**4.1 Overview of human EEG data**

Sleep spindles, the major hallmark of stage 2 of non-REM sleep, have been for long proposed to be associated with human individual cognitive abilities or intelligence. For example, Bódizs et al. (2005) found that both grouping of fast sleep spindles by cortical slow oscillation over the left frontopolar derivation (Fp1) and fast sleep spindle density over the right frontal area (Fp2, F4) during stage 2 of non-REM sleep, correlated positively with general mental ability. Further, a robust positive correlations were found between slow (< 13 Hz) and fast (> 13 Hz) spindle activity in stage 2 of non-REM sleep and both individual cognitive abilities and implicit/explicit memory-related abilities (Schabus et al., 2006). Later, Fogel et al. (2007a) showed first that number of spindles in stage 2 of non-REM sleep remains relatively stable within individuals from night to night. Second, the authors demonstrated that the number of spindles and EEG power of sigma (slow spindle) activity were positively correlated with performance intelligence quotient (PIQ), but not with verbal IQ. Also, perceptual/analytical skills measured by the PIQ accounted for most of the interindividual differences in spindles. Interestingly, in the same study, a relationship between rapid eye movements in REM sleep and VIQ in individuals with higher IQ scores was also demonstrated (Fogel et al., 2007a). However, Tucker & Fishbein (2009) demonstrated that while subject's intelligence correlated positively with pre-sleep acquisition and post-sleep retest performance on both procedural and declarative tasks, it did not correlate with over-stage 2 of non-REM sleep spindle events. These findings suggest that intelligence may not be a powerful modulator of sleep's effect on memory performance (Tucker & Fishbein, 2009).

One major question arising from the above described findings is whether sleep contributes to human intelligence in a state dependent manner (i.e., by providing neurobiological conditions for memory consolidation and reconsolidation), whether sleep is associated with intelligence in a trait dependent manner (i.e., strictly individual sleep characteristics are

The Memory, Cognitive and Psychological Functions of Sleep:

obtained (Durrant et al., 2011).

**4.2 Overview of human neuroimaging data** 

Update from Electroencephalographic and Neuroimaging Studies 167

paradigm, and searched for a predictive relationship between the type of sleep obtained and subsequent performance improvements (abstraction). Participants who consolidated over either a night of sleep or a nap improved significantly more than those who consolidated over an equivalent period of daytime wakefulness. Importantly, PSG revealed a significant correlation between the level of improvement or abstraction and the amount of SWS

The existing so far neuroimaging data concerning sleep and cognition do not reveal patterns of brain reactivation during sleep. Instead, they reveal altered brain activities associated with impaired cognitive processes in response to sleep deprivation. One of the most important findings is presented in the study of Yoo et al. (2007b). In this study, two groups of subjects were randomly assigned to either a sleep deprivation (SD) or a sleep control (SC) group. All subjects underwent an episodic memory encoding session during fMRI scanning, in which they viewed a series of picture slides and were retested two days later (after two recovery nights of sleep) for a recognition test session (without fMRI). Compared with SC group, SD subjects displayed much worse recognition at retest after two days, though they had two nights recovery sleep. This finding clearly shows that sleep deprivation impairs learning. The fMRI scans at learning/encoding demonstrated a significant impairment of activation in the hippocampal complex in the SD relative to SC group, a region known to be of critical importance for learning of new information (Yoo et al., 2007b). Another study showed that compared with normal sleep, SD produced impairment of spatial attention that correlated with a reduced activation in the posterior cingulate cortex, as measured by fMRI (Mander et al., 2008). Similarly, it is shown that SD produces lapses of attention manifested as delayed behavioral responses to salient stimuli. To identify changes in task-related brain activation associated with lapses after SD, fMRI scans during a visual selective attention task were conducted. It was demonstrated that SD-related lapses in attention corresponded to (1) reduced ability of frontal and parietal control regions to raise activation in response to lapses, (2) dramatically reduced visual sensory cortex activation and (3) reduced thalamic activation during lapses (Chee et al., 2008). Another fMRI study by the same group tested the hypothesis of whether SD impairs short-term memory due to reduced storage capacity or whether it affects processes contributing to appropriate information encoding. Scans were conducted during performing a short-term memory visual task and during presenting varying visual array sizes without engaging memory. Whereas the magnitude of intraparietal sulcus activation and memory capacity after normal sleep were highly correlated, SD elicited a diminished pattern of activation on both tasks, indicating that deficits in both visual processing and visual attention account for loss of short-term memory capacity (Chee & Chuah, 2007). Further, an fMRI study showed that SD abolishes selective attention in association with a decreased BOLD signal found within fronto-parietal cognitive control areas (the left intraparietal sulcus and the left inferior frontal lobe) and parahippocampal place area (PPA) during a selective attention task performance. Additionally, SD resulted in a significant decrement in functional connectivity between the PPA and the two cognitive control fronto-parietal areas (Lim et al., 2010). Interestingly, a single night of SD was shown to produce a strategy shift during risky decision making such that healthy human volunteers moved from defending against losses to seeking increased gains. An fMRI assessment revealed that this change in economic preferences was correlated

related to individual intelligence), or whether sleep and human intelligence are related in both ways (Geiger et al., 2011; Fogel & Smith, 2011). It has been previously proposed that sleep serves human intelligence by complex interactions between its unique physiological and mental states, and individual sleep patterns and cognitive traits (Kirov, 2007). One study has addressed this question providing some evidence that stage 2 of non-REM sleep spindle increase after learning is related to elaborate encoding before sleep, whereas individual's general learning ability is well reflected by inter-individual (trait-like) differences in absolute sleep spindle activity (Schabus et al., 2008). However,, the precise mechanisms involved in the complex association between sleep and intelligence remain so far poorly understood (Kirov, 2007). As will be seen in the paragraph below, the role for sleep in human heuristic creativity and insightful behavior can further illuminate this issue. Importantly, it has been consistently shown that sleep provides unique conditions for development of human heuristic creativity (Cai et al., 2009; Stickgold et al., 1999; Walker et al., 2002), among which the insightful behavior, as a higher form of human intelligence, is of major importance (Wagner et al., 2004; Yordanova et al., 2008; 2009; 2010; in press). Human insight refers to discovering of regularities that are out of awareness. Notably, it has been demonstrated that as twice as many subjects who had slept after initial learning of a number reduction task (NRT) gained insight into a hidden regularity relative to subjects who had been sleep deprived (Wagner et al., 2004). However, Wagner and co-authors (2004) have not objectively assessed sleep architecture by PSG. Thus, it has remained unclear how and through which mechanisms sleep promotes insight. This question was addressed in a series of later studies using the so-called split night design (Plihal & Born, 1997; 1999). The first study investigating sleep's role for insight by the split night design demonstrated that implicit knowledge acquired at encoding of NRT before sleep was transformed into insight throughout the early night sleep rich in SWS, thus pointing a role for SWS in insight (Yordanova et al., 2008). Further, the same study demonstrated that implicit learning of NRT acquired at encoding (during awakening before the second half of night) was not further transformed but was preserved by the late night sleep rich in REM sleep, thus indicating that REM sleep stabilizes implicit learning (Yordanova et al., 2008). In two later studies, Yordanova and co-authors demonstrated that SWS contributes to gaining insight in a state dependent instead of a trait-dependent manner. First, they revealed a topographic redistribution of slow cortical potentials (SPs) indicating that a spatial reorganization occurred only after early sleep rich in SWS, but not after late sleep, and only for predictable responses on NRT. This SPs reorganization correlated with the amount of SWS (Yordanova et al., 2009). Second, they showed that only after SWS a pattern of brain activation shown as a precondition for insight (increased alpha and beta EEG desynchronization at the right hemisphere and a lack of such at the left) occurred (Yordanova et al., 2010). Finally, these authors were able to extract a specific for SWS EEG rhythm, slow (8-12 Hz) alpha activity at the right hemisphere, which was associated with transformation of implicit knowledge into insight on the NRT, and which was distinctly different form use-dependent (SWA and 12-15 Hz spindle activity) found at the left hemisphere in response to task performance (Yordanova et al., in press). Collectively, these findings strongly indicate that sleep promotes insight, as a higher form of human intelligence, in a state dependent rather than in a trait dependent way.

Finally, a recent study questioned which sleep mechanisms may play a role for abstraction of an implicit probabilistic structure in sequential stimuli using a statistical learning

paradigm, and searched for a predictive relationship between the type of sleep obtained and subsequent performance improvements (abstraction). Participants who consolidated over either a night of sleep or a nap improved significantly more than those who consolidated over an equivalent period of daytime wakefulness. Importantly, PSG revealed a significant correlation between the level of improvement or abstraction and the amount of SWS obtained (Durrant et al., 2011).

### **4.2 Overview of human neuroimaging data**

166 Neuroimaging – Cognitive and Clinical Neuroscience

related to individual intelligence), or whether sleep and human intelligence are related in both ways (Geiger et al., 2011; Fogel & Smith, 2011). It has been previously proposed that sleep serves human intelligence by complex interactions between its unique physiological and mental states, and individual sleep patterns and cognitive traits (Kirov, 2007). One study has addressed this question providing some evidence that stage 2 of non-REM sleep spindle increase after learning is related to elaborate encoding before sleep, whereas individual's general learning ability is well reflected by inter-individual (trait-like) differences in absolute sleep spindle activity (Schabus et al., 2008). However,, the precise mechanisms involved in the complex association between sleep and intelligence remain so far poorly understood (Kirov, 2007). As will be seen in the paragraph below, the role for sleep in human heuristic creativity and insightful behavior can further illuminate this issue. Importantly, it has been consistently shown that sleep provides unique conditions for development of human heuristic creativity (Cai et al., 2009; Stickgold et al., 1999; Walker et al., 2002), among which the insightful behavior, as a higher form of human intelligence, is of major importance (Wagner et al., 2004; Yordanova et al., 2008; 2009; 2010; in press). Human insight refers to discovering of regularities that are out of awareness. Notably, it has been demonstrated that as twice as many subjects who had slept after initial learning of a number reduction task (NRT) gained insight into a hidden regularity relative to subjects who had been sleep deprived (Wagner et al., 2004). However, Wagner and co-authors (2004) have not objectively assessed sleep architecture by PSG. Thus, it has remained unclear how and through which mechanisms sleep promotes insight. This question was addressed in a series of later studies using the so-called split night design (Plihal & Born, 1997; 1999). The first study investigating sleep's role for insight by the split night design demonstrated that implicit knowledge acquired at encoding of NRT before sleep was transformed into insight throughout the early night sleep rich in SWS, thus pointing a role for SWS in insight (Yordanova et al., 2008). Further, the same study demonstrated that implicit learning of NRT acquired at encoding (during awakening before the second half of night) was not further transformed but was preserved by the late night sleep rich in REM sleep, thus indicating that REM sleep stabilizes implicit learning (Yordanova et al., 2008). In two later studies, Yordanova and co-authors demonstrated that SWS contributes to gaining insight in a state dependent instead of a trait-dependent manner. First, they revealed a topographic redistribution of slow cortical potentials (SPs) indicating that a spatial reorganization occurred only after early sleep rich in SWS, but not after late sleep, and only for predictable responses on NRT. This SPs reorganization correlated with the amount of SWS (Yordanova et al., 2009). Second, they showed that only after SWS a pattern of brain activation shown as a precondition for insight (increased alpha and beta EEG desynchronization at the right hemisphere and a lack of such at the left) occurred (Yordanova et al., 2010). Finally, these authors were able to extract a specific for SWS EEG rhythm, slow (8-12 Hz) alpha activity at the right hemisphere, which was associated with transformation of implicit knowledge into insight on the NRT, and which was distinctly different form use-dependent (SWA and 12-15 Hz spindle activity) found at the left hemisphere in response to task performance (Yordanova et al., in press). Collectively, these findings strongly indicate that sleep promotes insight, as a higher form of human intelligence, in a state dependent rather than in

Finally, a recent study questioned which sleep mechanisms may play a role for abstraction of an implicit probabilistic structure in sequential stimuli using a statistical learning

a trait dependent way.

The existing so far neuroimaging data concerning sleep and cognition do not reveal patterns of brain reactivation during sleep. Instead, they reveal altered brain activities associated with impaired cognitive processes in response to sleep deprivation. One of the most important findings is presented in the study of Yoo et al. (2007b). In this study, two groups of subjects were randomly assigned to either a sleep deprivation (SD) or a sleep control (SC) group. All subjects underwent an episodic memory encoding session during fMRI scanning, in which they viewed a series of picture slides and were retested two days later (after two recovery nights of sleep) for a recognition test session (without fMRI). Compared with SC group, SD subjects displayed much worse recognition at retest after two days, though they had two nights recovery sleep. This finding clearly shows that sleep deprivation impairs learning. The fMRI scans at learning/encoding demonstrated a significant impairment of activation in the hippocampal complex in the SD relative to SC group, a region known to be of critical importance for learning of new information (Yoo et al., 2007b). Another study showed that compared with normal sleep, SD produced impairment of spatial attention that correlated with a reduced activation in the posterior cingulate cortex, as measured by fMRI (Mander et al., 2008). Similarly, it is shown that SD produces lapses of attention manifested as delayed behavioral responses to salient stimuli. To identify changes in task-related brain activation associated with lapses after SD, fMRI scans during a visual selective attention task were conducted. It was demonstrated that SD-related lapses in attention corresponded to (1) reduced ability of frontal and parietal control regions to raise activation in response to lapses, (2) dramatically reduced visual sensory cortex activation and (3) reduced thalamic activation during lapses (Chee et al., 2008). Another fMRI study by the same group tested the hypothesis of whether SD impairs short-term memory due to reduced storage capacity or whether it affects processes contributing to appropriate information encoding. Scans were conducted during performing a short-term memory visual task and during presenting varying visual array sizes without engaging memory. Whereas the magnitude of intraparietal sulcus activation and memory capacity after normal sleep were highly correlated, SD elicited a diminished pattern of activation on both tasks, indicating that deficits in both visual processing and visual attention account for loss of short-term memory capacity (Chee & Chuah, 2007). Further, an fMRI study showed that SD abolishes selective attention in association with a decreased BOLD signal found within fronto-parietal cognitive control areas (the left intraparietal sulcus and the left inferior frontal lobe) and parahippocampal place area (PPA) during a selective attention task performance. Additionally, SD resulted in a significant decrement in functional connectivity between the PPA and the two cognitive control fronto-parietal areas (Lim et al., 2010). Interestingly, a single night of SD was shown to produce a strategy shift during risky decision making such that healthy human volunteers moved from defending against losses to seeking increased gains. An fMRI assessment revealed that this change in economic preferences was correlated

The Memory, Cognitive and Psychological Functions of Sleep:

sensitivity of human brain to specific emotions (Gujar et al., 2011a).

**5.1 Overview of human EEG data** 

across different states of consciousness.

**5.2 Overview of human neuroimaging data** 

Update from Electroencephalographic and Neuroimaging Studies 169

Notably, human REM sleep has been so far shown to consolidate declarative (episodic or semantic) memory only when items have emotional salient components, and only when these emotional components have negative valence (Wagner et al., 2001; 2006; Nishida et al., 2009). Recently, REM sleep was shown to be very important for recalibration of the

Data concerning human sleep EEG findings about the role of REM sleep in psychological functioning are still scarce. One study investigated the effect REM sleep portion of a daily nap on episodic memory consolidation. No memorizing effects on emotionally neutral factbased information were found. However, when episodic items were associated with negative emotional components, REM sleep strongly consolidated them. Moreover, the improvement strongly and positively correlated with all, REM sleep latency, amount of REM sleep, and importantly, with the theta (4-8 Hz) REM sleep EEG signature (Nishida et al., 2009). A more recent study tested whether boosting REM sleep EEG signatures could produce overnight improvement of memory consolidation. In this study, REM sleep EEG was potentiated by delivering theta (5 Hz) anodal trans-cranial direct current stimulation during REM sleep periods in the second half of night. The stimulation did enhanced gamma (25-45 Hz) EEG activity during REM sleep, but it did not improve either declarative or procedural memory consolidation at morning recalls. Instead, this increase in gamma EEG during REM sleep resulted in a worsened mood, as assessed by positive and negative affect scale, in the morning after sleep, accompanied by worsening on working memory, as assessed by a word fluency test (Marshall et al., 2011). Collectively, these studies demonstrate that specific REM sleep EEG rhythms are involved in affective behavior. Interestingly, a recent study demonstrated the EEG signatures of dream recall from both REM sleep (theta, 5-7 Hz EEG activity) and stage 2 of non-REM sleep (alpha, 8-12 Hz EEG activity) (Marzano et al., 2011). These findings document different modes of mentality in REM sleep versus non-REM sleep. Furthermore, they suggest that the neurophysiological mechanisms underlying encoding and recall of episodic memories may remain the same

The existing so far human neuroimaging studies about the role for sleep in emotional regulation demonstrate deteriorated patterns of brain activation after total sleep deprivation (SD). By using fMRI, it has been shown that a disconnection between medial prefrontal cortex and amygdala following SD has been associated with improper response to negative emotional stimuli (Yoo et al., 2007a). Further, Sterpenich et al. (2007) showed that a successful recollection of emotional stimuli elicited larger BOLD responses in the hippocampus and various cortical areas, including the medial prefrontal cortex, in a sleep control (SC) group than in a sleep deprived (SD) group. In contrast, the recollection of negative items elicited larger responses in the amygdala and in occipital areas in the SD relative to the SC group (Sterpenich et al., 2007). A later fMRI study examined the effect of a single night SD on consolidation of aversive emotional stimuli and corresponding patterns of brain activation at retest that took place six months after encoding. At retest 6 months later, the recollection of subjects allowed to sleep, compared with SD subjects, was associated with significantly larger fMRI responses in the ventral medial prefrontal cortex (vMPFC) and precuneus, areas involved in memory retrieval, and in the extended amygdala and occipital cortex, areas involved in emotion modulation at encoding. These results

with the magnitude of an SD-driven increase in ventromedial prefrontal activation as well as by an SD-driven decrease in anterior insula activation during decision making (Venkatraman et al., 2011).

Regarding the above described neuroimaging findings together, it can be concluded that sleep is of critical importance for almost all, if not for all types of cognitive processes, including decision making. However, which mechanisms are responsible for substantial deficits of the many cognitive functions seen after sleep deprivation is still elusive.
