sign in sign up
<<
Home , Encoding (memory), Forgetting, Isolation (psychology), Memory, Memory consolidation, Memory development

G. A. Kerkhof and H. P. A. Van Dongen (Eds.) in Research, Vol. 185 ISSN: 0079-6123 Copyright Ó 2010 Elsevier B.V. All rights reserved.

CHAPTER 4

Sleep, and

Matthew P. Walker*

Sleep and Laboratory, Department of , Helen Wills Institute, University of California, Berkeley, CA, USA

Abstract: As critical as waking brain function is to , an extensive literature now indicates that sleep supports equally important, different, yet complementary operations. This review will consider recent and emerging findings implicating sleep, and specific sleep-stage physiologies, in the modulation, regulation and even preparation of cognitive and emotional brain processes. First, evidence for the role of sleep in memory processing will be discussed, principally focusing on declarative memory. Second, at a neural level, several mechanistic models of sleep-dependent plasticity underlying these effects will be reviewed, with a synthesis of these features offered that may explain the ordered structure of sleep, and the orderly evolution of memory stages. Third, accumulating evidence for the role of sleep in associative memory processing will be discussed, suggesting that the long-term goal of sleep may not be the strengthening of individually memory items, but, instead, their abstracted assimilation into a of generalized . Forth, the newly emerging benefit of sleep in regulating emotional brain reactivity will be considered. Finally, and building on this latter topic, a novel hypothesis and framework of sleep-dependent affective brain proces- sing will be proposed, culminating in testable predictions and translational implications for mood disorders.

Keywords: Sleep; emotion; ; memory; ; consolidation; integration; SWS; REM; reactivation

Introduction sleep is a brain phenomenon, and over the past 20 years, an exciting revival has taken place within the Despite the vast amount of this state takes , focusing on the question of why we from our lives, we still lack any consensus function sleep, and specifically targeting the role of sleep in a for sleep. In part, this is perhaps because sleep, like number of cognitive and emotional processes. This its counterpart , may serve not one but chapter aims to provide a synthesis of these recent many functions, for brain and body alike. Centrally, findings in , with the goal of extracting

* Corresponding author. Tel.: (+) 510 642 5292; Fax: (+) 510 642 5293. E-mail: [email protected]

DOI:10.1016/B978-0-444-53702-7.00004-X 49 50 consistent themes across domains of brain function that ‘temporal memory’ (memory for when events that appear to be regulated by sleep. ‘Memory pro- occur) was significantly disrupted by a night of pre- cessing and brain plasticity’ section will explore the training sleep loss. These findings have been revis- role of sleep in memory and brain plasticity, and ited in a more rigorous study by Harrison and also examine competing models of sleep-dependent Horne (2000), again using the temporal memory . ‘, integration and ’ . Significant impairments in retention section will address the role of sleep beyond mem- were evident in a group of subjects deprived of ory consolidation, in processes of association, inte- sleep for 36 h, scoring significantly lower than con- gration and creativity. Finally, ‘Emotional regula- trols, even in a subgroup that received to tion’ section will discuss the more recent and overcome non-specific effects of lower alertness. emerging role for sleep in emotional and affective Furthermore, the sleep-deprived subjects dis- brain regulation. played significantly worse insight into their mem- ory encoding performance, resulting in lower pre- dictive ability of performance. Memory processing and brain plasticity Pioneering work by Drummond and colleagues has examined the neural basis of similar memory When considering the role of sleep in memory impairments using fMRI, investigating the effects processing, it is pertinent to appreciate that mem- of 35 h of total on verbal learning ories evolve (Walker and Stickgold, 2006). (Drummond et al., 2000). In those who were sleep Specifically, pass through discrete stages deprived, regions of the medial in their ‘lifespan’. The conception of a memory (MTL) were significantly less active during learn- begins with the process of encoding, resulting in ing, relative to a control group that had slept, while an initial stored representation of an the actually expressed greater within the brain. However, it is now understood activation. Most interesting, the parietal lobes, that a vast number of post-encoding memory pro- which were not activated in the control group dur- cesses can take place. For memories to persist over ing learning, were significantly active in the depri- the longer time course of minutes to years, an off- vation group. Such findings suggest that inade- line, non-conscious operation of consolidation quate sleep (at least following one night) prior to appears to be necessary, affording memories learning produces bi-directional changes in epi- greater resistance to decay (a process of stabiliza- sodic encoding activity, involving the inability of tion), or even improved recollection (a process of the MTL to engage normally during learning, com- enhancement). Sleep has been implicated in both bined with potential compensation attempts by the encoding and consolidation of memory. prefrontal regions, which in turn may facilitate recruitment of function (Drummond and Brown, 2001). encoding The impact of sleep deprivation on memory for- mation may be especially pronounced for emo- In contrast to the role of sleep after learning pro- tional material. We have investigated the impact moting offline consolidation (discussed below), of sleep deprivation on the encoding of emotion- emerging evidence indicates an important need ally negative, positive and neutral words (Walker for sleep before learning in preparing specific brain and Stickgold, 2006). When combined across all networks for initial encoding of . One types, subjects in the sleep-deprived con- of the earliest studies to report the effects of dition exhibited a striking 40% reduction in the sleep and sleep deprivation on declarative memory ability to form new human memories under condi- encoding was by Morris et al. (1960), indicating tions of sleep deprivation. When these data were 51 separated into the three affective categories (neg- phylogeny. Perhaps the earliest reference to the ative, positive or neutral), the magnitude of encod- beneficial impact of sleep on memory is by the ing impairment differed. In those that had slept, Roman rhetoritician, Quintilian, stating that both positive and negative stimuli were associated with superior retention levels relative to the neu- ...[it] is a curious fact, of which the is tral condition, consonant with the notion that emo- not obvious, that the interval of a single night tion facilitates memory encoding. However, there will greatly increase the strength of the mem- was severe disruption of encoding and hence later ory... Whatever the cause, things which retention for neutral and especially positive emo- could not be recalled on the spot are easily tional memory in the sleep-deprived group. In con- trast, a relative resistance of negative emotional coordinated the next day, and time itself, memory was observed in the deprivation group. which is generally accounted one of the These data suggest that, while the effects of sleep causes of forgetfulness, actually serves to deprivation are directionally consistent across strengthen the memory. memory sub-categories, the most profound impact is on the encoding of positive emotional stimuli, A robust and consistent literature has demon- and to a lesser degree, emotionally neutral stimuli. strated the need for sleep after learning in the In contrast, the encoding of negative memory subsequent consolidation and enhancement of appears to be more resistant to the effects of prior procedural memories; the evidence for which has sleep loss, at least following one night. recently been reviewed elsewhere (Walker and The impact of sleep deprivation on the neural Stickgold, 2006). Early work focusing on the role dynamics associated with declarative memory for sleep in declarative memory processing was encoding has recently been examined using somewhat less consistent, but more recent find- event-related fMRI (Yoo et al., 2007a). In addition ings have now begun to reveal a robust beneficial to performance impairments under condition of effect of sleep on the consolidation of declarative sleep deprivation, and relative to a control group memory – our focus here. that slept, a highly significant and selective deficit Several reports by Born and his colleagues have was identified in bilateral regions of the hippocam- shown offline improvement on a word-pair associ- pus – a structure known to be critical for learning ates task following sleep, attributed to early night new episodic information. When taken together, sleep, rich in SWS (Diekelmann and Born, 2010). this collection of findings indicate the critical need More recently, the same group has demonstrated for sleep before learning in preparing key neural that, in addition to classically defined slow delta structures for efficient next-day learning. Without waves (0.5–4 Hz), the very slow cortical oscillation adequate sleep, hippocampal function becomes (<1 Hz) appears to be important for the consolida- markedly disrupted, resulting in the decreased tion of declarative memories. Following learning of ability for recording new , the extent a word-pair list, a technique called ‘direct current of which appears to be further governed by altera- ’ was used to induce slow oscillation-like tions in prefrontal encoding dynamics. field potentials in the prefrontal cortex (in this case, at 0.75 Hz) during early night SWS (Diekelmann and Born, 2010). Direct current stimulation not Sleep and only increased the amount of slow oscillations dur- ing the simulation period (and for some time after) Using a variety of behavioural , evi- but also enhanced next-day word-pair retention, dence for the role of sleep in memory consolidation suggesting a critical role for SWS has now been reported across a diverse range of in the offline consolidation of episodic facts. 52

Rather than simply testing memory , significant protective benefit was seen in those that Ellenbogen et al. (2006) have since revealed the slept (Fig. 1B). Thus, memories tested after a night extent of sleep’s ability to protect declarative mem- of sleep were significantly more resistant to inter- ories using experimentally induced learning disrup- ference, whereas across a waking day, memories tion. Taking advantage of a classic interference were far more susceptible to this antagonistic learn- technique called the A-B–A-C paradigm, subjects ing challenge. Yet, it was only by using an interfer- first learned unrelated word-paired associates, des- ing challenge, the A-C list, that the true benefit of ignated as list A-B (e.g. leaf-wheel etc.). After sleep sleep’s protection of memory was revealed, a ben- at night, or wakefulness during the day, half of the efit that would not necessarily have been evident in subjects in each group learned a new, interfering list a standard study-test memory paradigm. containing a new associate paired with the first One mechanism proposed to underlie these word, designated as list A-C (e.g. leaf-nail etc.), effects on hippocampal-dependent learning tasks before being tested on the original A-B list (e.g. (and see next section) is the reactivation of mem- leaf-wheel etc.). In the groups that did not experi- ory representations at night. A considerable num- ence the interfering challenge – simply being ber of reports have investigated the firing patterns trained and then tested on list A-B – sleep provided of large networks of individual across the a modest benefit to memory recollection (Fig. 1A). wake- in animals. The signature firing However, when testing the groups that were patterns of these hippocampal and cortical net- exposed to interfering list learning (list A-C) prior works, expressed during waking performance of to recalling the original list (list A-B), a large and spatial tasks and novel experiences, appear to be [(Fig._1)TD$IG] ‘replayed’ during subsequent SWS (and in some studies, also REM) (Dave and Margoliash, 2000; Dave et al., 1998; Ji and Wilson, 2007; Jones and Wilson, 2005; Louie and Wilson, 2001; Poe et al., 2000; Ribeiro et al., 2004; Skaggs and McNaughton, 1996; Wilson and McNaughton, 1994). Homologous evidence has been reported in the using a virtual maze task in combination with positron emission tomography (PET) scanning (Peigneux et al., 2004). Daytime learning was initially associated with hippocampal activity. Then, during post-training sleep, there was a re-emergence of hippocampal activation, specifically during SWS. Most compelling, how- ever, the amount of SWS reactivation in the was proportional to the amount of next-day task improvement, suggesting that this reactivation is associated with offline . Building on the framework that memories, par- ticularly those involving the hippocampus, are Fig. 1. Impact of sleep on the consolidation and stabilization of reactivated at night during sleep, Rasch et al. declarative memory. Percent correct recall for B words from the (2007) have taken advantage of the classical psy- original A-B pair after a 12 h retention interval of either wake or sleep following no interference or interference learning chology effect of cue-dependent recall and trans- y (list A-C). p < 0.10, *p < 0.05, **p < 0.001; error bars indicate lated it into a sleep-dependent consolidation par- s.e.m. Modified from Ellenbogen et al. (2007). adigm. It is well known that memory can be 53 strongly modulated by smell; most of us have asso- and debate. It is perhaps unlikely that multiple ciated the smell of a certain perfume or cologne different memory systems, involving diverse corti- with a particular person, and when we encounter cal and/or subcortical networks, require the same that same perfume again, it often results in the underlying neural mechanisms for their modula- powerful cued recall of memories of that particu- tion. Even if they do, it is not clear that this process lar person. In this study, however, following learn- would rely on just one type of sleep-stage physiol- ing of a task that was paired with ogy. At present, two intriguing models of sleep- the smell of rose, the odour was not re-presented dependent plasticity, relevant to declarative mem- again at retrieval, but instead, during subsequent ory, have been offered to account for the overnight SWS that night – a time when consolidation was facilitation of recall, which build on different presumed to be occurring. Relative to a control aspects of neural activity during sleep: (1) hippo- condition where the odour was not presented campal-neocortical dialogue and (2) synaptic again during SWS, the re-perfusion of the rose homeostasis hypothesis. scent at night resulted in significantly improved recall the following day. Moreover, the re-presen- Hippocampal-neocortical dialogue tation of the odour resulted in greater (re)activa- tion of the hippocampus during SWS. These find- There is considerable agreement that structures ings support the role of SWS in the consolidation within the MLT, most notably the hippocampal of individual declarative memories, and may indi- complex, are crucial for the formation and retrieval cate an active reprocessing of hippocampal-bound of new declarative memories. These structures are information during SWS. believed to guide the reinstatement of recently formed memories by binding together patterns of cortical activation that were present at the time of Models of sleep-dependent memory processing initial learning. A classical model of declarative memory consolidation suggests that information Elucidating the neural mechanisms that control initially requires MTL binding, but over time, and promote sleep-dependent human memory and by way of slow offline processes, is eventually consolidation remains an active topic of research, integrated into neocortical circuits (Fig. 2). [(Fig._2)TD$IG]

Fig. 2. Model of sleep-dependent hippocampal-neocortical memory consolidation. At encoding the hippocampus rapidly integrates information within distributed cortical modules. Successive sleep-dependent reactivation of this hippocampal-cortical network leads to progressive strengthening of cortico-cortical connections, which, over time, allow these memories to become independent of the hippocampus and gradually integrated with pre-existing cortical memories. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this book.) 54

Neocortical structures thus become the eventual encoding activity when sleep has not taken place site for consolidated episodic memories (through deprivation), being associated with a through cross-cortical connections, and as a conse- decreased ability to form new episodic memories. quence, the MTL is not necessary for their Two recent reports have provided further evi- retrieval. Therefore, the classical model of memory dence in support of this sleep-dependent dialogue consolidation holds that neocortical structures and neural transformation of declarative memory. become increasing important for the retention In the first such report, Takashima et al. (2006) and retrieval of successfully consolidated episodic examined the benefit of daytime on episodic memories, while the corresponding contribution of declarative memory consolidation. In addition to a the hippocampus progressively decreases. In addi- long-term evaluation of memory over 3 months, tion to its role in binding distributed cortical mem- there was also a short-term evaluation of memory ory components, the hippocampus plays a critical across the first day, which included an intervening role in reactivating these networks, specifically period (90 min) between training and testing during sleep. This process of reactivation would, of the original studied (remote) stimuli. Inter- over multiple sleep cycles across a night, and/or estingly, the duration of NREM SWS during the multiple occurrences of sleep over many nights, intervening nap correlated positively with later gradually strengthen the initially weak connections performance, yet negatively between neocortical sites, thereby reinforcing with retrieval-related activity in the hippocampus. them (Fig. 2). Eventually, this strengthening would Advancing on these findings, Maquet and collea- allow the original information to be activated in the gues have since demonstrated that one night of cortex, independent of the hippocampus. post-training sleep deprivation, even following Interestingly, these models make two predic- recovery sleep, significantly impairs the normal tions about the impact of sleep on declarative modulation of hippocampal activity associated memory. The first is that declarative memories with recollection (Gais et al., from the day prior should be more resistant to 2007). Furthermore, first-night sleep deprivation interference the next day, due to the increased also prevented an increase in hippocampal connec- cortico-cortical connections formed during over- tivity with the medial prefrontal cortex (mPFC), a night consolidation (Fig. 1). It is precisely this development that was only observed in those who behavioural effect that was reported in the study sleep after learning. by Ellenbogen et al. (2006), showing greater post- No study has yet demonstrated that the neural sleep resistance to interference, using the A-B– signature of learning during the day is subse- A-C paradigm. A second and far less considered quently reactivated and driven by characteristics benefit of this sleep-dependent dialogue is the of SWS at night, and that the extent of these prop- encoding capacity of the hippocampus (Fig. 2). If erties is consequently proportional to the degree the strengthening of cortico-cortical connections of next-day recall and memory reorganization. takes place during sleep, albeit iteratively, then Nevertheless, collectively, they offer an empirical blocking sleep after hippocampal learning should foundation on which to entertain this possibility. negate this offline transfer, preventing the develop- ment of independence from (or ‘refreshing’ of) the Synaptic homeostasis hypothesis hippocampus, and, by doing so, decrease the capac- ity for new hippocampal learning the next day. This In recent years, an orthogonal theory of SWS and second premise appears to accurately explain the learning has emerged, which postulates a role for findings discussed in the ‘Sleep and memory encod- sleep in regulating the synaptic connectivity of ing’ section above (Yoo et al., 2007b), which the brain – principally the (Tononi describe a significant impairment of hippocampal and Cirelli, 2003, 2006). The model by Tononi 55 and Cirelli considers NREM SWS, and specifi- day, affording improved recall. A number of cally the magnitude of slow-wave activity human studies by Huber, Tononi and colleagues (SWA) of SWS, as a brain state that promotes have provided evidence supporting their model. the decrease of synaptic connections, not their For example, it has been shown that learning of a increase. Accordingly, plastic processes, such as motor-skill adaptation task during the day subse- learning and memory occurring during wakeful- quently triggers locally specific increases in cor- ness, result in a net increase in synaptic strength tical SWA at night, the extent of which is propor- in diffuse brain circuits. The role of subsequent tional to both the amount of initial daytime SWS, therefore, and the slow oscillation in par- learning and the degree of next-day improvement ticular, is to selectively downscale or ‘depotenti- (Fig. 3)(Huber et al., 2004). Taken together with ate’ synaptic strength back to baseline levels, pre- complementary cellular evidence demonstrating venting synaptic over-potentiation, which would sleep-dependent reductions in synaptic strength result in saturated brain plasticity. In doing so, (Liu et al., 2010), such findings are considered this re-scaling would leave behind more efficient supportive of a homeostatic function of NREM and refined memory representations the next slow waves in promoting synaptic downscaling. [(Fig._3)TD$IG]

Fig. 3. Slow-wave activity and motor-skill memory. Topographical high-density EEG maps of slow-wave activity (SWA) during NREM sleep following either (A) motor-skill learning or (B) a non-learning control condition, and (C) the subtracted difference between SWA in the learning versus non-learning condition, demonstrating a local homeostatic increase above the learning-related central-parietal brain region. (D) The correlation between the amount of overnight improvement on the task (measured the next day) and the extent of increase in SWA across subjects. Modified from Huber et al., 2004. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this book.) 56

How can the two of neural reactivation learning and the hallmark feature of stage-2 (such as increased fMRI activity during SWS) and NREM – sleep spindles; short (1 s) synchronous neural homeostasis (such as increased slow-wave bursts of activity expressed in the EEG in the 10– EEG activity) be interpreted at a neural synaptic 16 Hz frequency range. For example, following level? It could be argued that both these reported learning of a motor-skill memory task, Nishida changes reflect either an increased neural reactiva- and Walker (2007) have examined post-training tion or an increase in SWA associated with homeo- sleep spindles over the motor cortex, evaluating stasis and synaptic downscaling. Furthermore, the difference in spindle activity in the right, learn- both hypotheses, while distinctly different, mecha- ing hemisphere (since subjects performance the nistically, could offer complementary benefits at task with their left non-dominant hand), relative the network level in terms of signal-to-noise ratio to the left, non-learning hemisphere. Remarkably, (SNR), and may account for overnight memory when sleep-spindle power at electrode sites above improvements. Specifically, homeostatic synaptic the primary motor cortex of the non-learning downscaling could result in the removal of super- hemisphere (left) were subtracted from those in fluous neural connections, resulting in improved the learning hemisphere (right), representing the SNR. However, neural reactivation and strength- within-subject, between-hemisphere difference in ening of experience-dependent circuits, without spindle activity following learning, a strong predic- removing redundant synaptic connects, may tive relationship with the amount of memory equally improve SNR. Therefore, both mechan- improvement emerged. isms, while different, could produce a similar out- Such findings indicate that the enhancement of come-enhanced fidelity of the memory represen- specific memory representations is associated with tation. Presumably the combination of both would electrophysiological events expressed at local, produce perhaps the most optimal and efficient anatomically discrete locations of the brain. memory trace, yet a careful delineation of these Contrasting with the proposed impact of SWS, possibilities remains an important goal for future the mechanistic benefit of sleep spindles may be studies. related to their faster stimulating frequency; a Interestingly, the homeostasis model would range suggested to facilitate long-term potentia- also predict that sleep deprivation, specifically tion (LTP; a foundational principle of synaptic the prevention of SWS, would also negate effec- strengthening in the brain), and not synaptic tive new learning the next day, due to over-poten- . This increase in spindle activity may tiation of synaptic connections. Thus, any region represent a local, endogenous trigger of intrinsic that exhibits SWA, and is involved in represent- , again corresponding topo- ing memory (e.g. hippocampus), would display a graphically to the underlying memory representa- corresponding inability to code further informa- tion (Nishida and Walker, 2007). tion beyond a normal waking duration (16 h in Increases in post-training spindle activity are humans). Such a premise may offer an alternative not limited to tasks. For explanation to the marked hippocampal encod- example, Gais et al. have shown that there is ing deficits reported under conditions of sleep significantly higher sleep-spindle density in sub- loss (Yoo et al., 2007b). jects that underwent a daytime episodic learning session (encoding of word-pair associates) com- A role for sleep spindles? pared to a control group that did not perform the learning session. Moreover, the spindle density Independent of both these models, which consider was associated with the proficiency of memory the role of SWS in memory processing, a number of recalled the next day in the learning group reports have also described an association between (Gais et al., 2002). 57

Reconciling models computational modelling, would offer improved signal-to-noise quality within the system. Such a None of these models needs necessarily to be cooperative mechanism, which appreciates the wrong. Instead, aspects of each may afford com- temporal order of the wake-sleep cycle (acquisi- plementary and synergistically beneficial out- tion, followed by post-processing), and, within comes for memory. Clues to this possibility lie sleep, the ultradian pattern of sleep-stage progres- within the ordered structure of human sleep, with sion across a night (selection and removal, fol- NREM SWS dominating early in the night, and lowed by strengthening) would produce a network stage-2 NREM and REM prevailing later in the of stored information that is not only more effi- night. When placed in this temporal framework, a cient, but, for those representations remaining, progression of events emerges that may be optimal more enhanced. Both these processes would pre- for the neuroplastic modulation of memory repre- dict improved recall of remodelled individual sentations. From a reactivation perspective, the memories from the prior day and further afford predominance of hippocampal-neocortical inter- the synaptic capacity for efficient acquisition of action would take place in the early SWS-rich new ‘information clay’ the next day. phase of the night, leaving cortico-cortical connec- tions on offer for later processing during stage-2 NREM and REM. Similarly, and even in coinci- Association, integration and creativity dence, SWS may downscale cortical (and possibly subcortical) plasticity, and do so in a learning- As critical as consolidation may be – an operation dependent manner, again leaving only those repre- classically concerned with individual memory sentations (or aspects of these representations) items – the association and integration of new which are strongest – including those strengthened experience into pre-existing networks of knowl- by hippocampal-neocortical interplay – for proces- edge is equally, if not more, important. The result- sing during these latter periods of sleep, dominated ing creation of associative webs of information by faster frequency oscillations. offers numerous and powerful advantages. This is analogous to the art of sculpture. Indeed, the end goal of sleep-dependent memory During the day, through experience, substantial processing may not be to enhance individual mem- informational ‘clay’ is acquired on the cortical ped- ories in isolation, but, instead, to integrate them estal; some relevant, some not. Once accumulated, into a common schema, and by doing so, facilitate the next step is to carve out and select the strongest the development of universal concepts; a process and most salient memory representations (statues) which forms the basis of generalized knowledge, for the organisms – a mechanism that SWS, occur- and even creativity. ring first and predominately early in the night, may be ideally suited for. Following such downscaling and/or dynamic selection of memory through Association and integration translocation, the remaining cortical representa- tions – the rough outline of the sculpted form – Perhaps the earliest demonstration that sleep may may finally be strengthened by faster frequency be involved in a form of memory generalization oscillations, including those of sleep spindles (and was by Fenn et al. (2003). Utilizing an artificial potentially PGO-wave burst during REM), more grammar task, subjects were trained and later associated with the potentiation of synaptic con- tested on their ability to transfer phonological cat- nections, not their depotentiation. This final step is egories across different acoustic patterns. The task akin to polishing and improving the detailed fea- required forming new mappings from complex tures of the memory statue, which, in terms of acoustical sounds to pre-existing linguistic 58 categories, which then generalized to new stimuli. to novel situations. Participants initially learned As such, it involved both a declarative process of five ‘premise pairs’ (A > B, B > C, C > D, forming specific memories associated with the D > E, E > F). Unknown to subjects, the pairs learned stimuli, together with a procedural compo- contained an embedded hierarchy nent involving mapping across the of learned (A > B > C > D > E > F). Following an offline sounds that supports generalization to novel sti- delay of 20 min, 12 h across the day or 12 h con- muli. During the initial training session there was taining a night of sleep, knowledge of this hierar- a significant improvement in recognition perfor- chy was tested by examining relational judge- mance on the task. However, when retested after ments for novel ‘inference’ pairs, separated a 12 h waking interval, performance had decayed. either by 1 of associative distance (B > D, Yet, if subjects were retested following a night of C > Epairs)orby2 of associative distance sleep, this ability for memory generalization was (B > E pair). Despite all groups achieving near restored. Supporting this concept, Dumay and identical premise-pair retention after the offline Gaskell (2007) have also demonstrated that sleep, delay (i.e. the building blocks of the hierarchy), a and not equivalent time awake, can integrate striking dissociation was evident in the ability to related but novel phonemes into pre-existing make relational inference judgements. Subjects long-term lexical memory stores overnight. that were tested soon after learning in the In a related study by Gomez et al. (2006), infants 20 min group showed no evidence of inferential were exposed to ‘phrases’ from an artificial lan- ability, performing at chance levels (Fig. 4A). In guage during a learning session – for example contrast, the 12 h groups displayed highly signif- phrases like ‘pel-wadim-jic’ – until the infants icant relational . Most became familiar (as indexed by look responses). remarkable, however, if the 12 h period con- However, these three syllable units had an embed- tained a night of sleep, a near 25% advantage in ded rule, which was that the first and last unit relational memory was seen for the most distantly formed a relationship of non-adjacency; in this connected inferential judgement (the B > Epair; case, pel predicts jic. The infants were then Fig. 4A). Together, these findings demonstrate retested several hours later, yet some infants took that human memory integration takes time to normally scheduled naps, while others were sched- develop, requiring slow, offline associative pro- uled at a time when they would not sleep after cesses. Furthermore, sleep appears to preferen- learning. At later testing, infants again heard the tially facilitate this integration by enhancing recordings, along with novel phrases in which the hierarchical memory binding, biasing the devel- predictive relationship between the first and last opment of the most distant/weak associative links word was new. Infants who did not sleep recog- amongst related yet separate memory items (Fig. nized the phrases they had learned earlier, yet 4B). It is also interesting to note a further advan- those who had slept demonstrated a generalization tage of this sleep-dependent assimilation process. of the predictive relationship to new phrases, sug- When stored as individual premise pairs (top gesting that the intervening process of sleep row, Fig. 4B), the size/number of items (bits) of allowed the reinterpretation of prior experience, information to code is 10 (A-B, B-C, C-D, D-E, and supported the of commonalities – E-F). However, when formed into a hierarchy, that is the ability to detect a general pattern in new the informational load is compressed, reduced information. by nearly 50% to just six bits (A-B-C-D-E-F). Ellenbogen et al. have since tested this sleep- Therefore, a supplementary benefit of sleep- dependent hypothesis of integration by exam- dependent memory association may be the ining human relational memory – the ability improved efficiency of memory storage, in addi- to generalize previously acquired associations tion to a more generalized representation. 59

Thus, the overnight strengthening and consoli- when considering that declarative (non-emotional) dation of individual item memories (reviewed memories decay over the long term. It is then inter- above) may not be the ultimate objective of esting to speculate whether sleep serves to facili- sleep-dependent memory processing, especially tate two complementary objectives for declarative [(Fig._4)TD$IG]

Fig. 4. Sleep-dependent integration of human relational memory. (A) Delayed inference (associative) memory performance (% correct) in a relational memory task following different offline delays. Immediate testing after just a 20 min offline delay demonstrated a lack of any inferential ability resulting in chance performance on both 1 (first order) and 2 and 2 (second order) associative judgements. Following a more extended 12 h delay, across the day (wake group), performance was significantly above chance across both the 1 and 2 inference judgements. However, following an equivalent 12 h offline delay, but containing a night of sleep (sleep group), significantly better performance was expressed on the more distant 2 inference judgement compared with the 1 judgement. (B) A conceptual model of the effects of sleep on memory integration. Immediately after learning, the representation of each premise is constituted as the choice of one item over another (A > B etc.), and these premises are isolated from one another despite having overlapping elements. After a 12 h period with no sleep, the premise representations are partially integrated by their overlapping elements, sufficient to support first-order transitive inferences. However, following a 12 h offline period with sleep, the premise representations are fully interleaved, supporting both first- and second-order transitive inferences. *p < 0.05; error bars indicate s.e.m. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this book.) 60 memory, which span different time courses. The 2002). Similarly, a study of semantic has first may be an initial process of consolidating indi- demonstrated that, in contrast to the situation in vidual item (episodic) memories that are novel, waking, performance following REM sleep awa- which may occur in the relative short term. Over kenings shows a greater priming effect by weakly a longer time course, however, and utilizing these related words than by strong primes, while strong recently consolidated item memories prior to their priming exceeds weak priming in NREM sleep fading, sleep may begin the process of extraction (Stickgold et al., 1999), again indicating the highly (of meaning) and abstraction (building associ- associative properties of the REM sleep brain. ational links with existing information), thereby Even the study of mental activity () from creating more adaptive semantic networks REM sleep indicates that there is not a concrete (Walker and Stickgold, 2010). Ultimately, individ- episodic replay of daytime experiences, but ual item memories would no longer be necessary instead, a much more associative process of seman- for the goal that sleep is trying to achieve, and only tic integration during sleep (Fosse et al., 2003). the conceptual meaning of such experiences would Yet, the most striking experimental evidence of remain. Whether the subsequent loss of item mem- sleep-inspired insight is arguably that reported ories is passive, or whether sleep plays an active by Wagner et al. (2004). Using a mathematical role in this process remains to be examined, but ‘Number Reduction Task’, a process of sleep- this is a testable hypothesis, that is (indi- dependent creative insight was elegantly demon- vidual items) is the price we pay for remembering strated. Subjects analysed and worked through a (general rules). series of 8-digit string problems, using specific addition rules. Following initial training, after var- ious periods of wake or sleep, subjects returned for Creativity an additional series of trials. When retested after a night of sleep, subjects solved the task, using this One potential advantage of testing associative con- ‘standard’ procedure, 16.5% faster. In contrast, nections and building cross-linked systems of subjects who did not sleep prior to retesting aver- knowledge is creativity – the ability to take existing aged less than a 6% improvement. However, hid- pieces of information and combine them in novel den in the construction of the task was a much ways that lead to greater and offer simpler way to solve the problem. On every trial, new advantageous behavioural repertoires. The the last three response digits were the mirror image link between creativity and sleep, especially - of the preceding three. As a result, the second ing, has long been a topic of intense speculation. response digit always provided the answer to the From the dreams of both August Kekule, which led problem, and using such ‘insight’, subjects could to the conception of a simple structure for benzene, stop after producing the second response digit. and Dmitry Mendeleyev, which initiated the crea- Most dramatically, nearly 60% of the subjects tion of the periodic table of elements, to the late- who slept for a night between training and retest- night dreaming of Otto Loewi, which inspired the ing discovered this short cut the following morning. experimental demonstration of neurochemical In contrast, no more than 25% of subjects in any of transmission, even scientific examples of creativity four different control groups who did not sleep had occurring during sleep are not uncommon. this insight. Sleeping after exposure to the problem Quantitative data have further demonstrated therefore more than doubled the likelihood of that solution performance on tests of cognitive solving it (although it is interesting to note that this flexibility using anagram word puzzles is more than insight was not present immediately following 30% better following awakenings from REM sleep sleep, but took over 100 trials on average to compared with NREM awakenings (Walker et al., emerge the next day). 61

In summary, substantial evidence now suggests investigated the effects of sleep disruption on emo- that sleep serves a meta-level role in memory pro- tional reactivity to daytime work events in medical cessing that moves far beyond the consolidation residents. Sleep loss was shown to amplify negative and strengthening of individual memories, and, emotional consequences of disruptive daytime instead, aims to intelligently assimilate and gener- events while blunting the positive benefit associ- alize these details offline. In doing so, sleep may ated with rewarding or goal-enhancing activities. offer the ability to test and build common informa- Although these findings help to characterize the tional schemas of knowledge, providing increas- behavioural irregularities imposed by sleep loss, ingly accurate statistic predictions about the world, evidence for the role of sleep in regulating our and allowing for the discovery of novel, even cre- emotional brain is surprisingly scarce. To date, ative next-day solution insights. only one such study has investigated whether a lack of sleep inappropriately modulates human emotional brain reactivity (Yoo et al., 2007a). Emotional regulation Healthy young participants were allowed to sleep normally prior to a functional MRI (fMRI) scan- Despite substantial research focusing on the inter- ning session, or were sleep deprived for one night action between sleep and cognition, especially (accumulating approximately 35 h of wakeful- memory, the impact of sleep and sleep loss on ness). During scanning, subjects performed an affective and emotional regulation has received affective stimulus-viewing task involving the pre- more limited research . This absence of sentation of picture slides ranging in a gradient investigation is perhaps surprising considering that from emotionally neutral to increasingly negative nearly all psychiatric and neurological mood dis- and aversive. orders express co-occurring abnormalities of sleep, While both groups expressed significant amyg- suggesting an intimate relationship between sleep dala activation in response to increasingly negative and emotion. Nevertheless, a number of recent picture stimuli, those in the sleep-deprived condi- studies evaluating subjective as well as objective tion exhibited a remarkable +60% greater magni- measures of mood and affect, combined with tude of reactivity, relative to the control insights from clinical domains, offer an emerging group. In addition to this increased intensity of understanding for the critical role of sleep in regu- activation, there was also a threefold increase in lating emotional brain function. the extent of amygdala volume recruited in response to the aversive stimuli in the sleep- deprived group. Perhaps most interestingly, rela- Affective reactivity tive to the sleep-control group, there was a signifi- cant loss of functional connectivity identified Together with impairments of attention and alert- between the amygdala and the mPFC in those ness, sleep deprivation is commonly associated who were sleep deprived – a region known to have with increased subjective reports of strong inhibitory projections and hence modulatory and affective volatility (Horne, 1985). Using a impact on the amygdala. In contrast, significantly sleep restriction paradigm (5 h/night), Dinges greater connectivity in the deprivation group was et al. (1997) have reported a progressive increase observed between the amygdala and the auto- in emotional disturbance across a 1-week period nomic-activating centres of the . on the basis of questionnaire mood scales. In addi- Thus, without sleep, an amplified hyper-limbic tion, subjective descriptions in participants’ daily reaction by the human amygdala was observed journals also indicated increasing complaints of in response to negative emotional stimuli. emotional difficulties. Zohar et al. (2005) have Furthermore, this altered magnitude of limbic 62 activity is associated with a loss of functional con- and not across an equivalent time awake. nectivity with the mPFC in the sleep-deprived con- Wagner et al. (2001) have also shown that sleep dition; implying a failure of top-down inhibition by selectively favours the retention of previously the prefrontal lobe. It would therefore appear that learned emotional texts relative to neutral texts, a night of sleep may ‘reset’ the correct affective and that this affective memory benefit is only pres- brain reactivity to next-day emotional challenges ent following late-night sleep (a time period rich in by maintaining functional integrity of this mPFC- stage-2 NREM and REM sleep). amygdala circuit and thus govern appropriate Using a nap paradigm, it has most recently been behavioural repertoires (e.g. optimal social judge- demonstrated that sleep, and specifically REM ments and rational decisions). Intriguingly, a sim- neurophysiology, may underlie this consolidation ilar pattern of anatomical dysfunction has been benefit (Nishida and Walker, 2007). Subjects per- implicated in a number of psychiatric mood disor- formed two study sessions in which they learned ders which express co-occurring sleep abnormali- emotionally negative and neutral picture stimuli: ties, directly raising the issues of whether such fac- one 4 h prior, and one 15 min prior to a recognition tors (sleep loss and clinical mood disorders) are memory test. In one group, participants slept causally related. (90 min nap) after the first study session, while in the other group, participants remained awake. Thus, items from the first (4 h) study sessions tran- Emotional sitioned through different brain states in each group prior to testing, containing sleep in the nap Sleep’s role in declarative memory consolidation, group and no sleep in the no-nap group, yet expe- rather than being absolute, may depend on more rienced identical brain-state conditions following intricate aspects of the information being learned, the second (15 min) study session prior to testing. such as novelty, meaning to extract, and also the No change in memory for emotional (or neutral) affective of the material. Independent of stimuli occurred across the offline delay in the no- the field of sleep and memory, there is a wealth of nap group. However, a significant and selective evidence demonstrating that memory processing is offline enhancement of emotional memory was modulated by emotion. Experiences which evocate observed in the nap group, the extent of which not only encode more strongly but was correlated with the amount of REM sleep appear to persist and even improve over time as and the speed of entry into REM (latency). the delay between learning and testing increases Furthermore, spectral analysis of the EEG demon- (hours/days). strated that the magnitude of right-dominant pre- Although these findings indicate a strong influ- frontal theta power during REM (activity in the ence of emotion on slow, time-dependent consoli- frequency range of 4.0–7.0 Hz) exhibited a signifi- dation processes, based on the coincident neuro- cant and positive relationship with the amount of physiology that REM sleep provides and the emotional memory improvement. neurobiological requirements of emotional mem- These findings go beyond demonstrating that ory processing, work has now begun to test a selec- affective memories are preferentially enhanced tive REM-dependent hypothesis of affective across periods of sleep, and indicate that the extent human memory consolidation. For example, of emotional memory improvement is associated Hu et al. (2006) have compared the consolidation with specific REM sleep characteristics – both of emotionally arousing and non-arousing picture quantity and quality. Corroborating these correla- stimuli following a 12 h period across a day or tions, it has previously been hypothesized that following a night of sleep. A specific emotional REM sleep represents a brain state particularly memory benefit was observed only following sleep amenable to emotional memory consolidation, 63 based on its unique biology. Neurochemically, tends not to be associated with anywhere near levels of limbic and forebrain the same magnitude of autonomic (re)activation (ACh) are markedly elevated during REM, as that elicited at the moment of learning/experi- reportedly quadruple those seen during NREM ence – suggesting that, over time, the affective and double those measured in quite waking. ‘’ previously enveloped around the mem- Considering the known importance of ACh in the ory during encoding has been removed, while the long-term consolidation of emotional learning, this information contained within that experience (the pro- REM state may result in a selective memory) remains. facilitation of affective memories, similar to that For example, neuroimaging studies have shown reported using experimental manipulations of that initial exposure and learning of emotional sti- ACh. Neurophysiologically, theta oscillations have muli is associated with substantially greater activa- been proposed as a carrier frequency allowing dis- tion in the amygdala and hippocampus, relative to parate brain regions that initially encode informa- neutral stimuli (Dolcos et al., 2004, 2005; Kilpatrick tion to selectively interact offline, in a coupled and Cahill, 2003). In one of these studies relationship. By doing so, REM theta may afford (Dolcos et al., 2004), however, when participants the ability to promote the strengthening of spe- were re-exposed to these same stimuli during rec- cific memory representations across distributed ognition testing many months later, a change in the networks. profile of activation occurred (Dolcos et al., 2005). Although the same magnitude of differential activ- ity between emotional and neutral items was A hypothesis of emotional memory: sleep to forget observed in the hippocampus, this was not true in and sleep to remember, respectively the amygdala. Instead, the difference in amygdala (re)activity compared with neutral items had dissi- Beyond the strengthening of emotional memories, pated over time. This would support the that there may be an additional consequence of sleep- the strength of the memory (hippocampal-associ- dependent affective modulation, and one that has ated activity) remains at later recollection, yet the significant implications for mood disorders – that is associated emotional reactivity to these items sleeping to forget. Based on the emerging interac- (amygdala activity) is reduced over time. tion between sleep and emotion, below I outline a The hypothesis predicts that this decoupling model of affective information processing that may preferentially takes place overnight, such that we offer brain-based explanatory insights regarding sleep to forget the emotional tone, yet sleep to the impact of sleep abnormalities, particularly remember the tagged memory of that episode REM, for the initiation and/or maintenance of (Fig. 5). The model further argues that if this pro- mood disturbance. cess is not achieved, the magnitude of visceral While there is abundant evidence to suggest that autonomic ‘charge’ remaining within autobio- emotional experiences persist in our autobiogra- graphical memory networks will persist, resulting phies over time, an equally remarkable but far less in the potential condition of chronic . Based noted change is a reduction in the affective tone on the consistent relationship identified between associated with their recall. The reason that affec- REM and emotional processing, combined with its tive experiences appear to be encoded and consol- unique neurobiology, the hypothesis proposes that idated more preferentially than neutral memories REM sleep provides an optimal biological state for is due to autonomic neurochemical reactions eli- achieving such affective ‘therapy’. Specifically, cited at the time of the experience, creating what increased activity within limbic and paralimbic we commonly term an ‘emotional memory’. structures (including the hippocampus and amyg- However, the later recall of these experiences dala) during REM may first offer the ability for 64 [(Fig._5)TD$IG]

Fig. 5. The sleep to forget and sleep to remember (SFSR) model of emotional memory processing. (a) Neural dynamics. Waking formation of an episodic emotional memory involves the coordinated encoding of hippocampal-bound information within cortical modules, facilitated by the extended limbic system, including the amygdala, and modulated by high concentrations of aminergic neurochemistry. During subsequent REM sleep, these same neural structures are reactivated, the coordination of which is made possible by synchronous theta oscillations throughout these networks, supporting the ability to reprocess previously learned emotional experiences. However, this reactivation occurs in a neurochemical milieu devoid of aminergic modulation, and dominated by cholinergic neurochemistry. As a consequence, emotional memory reprocessing can achieve, on the one hand, a depotentiation of the affective tone initially associated with the event(s) at encoding, while on the other hand, a simultaneous and progressive neocortical consolidation of the information. The latter process of reactivation, resulting in next-day stronger cortico-cortical connections, additionally supports integration into previous acquired autobiographical experiences, further aiding the assimilation of the affective event(s) in the context of past knowledge, the conscious expression of which may contribute to the experience of dreaming. Cross- connectivity between structures before and after sleep is represented by number and thickness of lines. Circles within cortical and hippocampal structures represent information nodes; shade reflects extent of connectivity: strong (filled), moderate (grey) and weak (clear). Fill of limbic system and arrow thickness represents magnitude of co-activation with and influence on the hippocampus. (b) Conceptual outcome. Through multiple iterations of this REM mechanism across the night, and/or across multiple nights, the long- term consequence of such sleep-dependent reprocessing would allow for the strengthening and retention of salient information previously tagged as emotional at the time of learning. However, recall no longer maintains an affective, aminergic charge, allowing for post-sleep recollection with minimal autonomic reactivity (unlike encoding), thereby preventing a state of chronic anxiety. reactivation of previously acquired affective offer large-scale network cooperation at night, experiences. Second, the neurophysiological signa- allowing the integration and, as a consequence, ture of REM involving dominant theta oscillations greater understanding of recently experienced within subcortical as well as cortical nodes may emotional events in the context of pre-existing 65 neocortically stored . Third, aminergic demodulation, that prevents the proces- these interactions during REM critically, and per- sing and separation of emotion from memory. haps most importantly, take place within a brain Indeed, this hypothesis has gained support from that is devoid of aminergic neurochemical concen- recent pharmacological studies in PTSD patients, tration, particularly noradrenergic input from the demonstrating that nocturnal alpha-adrenergic locus coeruleus; the influence of which has been blockade using prazosin (i.e. reducing adrenergic linked to states of high and anxiety disorders. activity during sleep) both decreases trauma- In summary, the described neuroanatomical, dream symptomatology and restores characteris- neurophysiological and neurochemical conditions tics of REM sleep (Raskind et al., 2007; Taylor of REM sleep offer a unique biological theatre in et al., 2008). which to achieve, on the one hand, a balanced This model also makes specific experimental neural potentiation of the informational core of predictions as to the fate of these two components emotional experiences (the memory), yet may also – the memory and the emotion. As partially dem- depotentiate and ultimately ameliorate the auto- onstrated, the first prediction would be that, over nomic arousing charge originally acquired at the time, the veracity of the memory itself would time of learning (the emotion), negating a long- improve, and the extent to which these [negative] term state of anxiety (Fig. 5). emotional experiences are strengthened would be The model also asserts that if the process of proportional to the amount of post-experience divorcing emotion from memory is not achieved REM sleep obtained, as well as how quickly it is across the first night following an emotional event, achieved (REM latency). a repeat attempt would occur on the second night, Secondly, using autonomic physiology mea- since the strength of emotional ‘tag’ associated sures, these same predictions would hold in the with the memory would remain high. Should this inverse direction for the magnitude of emotional process fail a second time, the same events would reactivity induced at the time of recall. Together continue to repeat across ensuing nights, poten- with the neuroimaging studies of emotional mem- tially with an increasing progressive amount of ory recall over time, and the psychological studies REM in response. It is just such a cycle of REM- investigating the role of REM sleep dreaming in sleep dreaming () that represents a mood regulation, a recent fMRI study offers per- diagnostic key feature of post-traumatic stress dis- haps the strongest preliminary support of this order (PTSD) (Lavie, 2001). It may not be coinci- sleep-dependent model of emotional memory pro- dental, therefore, that these patients continue to cessing (Sterpenich et al., 2007). The investigation display hyperarousal reactions to associated demonstrated that subjects who were deprived of trauma cues, indicating that the process of separat- sleep the first night after learning arousing emotion ing the affective tone from the emotional experi- picture slides showed not only reduced recall of the ence has not been accomplished. The reason why information (the sleep to remember component of such a REM mechanism may fail in PTSD remains the hypothesis) but also a lack of reduction in unknown, although the exceptional magnitude of amygdala reactivity when re-exposed to these trauma-induced emotion at the time of learning same emotional picture slide at recognition testing may be so great that the system is incapable of – as compared to a control group that did sleep (the initiating/completing one or both these processes, sleep to forget component of the hypothesis). leaving some patients unable to depotentiate, Thirdly, the model predicts that a pathological integrate and hence ‘overcome’ the experience. increase in REM (as commonly occurs in depres- Alternatively, it may be the hyperarousal status sion) may disproportionately amplify the strength of the brain during REM sleep in these patients of negative memories so much that, despite con- (Harvey et al., 2003; Pole, 2007; Strawn and comitant attempts at ameliorating the associated Geracioti, 2008), potentially lacking sufficient affective tone, it would still create a perceived 66 autobiographical history dominated by negative emotional experiences may be possible, redressing memory excess (which may also facilitate disad- and maintaining the appropriate connectivity and vantageous waking rumination). In contrast, the hence next-day reactivity throughout limbic and selective decrease of REM, as occurs with many associated autonomic systems. anti-depressants, would predict a reduction of such Ultimately, the timeless maternal wisdom of negative memory consolidation and bias, although mothers alike may have long held the answers to it may curtail the degree of affect decoupling that Allan Rechtschaffen’s original question; that is can occur. Long term, the balanced extent of accu- ‘you should sleep on a problem’, and when trou- mulated REM should, therefore, not only correlate bled ‘get to , you’ll feel better in the morning’. with the persistence, in memory, of the emotional experience, but also be associated with a decreased magnitude of autonomic response associated with Acknowledgements recall – all of which are testable objectives for future research. The author wishes to thank Edwin Robertson, If such a hypothesis is correct, there would be , Allison Harvey, Ninad Gujar implications for psychiatric and mood and Els van der Helm for thoughtful insights. disorders. This would require a new appreciation This work was supported in part by grants from for the palliative role of sleep in treatment regimes, the National Institutes of Health (NIA AG31164); and a consideration of whether altering sleep and the University of California, Berkeley. architecture to regulate the balance of of past experience is a useful and viable possibility. References

Conclusions Dave, A. S., & Margoliash, D. (2000). Song replay during sleep and computational rules for sensorimotor vocal learning. – While not fully complete, we will soon have a new Science, 290, 812 816. Dave, A. S., Yu, A. C., & Margoliash, D. (1998). Behavioral taxonomy of sleep-dependent memory processing, state modulation of auditory activity in a vocal motor system. and one that will supersede the polarized all-or- Science, 282, 2250–2254. none views of the past. With such findings, we Diekelmann, S., & Born, J. (2010). The memory function of can come to a revised appreciation of how both sleep. Reviews Neuroscience, 11, 114–126. wake and sleep unite in a symbiotic alliance to Dinges, D. F., Pack, F., Williams, K., Gillen, K. A., Powell, J. W., & Ott, G. E., et al., (1997). Cumulative sleepiness, mood coordinate the encoding, consolidation and inte- disturbance, and psychomotor vigilance performance decre- gration of our memories, the ultimate aim of which ments during a week of sleep restricted to 4–5 h per night. may be to create a generalized catalogue of stored Sleep, 20, 267–277. knowledge that does not rely on the verbose reten- Dolcos, F., LaBar, K. S., & Cabeza, R. (2004). Interaction tion of all previously learned facts. between the amygdala and the medial temporal lobe memory system predicts better memory for emotional events. , Beyond memory and plasticity, a growing num- 42, 855–863. ber of human neuroscience studies, set on a foun- Dolcos, F., LaBar, K. S., & Cabeza, R. (2005). Remembering dation of clinical insights, point to an exciting role one year later: Role of the amygdala and the medial temporal for sleep in regulating affective brain function and lobe memory system in retrieving emotional memories. emotional experience. Based on the remarkable Proceedings of the National Academy of Sciences USA, 102, 2626–2631. neurobiology of sleep, and REM in particular, a Drummond, S. P., & Brown, G. G. (2001). The effects of total unique capability for the overnight modulation of sleep deprivation on cerebral responses to cognitive perfor- affective networks and previously encountered mance. , 25, S68–S73. 67

Drummond, S. P., Brown, G. G., Gillin, J. C., Stricker, J. L., Lavie, P. (2001). Sleep disturbances in the wake of trau- Wong, E. C., & Buxton, R. B. (2000). Altered brain response matic events. New England Journal of Medicine, 345, to verbal learning following sleep deprivation. Nature, 403, 1825–1832. 655–657. Liu, Z. W., Faraguna, U., Cirelli, C., Tononi, G., & Gao, X. B. Dumay, N., & Gaskell, M. G. (2007). Sleep-associated changes (2010). Direct evidence for wake-related increases and sleep- in the of spoken words. Psychological related decreases in synaptic strength in cortex. Science, 18,35–39. Journal of Neuroscience, 30, 8671–8675. Ellenbogen, J. M., Hulbert, J. C., Stickgold, R., Dinges, D. F., & Louie, K., & Wilson, M. A. (2001). Temporally structured Thompson-Schill, S. L. (2006). Interfering with theories of replay of awake hippocampal ensemble activity during rapid sleep and memory: Sleep, declarative memory, and associa- eye movement sleep. Neuron, 29, 145–156. tive interference. Current Biology, 16, 1290–1294. Morris, G. O., Williams, H. L., & Lubin, A. (1960). Ellenbogen, J., Hu, P., Payne, J. D., Titone, D., & Walker, M. P. Misperception and disorientation during sleep. Archives of (2007). Human relational memory requires time and sleep. General , 2, 247–254. The Proceedings of the National Academy of Sciences of the Nishida, M., & Walker, M. P. (2007). Daytime naps, motor United States of America, 104, 7723–7728. memory consolidation and regionally specific sleep spindles. Fenn, K. M., Nusbaum, H. C., & Margoliash, D. (2003). PLoS ONE, 2, e341. Consolidation during sleep of perceptual learning of spoken Peigneux, P., Laureys, S., Fuchs, S., Collette, F., Perrin, F., & . Nature, 425, 614–616. Reggers, J., et al., (2004). Are spatial memories strengthened Fosse, M. J., Fosse, R., Hobson, J. A., & Stickgold, R. J. (2003). in the human hippocampus during slow wave sleep? Neuron, Dreaming and episodic memory: A functional dissociation? 44, 535–545. Journal of , 15,1–9. Poe, G. R., Nitz, D. A., McNaughton, B. L., & Barnes, C. A. Gais, S., Albouy, G., Boly, M., Dang-Vu, T. T., Darsaud, A., & (2000). Experience-dependent phase-reversal of hippocam- Desseilles, M., et al., (2007). Sleep transforms the cerebral pal neuron firing during REM sleep. Brain Research, 855, trace of declarative memories. Proceedings of the National 176–180. Academy of Sciences USA, 104, 18778–18783. Pole, N. (2007). The psychophysiology of posttraumatic stress Gais, S., Molle, M., Helms, K., & Born, J. (2002). Learning- disorder: A meta-analysis. Psychological Bulletin, 133, dependent increases in density. Journal of 725–746. Neuroscience, 22, 6830–6834. Rasch, B., Buchel, C., Gais, S., & Born, J. (2007). Odor cues Gomez, R. L., Bootzin, R. R., & Nadel, L. (2006). Naps promote during slow-wave sleep prompt declarative memory consoli- abstraction in language-learning infants. Psychological dation. Science, 315, 1426–1429. Science, 17, 670–674. Raskind, M. A., Peskind, E. R., Hoff, D. J., Hart, K. L., Harrison, Y., & Horne, J. A. (2000). Sleep loss and temporal Holmes, H. A., & Warren, D., et al., (2007). A parallel memory. Quarterly Journal of , 53, group controlled study of prazosin for trauma 271–279. nightmares and sleep disturbance in combat veterans with Harvey, A. G., Jones, C., & Schmidt, D. A. (2003). Sleep and post-traumatic stress disorder. Biological Psychiatry, 61, posttraumatic stress disorder: A review. Clinical Psychology 928–934. Review, 23, 377–407. Ribeiro, S., Gervasoni, D., Soares, E. S., Zhou, Y., Lin, S. C., & Horne, J. A. (1985). Sleep function, with particular reference to Pantoja, J., et al., (2004). Long-lasting novelty-induced neu- sleep deprivation. Annals of Clinical Research, 17, 199–208. ronal reverberation during slow-wave sleep in multiple fore- Hu, P., Stylos-Allen, M., & Walker, M. P. (2006). Sleep facil- brain areas induction of hippocampal long-term potentiation itates consolidation of emotionally arousing declarative mem- during waking leads to increased extrahippocampal zif-268 ory. Psychological Science, 17, 891–898. expression during ensuing rapid-eye-movement sleep. PLoS Huber, R., Ghilardi, M. F., Massimini, M., & Tononi, G. (2004). Biology, 2, E24. Local sleep and learning. Nature, 430,78–81. Skaggs, W. E., & McNaughton, B. L. (1996). Replay of neuronal Ji, D., & Wilson, M. A. (2007). Coordinated memory replay in firing sequences in rat hippocampus during sleep following the visual cortex and hippocampus during sleep. Nature spatial experience. Science, 271, 1870–1873. Neuroscience, 10, 100–107. Sterpenich, V., Albouy, G., Boly, M., Vandewalle, G., Darsaud, Jones, M. W., & Wilson, M. A. (2005). Theta rhythms coordi- A., & Balteau, E., et al., (2007). Sleep-related hippocampo- nate hippocampal-prefrontal interactions in a spatial memory cortical interplay during emotional memory recollection. task. PLoS Biology, 3, e402. PLoS Biology, 5, e282. Kilpatrick, L., & Cahill, L. (2003). Amygdala modulation of Stickgold, R., Scott, L., Rittenhouse, C., & Hobson, J. A. (1999). parahippocampal and frontal regions during emotionally Sleep-induced changes in associative memory. Journal of influenced memory storage. Neuroimage, 20, 2091–2099. Cognitive Neuroscience, 11, 182–193. 68

Strawn, J. R., & Geracioti Jr., T. D. (2008). Noradrenergic Walker, M. P., & Stickgold, R. (2006). Sleep, memory and dysfunction and the of posttraumatic plasticity. Annual Review of Psychology, 10, 139–166. stress disorder. Depression and Anxiety, 25, 260–271. Walker, M. P., & Stickgold, R. (2010). Overnight alchemy: Takashima, A., Petersson, K. M., Rutters, F., Tendolkar, I., Sleep-dependent memory evolution. Nature Reviews Jensen, O., & Zwarts, M. J., et al., (2006). Declarative mem- Neuroscience, 11, 218. ory consolidation in humans: A prospective functional mag- Walker, M. P., Liston, C., Hobson, J. A., & Stickgold, R. (2002). netic resonance imaging study. Proceedings of the National Cognitive flexibility across the sleep-wake cycle: REM-sleep Academy of Sciences USA, 103, 756–761. enhancement of anagram . Brain Research. Taylor, F. B., Martin, P., Thompson, C., Williams, J., Mellman, Cognitive Brain Research, 14, 317–324. T. A., & Gross, C., et al., (2008). Prazosin effects on objective Wilson, M. A., & McNaughton, B. L. (1994). Reactivation of sleep measures and clinical symptoms in civilian trauma post- hippocampal ensemble memories during sleep. Science, 265, traumatic stress disorder: A placebo-controlled study. 676–679. Biological Psychiatry, 63, 629–632. Yoo,S.S.,Gujar,N.,Hu,P.,Jolesz,F.A.,&Walker,M.P. Tononi, G., & Cirelli, C. (2003). Sleep and synaptic homeostasis: (2007a). The human emotional brain without sleep – A A hypothesis. Brain Research Bulletin, 62, 143–150. prefrontal amygdala disconnect. Current Biology, 17, Tononi, G., & Cirelli, C. (2006). Sleep function and synaptic R877–878. homeostasis. Reviews, 10,49–62. Yoo, S. S., Hu, P. T., Gujar, N., Jolesz, F. A., & Walker, M. P. Wagner, U., Gais, S., & Born, J. (2001). Emotional memory (2007b). A deficit in the ability to form new human memories formation is enhanced across sleep intervals with high without sleep. Nature Neuroscience, 10, 385–392. amounts of . Learning & Zohar, D., Tzischinsky, O., Epstein, R., & Lavie, P. (2005). Memory, 8, 112–119. The effects of sleep loss on medical residents’ emotional Wagner, U., Gais, S., Haider, H., Verleger, R., & Born, J. reactions to work events: A cognitive-energy model. Sleep, (2004). Sleep inspires insight. Nature, 427, 352–355. 28,47–54.