2006). Other studies that use selectively sleep-deprived subjects on the first night prior to learning show that retention of a visuomotor adaptation task suffers more with the deprivation of stage 2 NREM and that the learning of a motor-reaching adaptation task was accompanied by a discrete increase of the subsequent SWS activity at the start of the night, where this increase was proportional to the enhancement of the tested motor skill on the next day (Walker and Stickgold 2006).

In other studies, areas involved in learning prior to sleep were seen to be reactivated during REM in subjects previously trained on motor learning tasks while subjects who were not trained showed no activation (Maquet et al. 2000). During REM after procedural task learning, the firing sequence of hippocampal neurons seemed similar to that of the task learning phase (Battaglia et al. 2004). According to Paller, this can serve to strengthen synaptic connections between neurons that process information in a manner relevant to performing the task by mechanisms for circuitry remodeling produced by LTP (Paller and Voss 2004).

Visual and auditory sleep-dependent skill learning is also confirmed by studies of the same nature as the previous ones. Learning a visual or pitch discrimination task also shows significant improved results following a night of sleep compared to the same amount of wake time, where no improvement is shown (Walker and Stickgold 2006).

Daytime naps also seem to improve procedural memory. There is an interesting effect seen in a study that applied a sequential finger tapping task (Walker and Stickgold 2005). Two groups were trained on a task in the morning and tested. One of the groups remained awake until the retesting, later in the day, while the other underwent a midday nap of 60–90 min. When retested, the group whose subjects napped showed 16% improvement whereas those who did not nap showed no significant improvement. The interesting fact is that on the next day, after both groups slept through the night, when retested, the control group showed 24% improvement and the nap subjects showed 7% improvement (improvements compared to the previous retest, not the first). So, on the next day, both groups showed more or less the same improvement. The difference is that the first had 23% improvement split in 16% nap plus 7% night sleep while the second had the whole 24% improvement in night sleep.

Still in the field of procedural memory improvement due to sleep, one characteristic that is sometimes interpreted as a limitation of this learning is that it is specific to the sequence learned and to the limb used (Vertes 2004). Improvements gained in one procedural task are not present on other tasks, and similarly for the limb used.

Benefits from sleep were also found on working memory (Kuriyama et al. 2008). In a test used to determine working memory capacity and response time, subjects were divided into three different groups. Group A trained at 8 a.m. and were retested at 3 p.m. and 10 p.m. without any sleep in between. Group B trained at midday and were retested at 10 p.m., went to sleep and were again retested at 8 a.m. on the next morning. Finally, group C trained at 10 p.m., went to sleep and were retested at 8 a.m. and later at 6 p.m. This different training and retesting hours are important to see in which case working memory benefited most. The results showed that improved working memory results only showed up in retests after sleep and there was no time of day specific to better results (which excludes a circadian influence).

Besides declarative, procedural and short-term memory, there are also other aspects of cognition that seem to benefit from sleep. In a study by Wagner, subjects were required to perform a mathematical operation on a string of digits (Wagner et al. 2004). The time spent on the operations could be significantly reduced by applying a certain rule that was left for the subjects to discover. The insight of this rule was more frequent when the training was followed by a night’s sleep than by a sleepless night or the equivalent time awake during the daytime.

Now it seems evident that the role of sleep is more than physiological recuperation or a period of inactivity to keep our primitive ancestors idle during the night, and that dreams can be the manifestation of some of the processes that occur during this period.

Having these findings in mind, it is possible to propose some characteristic relations between sleep, memory and known dream elements and to test new possibilities for their interplay. For example, some environments or logical sequences of episodes during a dreaming episode can be due to the activation of a memory trace and further chain of activation of the linked traces. Because these traces were linked together by some logical or temporal relation, when reactivated, the result also seems logical instead of the random activation proposed by the activation-synthesis theory. This could be one way to create a coherent dreaming scenario.

5.5 Sleep and Dreaming as Targets for Interventions

The main result of our integrative approach is a conceptual and practical framework that enables different analytic perspectives of the brain through the EEG recordings. The aim is to segment the continuous EEG into different functional sections, for restorative processes, memory related functions (storage, consolidation, recall, association, etc.), and cognitive processes (expectation, attention, decision, deduction, etc.), not only during the awake state but especially during sleep.

5.5.1 Cognitive Processing During Sleep and Dreaming

The degree of consciousness or cognitive function in the brain during sleep is still a controversial and debated topic. One possible source of answers or new arguments for the discussion is research conducted in accordance with the following hypothesis:

  • Cognitive processing in the waking state can be characterized by its EEG characteristics and these are assumed to be the same during dreaming.

  • The dream recall immediately after a REM episode is a faithful report of the dream content at least regarding the time sequence of the salient contents.

  • The Van Castle dream classification is an adequate tool for objective dream content classification into categories that can be associated with cognitive processes.

  • NREM dreaming can be treated in the same way as REM dreaming.

The research protocol is straightforward. Subjects under study are monitored with standard EEG recordings while awake and performing several cognitive tasks and again during sleep, when they are awakened every 10 min during REM episodes and their dream recall narration digitally recorded.

The dream content of each episode is analyzed by the Van Castle technique in order to establish the time sequence of cognitive states. The awake EEG is used to determine the individual alpha frequency (IAF) and the theta frequency (TF) in order to establish the alpha-1, alpha-2 and upper alpha bands, then the lead and significant electrodes and bands are calculated for the different cognitive tasks. This process is repeated for every REM dream episode resulting in sequences of cognitive processing states. A correlation between the sequences obtained by the dream recall or EEG characteristics is calculated in order to establish a plausible causality. A significant individual and/or group positive correlation indicates the presence of cognitive processing during the REM dream episode.

The same research can be applied to NREM sleep stages 2 and 3. If the correlation is still positive then we have strong evidence that the EEG characteristics identified are indeed descriptors of cognitive processing during sleep and the whole night’s sleep can be analyzed under this perspective.

A final procedure is the recording of uninterrupted sleep with whole-night dream recall the next morning. The recall is then correlated with the cognitive processing analysis of the whole night EEG. If a positive correlation is found, it might indicate the possibility of estimating the dream content in cognitive processing terms without the need for dream recall.

5.5.2 Memory Enhancement

As described above, short-term memory recall has been shown to be positively correlated with alpha frequency band. It has been reported that shorter reaction times and longer recall sequences are positively related to the IAF. The underlying hypothesis includes the assumption that the IAF can be changed by some sort of training procedure, i.e., that the neuro-feedback paradigm is an effective way to teach the subject how to control the brain EEG characteristics. Subjects with surface electrodes recording in specific regions of the scalp EEG are asked to increase in a sustained way and for a controlled period of time a feedback signal extracted from the EEG, which can be the power of a specific band of alpha, the frequency of the IAF, a alpha-theta ratio, or coherence measure between two EEG channel, etc. The objectives are progressively and adaptively increased between sessions. The subject usually learns to control the EEG characteristics after six training sessions and a total of 20–30 sessions are required to obtain the first significant results.

5.5.3 Creativity Enhancement

Creativity is still a much debated concept and many definitions exist. The concept of creativity used here is the production of any new idea or object by new means that are not exactly like any seen or proposed before. The precise definition of creativity is not very important, because it is almost impossible for anyone to create an idea or an object that has never been partially suggested or partially produced before. However defined, the creativity must be confirmed by peers. The basic idea is the production of different and varied ideas from existing knowledge during sleep, mainly by associative combinations of atoms or acquired ideas already present in the brain. The hypothesis is that during sleep the available ideas are combined or associated in a random way. More ideas mean more material for combinations, increasing the chance to produce new ideas or objects. Since these objects are obtained by free combinations, it is rather unlikely that these ideas will result in a useful or important idea. It is also hypothesized that while the subject is awake all the combinations are filtered by a more conscious or rational filter, thus limiting the chain of innovative ideas. By contrast, in REM sleep, there is no filtering, resulting typically in bizarre and mostly impractical ideas. The virtue lies in the middle, a state between being awake and REM sleep, known as light REM, where the brain state is between REM and awake states. The practical implementation of this experiment is to deliver subliminal stimuli to the subject during REM sleep. These subliminal stimuli will lighten the REM sleep toward a waking state, obtaining an intermediate filtering state, hopefully with combinatorial power and at the same time some realistic natural filtering. This setup will promote a more open and well behaved subconscious state.

5.6 Conclusion

During sleep it is possible to distinguish different stages occurring in a cyclic pattern. Each of these stages shows unique characteristics in the brain activity, heart rate and muscle tone. Although the exact reason for the presence of these different stages during sleep is not yet established, several studies assign different roles to each one of them. This brings up the question of their meaning and origin. If the length of sleep is due to our ancestors’ primitive need to remain inactive through the night, because of the lack of light, then today, as the problem of illumination no longer applies, its importance may be exaggerated. Is our circadian clock out of date, keeping us from being awake more of the time because of a self-defense mechanism that is no longer necessary? Without the knowledge presented in this chapter, one could easily agree with this observation. It seems that the restorative role of sleep does not act only on the metabolic processes of the body, since the cognitive processes are also influenced. Memory consolidation and reasoning capability are examples of processes that are affected by sleep deprivation or even selective sleep stage deprivation. It is possible that each stage has a more relevant role for each cognitive process, but in the case of memory consolidation the evidence points to the importance of the interplay between stages.

Before the recording of brain activity or heart rate was possible, dreams were the only window to the sleeping process. Today, all the recordings are taken into account and correlated with each other. The results enable us to see that some regions of the brain, responsible for some emotional states for example, are activated during sleep when the corresponding emotional state is present in the dream. Other previously known characteristics of the dreaming scenario are also explained by activation or deactivation of brain regions that are known to be responsible for those characteristics in the waking state. However, these activations are not totally random, since reports of dream episodes show some consistency in their plot. One possible explanation for this consistency is the activation in sequence of several memory traces that are connected to each other. Because these traces were connected during the same time period or in the same context, when reactivated they produce a consistent plot. In other words, everything we know is connected somehow in our brain. During sleep, these fragments of knowledge are activated and linked to each other, producing the dream episode. For example, it is possible that during a dream, if the right associations occur, a solution to a problem that seemed unsolvable is achieved. Such a solution would result from a very improbable event, but if the way associations are made could be induced somehow, this kind of event would be less improbable.

In summary, even if the time one should spend sleeping is debatable and varies between individuals, it is not fair to underestimate its underlying importance.