the role of sleep in cognition and emotion · the year in cognitive neuroscience 2009 the role of...

THE YEAR IN COGNITIVE NEUROSCIENCE 2009

The Role of Sleep in Cognition and EmotionMatthew P. Walker

Sleep and Neuroimaging Laboratory, Department of Psychology & Helen WillsNeuroscience Institute, University of California, Berkeley, California

As critical as waking brain function is to cognition, an extensive literature now indi-cates that sleep supports equally important, different yet complementary operations.This review will consider recent and emerging findings implicating sleep and specificsleep-stage physiologies in the modulation, regulation, and even preparation of cog-nitive and emotional brain processes. First, evidence for the role of sleep in memoryprocessing will be discussed, principally focusing on declarative memory. Second, at aneural level several mechanistic models of sleep-dependent plasticity underlying theseeffects will be reviewed, with a synthesis of these features offered that may explain theordered structure of sleep, and the orderly evolution of memory stages. Third, accumu-lating 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 individualmemory items, but, instead, their abstracted assimilation into a schema of generalizedknowledge. Fourth, the newly emerging benefit of sleep in regulating emotional brainreactivity will be considered. Finally, and building on this latter topic, a novel hypoth-esis and framework of sleep-dependent affective brain processing will be proposed,culminating in testable predictions and translational implications for mood disorders.

Key words: sleep; learning; memory; encoding; consolidation; association; integration;plasticity; emotion; affect; non-rapid eye movement (NREM) sleep; rapid eye movement(NREM) sleep; offline; slow wave sleep (SWS), slow-wave activity (SWA), sleep spindles

“If sleep does not serve an absolutely vital function, then it is

the biggest mistake the evolutionary process has ever made.”

Allan RechtschaffenUniversity of Chicago Sleep Laboratory

Smithsonian, November 1978

Introduction

A perplexing question continues to eludescientific judgment: “Why do we sleep?” Inaccepting the utility of evolution, as candidlystated by pioneering sleep researcher AllanRechtschaffen, sleep is likely to support a fun-damental need of the organism. Yet, despite

Address for correspondence: Matthew P. Walker, Sleep andNeuroimaging Laboratory, Department of Psychology & HelenWills Neuroscience Institute, University of California, Berkeley,California 94720-1650, USA. Voice: 510-642-5292; fax: [email protected]

the vast amount of time this state takes fromour lives, we still lack any consensus functionfor sleep. In part, this is perhaps because sleep,like its counterpart wakefulness, may serve notone but many functions, for brain and bodyalike.

Centrally, sleep is a brain phenomenon, andover the past 20 years, an exciting revival hastaken place within the neurosciences, one thatfocuses on the question of why we sleep, andspecifically targeting the role of sleep in a num-ber of cognitive and emotional processes. Thisreview aims to provide a synthesis of these re-cent findings in humans, with the goal of ex-tracting consistent themes across domains ofbrain function that appear to be regulated bysleep. Providing a mechanistic foundation onwhich to consider these findings, the first sec-tion of this chapter briefly summarizes the brainsubstrates of sleep: its neurochemistry, neuro-physiology, and functional anatomy. The next

The Year in Cognitive Neuroscience 2009: Ann. N.Y. Acad. Sci. 1156: 168–197 (2009).doi: 10.1111/j.1749-6632.2009.04416.x C© 2009 New York Academy of Sciences.

168

Walker: Sleep in Cognition and Emotion 169

Sle

ep S

tage REM

AWAKE

NREM-1

NREM-2

NREM-3

NREM-4

Time

12am 1am 2am 3am 4am 5am 6am 7am

R E M

S W S

stage-2 N R E M

Figure 1. The human sleep cycle. Across the night, NREM and REM sleep cycle every90 minutes in an ultradian manner, while the ratio of NREM to REM sleep shifts. During thefirst half of the night, NREM stages 3 and 4 NREM (SWS) dominate, while stage 2 NREMand REM sleep prevail in the latter half of the night. EEG patterns also differ significantlybetween sleep stages, with electrical oscillations such as slow delta waves developing inSWS, K-complexes and sleep spindles occurring during stage 2 NREM, and theta wavesseen during REM.

section explores the role of sleep in memoryand brain plasticity and also examines compet-ing models of sleep-dependent learning. Thethird section addresses the role of sleep beyondmemory consolidation, in processes of associa-tion, integration, and creativity. The final sec-tion discusses the more recent and emergingrole for sleep in emotional and affective brainregulation.

Sleep Neurobiology

The sleep of mammalian species has beenbroadly classified into two distinct types: non-rapid eye movement (NREM) sleep and rapideye movement (REM) sleep, with NREM sleepbeing further divided in primates and cats intofour substages (1–4) corresponding, in that or-der, to increasing depth of sleep (Rechtschaffen& Kales 1968). In humans, NREM and REMsleep alternate or “cycle” across the night inan ultradian pattern every 90 min (Fig. 1).Although this NREM−REM cycle length re-mains largely stable across the night, the ratioof NREM to REM within each 90-min cyclechanges, so that early in the night stages 3 and4 of NREM dominate, while stage 2 NREMand REM sleep prevail in the latter half of thenight. Interestingly, the functional reasons for

this organizing principal (deep NREM early inthe night, stage 2 NREM and REM late inthe night) remains unknown–another perplex-ing mystery of sleep.

As NREM sleep progresses, electroen-cephalographic (EEG) activity begins to slow infrequency. Throughout stage-2 NREM there isthe presence of phasic electrical events includ-ing K-complexes (large electrical sharp wavesin the EEG) and sleep spindles (short syn-chronized 10–16 Hz) EEG oscillations (Steri-ade & Amzica 1998). The deepest stages ofNREM, stages 3 and 4, are often groupedtogether under the term “slow wave sleep”(SWS), reflecting the occurrence of low fre-quency waves (1–4 Hz and <1 Hz), which havethemselves been termed “slow-wave activity”(SWA), representing an expression of underly-ing mass cortical synchrony (Amzica & Steri-ade 1995). During REM sleep, however, EEGwaveforms once again change in their com-position, associated with oscillatory activity inthe theta band range (4–7 Hz), together withhigher frequency synchronous activity in the30–80 Hz (“gamma”) range (Llinas & Ribary1993; Steriade et al. 1996). Periodic bursts ofREM also take place, a defining characteristicof REM sleep, associated with the occurrenceof phasic endogenous waveforms expressedin, among other regions, the pons (P), lateral

170 Annals of the New York Academy of Sciences

geniculate nuclei of the thalamus (G), and theoccipital cortex (O), and as such, have beentermed “PGO waves” (Callaway et al. 1987).

As the brain passes through these sleepstages, it also undergoes dramatic alterationsin neurochemistry. In NREM sleep subcorti-cal cholinergic systems in the brain stem andforebrain become markedly less active (Hobsonet al. 1975; Lydic & Baghdoyan 1988), whilefiring rates of serotonergic raphe neurons andnoradrenergic locus coeruleus neurons are alsoreduced relative to waking levels (Aston-Jones& Bloom 1981; Shima et al. 1986). DuringREM sleep both these aminergic populationsare strongly inhibited, while cholinergic sys-tems become as/more active compared to whatthey are during wake (Kametani & Kawamura1990; Marrosu et al. 1995), resulting in a brainstate largely devoid of aminergic modulationand dominated by acetylcholine.

At a whole-brain systems level, neuroimag-ing techniques have revealed complex anddramatically different patterns of functionalanatomy associated with NREM and REMsleep (for review, see Nofzinger 2005). DuringNREM SWS, rostral brain-stem regions, tha-lamic nuclei, basal ganglia, hypothalamus, pre-frontal cortex, cingulate corticies, and medialregions of the temporal lobe all appear to un-dergo reduced activity. However, during REMsleep significant elevations in activity have beenreported in the pontine tegmentum, thalamicnuclei, occipital cortex, mediobasal prefrontallobes, and associated limbic groups, includingthe amygdala, hippocampus, and anterior cin-gulate cortex. In contrast, the dorso−lateralprefrontal cortex, posterior cingulate, and pari-etal cortex appear least active in REM sleep.

Although this summary only begins to de-scribe the range of neural processes that are af-fected by the brain’s daily transit through sleepstates, it clearly demonstrates that sleep itselfcannot be treated as a homogeneous entity, onewhich may or may not alter cognitive and emo-tional processes. Instead, this constellation ofsleep stages offers a range of distinct neurobio-logical mechanisms that can potentially support

the modulation, regulation, and preparation ofnumerous brain functions.

Memory Processingand Brain Plasticity

In considering the role of sleep in memoryprocessing, it is pertinent one appreciate thatmemories evolve (Walker & Stickgold 2006).Specifically, memories pass through discretestages in their “life span.” The conception ofa memory begins with the process of encoding,resulting in a stored representation of an expe-rience within the brain (Paller & Wagner 2002).However, it is now understood that a vast num-ber of postencoding memory processes can takeplace (Stickgold & Walker 2005a). For mem-ories to persist over the longer time course ofminutes to years, an offline, nonconscious oper-ation of event consolidation appears to be nec-essary, affording memories greater resistanceto decay (a process of stabilization), or even im-proved recollection (a process of enhancement)(Robertson et al. 2004; Walker 2005). Sleep hasbeen implicated in both the encoding and con-solidation of memory.

Sleep and Memory Encoding

One of the earliest human studies to reportthe effects of sleep and sleep deprivation ondeclarative memory encoding was by Morriset al. (1960), indicating that “temporal mem-ory” (memory involving when events occur)was significantly disrupted by a night of pre-training sleep loss. These findings have beenrevisited in a more rigorous study by Harri-son and Horne (2000), again using the tem-poral memory paradigm. Significant impair-ments in retention were evident in a group ofsubjects deprived of sleep for 36 h, the sub-jects scoring significantly lower than controls,even in a subgroup that received caffeine toovercome nonspecific effects of lower alert-ness. Furthermore, the sleep-deprived subjectsdisplayed significantly worse insight into their


0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

POSITIVE NEGATIVE NEUTRAL

**

ALL

STIMULUS TYPES

n.s.Sleep

Sleep Deprived

0.0

0.2

0.4

0.6

0.8

1.0

1.2 *

STIMULUS TYPE

A BE

ncodin

g E

ffic

iency (

d')

Figure 2. Sleep deprivation and encoding of emotional and nonemotional declarative memory. Effectsof 38 h of total sleep deprivation on encoding of human declarative memory (A) when combined across allemotional and nonemotional categories; (B) When separated by emotional (positive and negative valence)and nonemotional (neutral valance) categories. †P < 0.08, ∗P < 0.05, ∗∗P < 0.01, error bars representSEM.

memory-encoding performance, resulting inlower predictive ability of performance.

Pioneering work by Drummond and col-leagues examined the neural basis of sim-ilar memory impairments using functionalMRI (fMRI), investigating the effects of 35 hof total sleep deprivation on verbal learning(Drummond et al. 2000). In those who weresleep deprived, regions of the medial temporallobe were significantly less active during learn-ing, relative to a control group that had slept,while the prefrontal cortex actually expressedgreater activation. Most interesting, the pari-etal lobes, which were not activated in the con-trol group during learning, were significantlyactive in the deprivation group. Such findingssuggest that inadequate sleep prior to learn-ing (at least following one night) produces bidi-rectional changes in episodic encoding activity,involving the inability of the medial temporallobe to engage normally during learning, com-bined with potential compensation attempts byprefrontal regions, which in turn may facilitaterecruitment of parietal lobe function (Drum-mond & Brown 2001).

The impact of sleep deprivation on mem-ory formation may be especially pronounced

for emotional material. We have investigatedthe impact of sleep deprivation on the encod-ing of emotionally negative, positive, and neu-tral words (Walker, unpublished results). Whencombined across all stimulus types, subjects inthe sleep-deprived condition exhibited a strik-ing 40% reduction in the ability to form newhuman memories under conditions of sleepdeprivation (Fig. 2A). However, when thesedata were separated into the three affectivecategories (negative, positive, or neutral), themagnitude of encoding impairment differed(Fig. 2B). In those that had slept, both posi-tive and negative stimuli were associated withsuperior retention levels relative to the neutralcondition, consistent with the notion that emo-tion facilitates memory encoding (Phelps 2004).However, there was severe disruption of en-coding and hence later retention for neutraland especially positive emotional memory inthe sleep-deprived group. In contrast, a relativeresistance of negative emotional memory wasobserved in the deprivation group. These datasuggest that, while the effects of sleep depriva-tion are directionally consistent across memorysubcategories, the most profound impact is onthe encoding of positive emotional stimuli, and


to a lesser degree, emotionally neutral stimuli.In contrast, the encoding of negative memoryappears to be more resistant to the effects ofprior sleep loss, at least following one night.

Intriguingly, these data may offer novel in-sights into affective mood disorders that expressco-occurring sleep abnormalities (Benca et al.1992; Buysse 2004). Indeed, if one comparesthe profiles of memory encoding in Fig. 2B,it is clear that those who slept encoded andretained a balanced mix of both positive andnegative memories. In contrast, those who didnot sleep displayed a skewed relative distribu-tion of encoding, resulting in an overridingdominance of negative memories, combinedwith a retention deficit of positive and neutralmemories. This selective alteration in mem-ory encoding may provide an experimentalexplanation for the higher incidence of depres-sion in populations that suffer sleep disruption(Shaffery et al. 2003; Buysse 2004), which, dueto these specific deficits, may impose a neg-ative remembering bias, despite the fact thatthese subjects experienced equally positive- andnegative-reinforcing event histories.

The impact of sleep deprivation on the neu-ral dynamics associated with declarative mem-ory encoding has recently been examined usingevent-related fMRI (Yoo et al. 2007a). In addi-tion to performance impairments under con-dition of sleep deprivation, and relative to acontrol group that slept, a highly significant andselective deficit was identified in bilateral re-gions of the hippocampus—a structure knownto be critical for learning new episodic infor-mation (Eichenbaum 2004) (Fig. 3A). Whilethese findings indicated that, at a group level,sleep deprivation markedly impairs hippocam-pal memory function, when examined withineach group separately, the success of encod-ing, from low to high, was further associatedwith activity in different regions of the pre-frontal lobe. In those that slept prior to learn-ing, the right dorsal/middle lateral prefrontalcortex showed a strong positive relationshipwith the proficiency of memory encoding. Incontrast, a region in the right inferior frontal

gyrus (IFG) displayed a significant positive, po-tentially compensatory, relationship with mem-ory performance in those who were sleep de-prived (Figs. 3B & C).

Taken together, this collection of findingsindicate the critical need for sleep-before-learning in the preparation of key neuralstructures for efficient next-day learning. With-out adequate sleep, hippocampal functionbecomes markedly disrupted, resulting in a de-creased ability for recording new experiences,the extent of which appears to be further gov-erned by alterations in prefrontal encodingdynamics.

Sleep and Memory Consolidation

Using a variety of behavioral paradigms, ev-idence for the role of sleep in memory consoli-dation has now been reported across a diverserange of phylogeny. Perhaps the earliest refer-ence to the beneficial impact of sleep on mem-ory is by the Roman rhetorician Quintilian,who stated:

[it] is a curious fact, of which the reason is not obvious, that

the interval of a single night will greatly increase the strength

of the memory. . . . Whatever the cause, things which could

not be recalled on the spot are easily coordinated the next day,

and time itself, which is generally accounted one of the causes

of forgetfulness, actually serves to strengthen the memory.

(Hammond 2004).

In the early eighteenth and twentieth cen-turies respectively, David Hartley (Hartley1801) and Jenkins and Dallenback (Jenkins &Dallenbach 1924) indicated that the strength ofa memory may be better preserved by periodsof sleep than it is by equivalent periods of timeawake. Following the discovery of discrete sleepstages (Aserinsky & Kleitman 1953), researchinvestigating the influence of sleep on memoryhas become gradually more complex at botha behavioral and mechanistic level. A robustand consistent literature has demonstrated theneed for sleep after learning in the subsequentconsolidation and enhancement of proceduralmemories; the evidence for which has recentlybeen reviewed elsewhere (Walker & Stickgold


Figure 3. Neural basis of sleep-deprivation-induced encoding deficits. (A) Regions of de-creased encoding activation in the sleep deprivation group relative to the sleep control groupin bilateral posterior hippocampus, together with a histogram of parameter estimates (effectsize) of averaged hippocampal activity in each group. Effects are significant at P < 0.001;>5 contiguous voxels. (B) correlation analysis with memory performance showing regionsof significant association between encoding-related activation and memory performance (d’)across subjects in the sleep control group (peak – right Middle/dorso-lateral prefrontal cortex),and (C) in the sleep deprivation group (peak – right Inferior frontal gyrus). Modified from Yooet al. 2007a.

2006). Early work focusing on the role for sleepin declarative memory processing was some-what less consistent, but more recent findingshave now begun to reveal a robust beneficial

effect of sleep on the consolidation of declar-ative memory—our focus here (Smith 2001;Ellenbogen et al. 2006b; Walker & Stickgold2006; Marshall & Born 2007).


Several reports by Born and his colleaguesshowed offline improvement on a word-pairassociation task following sleep, an improve-ment attributed to early night sleep, rich inSWS (Plihal & Born 1997, 1999; Gais & Born2004). More recently, the same group demon-strated that, in addition to classically definedslow delta waves (0.5–4 Hz), the very slow corti-cal oscillation (<1 Hz) appears to be importantfor the consolidation of declarative memories.Following subject learning of a word-pair list,a technique called “direct current stimulation”was used to induce slow oscillation-like fieldpotentials in the prefrontal cortex (in this case,at 0.75 Hz) during early night SWS (Marshallet al. 2006). Direct current stimulation not onlyincreased the amount of slow oscillations dur-ing the simulation period (and for some timeafter), but also enhanced next-day word-pairretention, suggesting a critical role for SWSneurophysiology in the offline consolidation ofepisodic facts.

Rather than simply testing memory recall,Ellenbogen and colleagues have since revealedthe extent of sleep’s ability to protect declar-ative memories using experimentally inducedlearning disruption (Ellenbogen et al. 2006a).Taking advantage of a classic interference tech-nique called the A-B–A-C paradigm, subjectsfirst learned unrelated word-paired associates,designated as list A-B (e.g., leaf-wheel, etc.). Af-ter sleep at night, or wakefulness during theday, half of the subjects in each group learned anew, interfering list containing a new associatepaired with the first word, designated as list A-C(e.g., leaf-nail, etc.), before being tested on theoriginal A-B list (e.g., leaf-wheel, etc.). In thegroups that did not experience the interferingchallenge—that is, those who were simply be-ing trained and then tested on list A-B—sleepprovided a modest benefit to memory recollec-tion (Fig. 4A). However, when testing the groupsthat were exposed to interfering list learning (listA-C) prior to recalling the original list (list A-B), a large and significant protective benefit wasseen in those that slept (Fig. 4B). Thus, mem-ories tested after a night of sleep were signifi-

100

80

60

40

20

0

**

*

Wake Sleep Wake-I Sleep-I

(no interference) (interference)

Condition

Pe

rce

nt

Re

ca

ll

Figure 4. Impact of sleep on the consolidationand stabilization of declarative memory. Percent cor-rect recall for B words from the original A-B pairafter a 12-h retention interval of either wake orsleep following no interference or interference learn-ing (list A-C). †P < 0.10, ∗P < 0.05, ∗∗P < 0.001;error bars indicate SEM. Modified from Ellenbogen2006a.

cantly more resistant to interference, whereas,across a waking day, memories were far moresusceptible to this antagonistic learning chal-lenge. Yet it was only by using an interferingchallenge, that of the A-C list, that the truebenefit of sleep’s protection of memory was re-vealed, a benefit that would not necessarily havebeen evident in a standard study−test memoryparadigm.

One mechanism proposed as underlyingthese effects on hippocampal-dependent learn-ing tasks (see next section, also) is the reactiva-tion of memory representations at night. A con-siderable number of reports have investigatedthe firing patterns of large networks of indi-vidual neurons across the wake−sleep cycle inanimals. The signature firing patterns of thesehippocampal and cortical networks, expressedduring waking performance of spatial tasks and


novel experiences, appear to be “replayed” dur-ing subsequent SWS (and in some studies, alsoREM) (Wilson & McNaughton 1994; Skaggs& McNaughton 1996; Dave et al. 1998; Dave& Margoliash 2000; Poe et al. 2000; Louie &Wilson 2001; Ribeiro et al. 2004; Jones & Wil-son 2005; Ji & Wilson 2007). Homologous ev-idence has been reported in the human brainusing a virtual maze task in combination withpositron emission tomography (PET) scanning(Peigneux et al. 2004). Daytime learning wasinitially associated with hippocampal activity.Then, during posttraining sleep, there was areemergence of hippocampal activation, specif-ically during SWS. Most compelling, however,was that the amount of SWS reactivation in thehippocampus was proportional to the amountof next-day task improvement, suggesting thatthis reactivation is associated with offline mem-ory improvement.

Building on the framework that memories,particularly those involving the hippocampus,are reactivated at night during sleep, Raschet al. have taken advantage of the classicalpsychology effect of cue-dependent recall, andtranslated it into a sleep-dependent consolida-tion paradigm (Rasch et al. 2007). It is wellknown that memory can be strongly modu-lated by smell (Cann & Ross 1989); most of ushave associated the smell of a certain perfumeor cologne with a particular person, and whenwe encounter that same perfume again, it oftenresults in the powerful cued recall of memoriesof that particular person. In this study, however,following learning of a spatial memory task thatwas paired with the smell of rose, the odor wasnot re-presented at retrieval, but instead dur-ing subsequent SWS that night—a time whenconsolidation was presumed to be occurring.Relative to a control condition where the odorwas not presented again during SWS, the re-perfusion of the rose scent at night resulted insignificantly improved recall the following day.Moreover, the re-presentation of the odor re-sulted in greater (re)activation of the hippocam-pus during SWS. These findings support therole of SWS in the consolidation of individ-

ual declarative memories, and may indicate anactive reprocessing of hippocampal-bound in-formation during SWS.

Models of Sleep-DependentMemory Processing

Elucidating the neural mechanisms thatcontrol and promote sleep-dependent humanmemory consolidation remains an active topicof research, and debate (Miller 2007). It is per-haps unlikely that multiple different memorysystems, involving diverse cortical and/or sub-cortical networks, require the same underlyingneural mechanisms for their modulation. Evenif they do, it is not clear that this process wouldrely on just one type of sleep-stage physiology(Giuditta et al. 1995). At present, two intriguingmodels of sleep-dependent plasticity, relevantto declarative memory, have been offered toaccount for the overnight facilitation of recall,which build on different aspects of neural activ-ity during sleep: (1) hippocampal−neocorticaldialogue, (2) synaptic homeostasis hypothesis.

Hippocampal−Neocortical Dialogue

There is considerable agreement that struc-tures within the medial temporal lobe (MTL),most notably the hippocampal complex, arecrucial for the formation and retrieval of newdeclarative memories. These structures are be-lieved to guide the reinstatement of recentlyformed memories by binding together pat-terns of cortical activation that were presentat the time of initial learning. A classical modelof declarative memory consolidation suggeststhat information initially requires MTL bind-ing, but over time, and by way of slow of-fline processes, it is eventually integrated intoneocortical circuits (Fig. 5). Neocortical struc-tures thus become the eventual storage site forconsolidated episodic memories through cross-cortical connections, and, as a consequence,the MTL is not necessary for these memo-ries’ retrieval. Therefore, the classical modelof memory consolidation holds that neocorti-cal structures become increasing important for


Figure 5. Model of sleep-dependent hippocampal-neocortical memory consolidation. Atencoding the hippocampus rapidly integrates information within distributed cortical modules.Successive sleep-dependent reactivation of this hippocampal–cortical network leads to pro-gressive strengthening of cortico-cortical connections, which over time, allow these memoriesto become independent of the hippocampus and gradually integrated with preexisting corticalmemories. Modified from Frankland & Bontempi 2005.

the retention and retrieval of successfully con-solidated episodic memories, while the corre-sponding contribution of the hippocampus pro-gressively decreases (Squire 1992; McClellandet al. 1995; Squire & Zola 1996; Squire 2004).It should be noted, however, that controversyremains about the role of these MTL structuresin the retrieval of declarative memories after thepassage of time. This has led to the emergenceof alternate consolidation models, most notablyNadel and Moscovitch’s “multiple trace the-ory” (Nadel & Moscovitch 1997; Moscovitch& Nadel 1998), which posits that hippocampalinvolvement is always critical for the retrievalof episodic (but not semantic) memories, andthat these memories remain permanently de-pendent on hippocampal−neocortical connec-tions (for discussion beyond the scope of thisreview see Frankl & Bontempi 2005).

In addition to its role in binding distributedcortical memory components, Marr and, also,McClelland et al. suggested that the hippocam-pus plays a critical role in reactivating thesenetworks, specifically during sleep (Marr 1970;McClelland et al. 1995). This process of re-activation, over multiple sleep cycles across anight and/or multiple occurrences of sleep overmany nights, would gradually strengthen theinitially weak connections between neocortical

sites, thereby reinforcing them (Fig. 5). Eventu-ally, this strengthening would allow the originalinformation to be activated in the cortex, in-dependent of the hippocampus. Buzsaki (1996)has since advanced on these ideas, proposing amodel of consolidation that involves two stagesor states of hippocampal activity, the first in-volving a mode of “recording” during wake,which shifts to a second stage, involving “play-back” mode during NREM SWS, specificallyduring bursts of neural activity called “sharp-waves.”

Interestingly, these models make two pre-dictions about the impact of sleep on declar-ative memory. The first is that declarativememories from the day prior should be moreresistant to interference the next day, dueto the increased cortico−cortical connectionsformed during overnight consolidation (Fig. 5).It is precisely this behavioral effect that wasreported in the study by Ellenbogen et al.(2006a) showing greater post-sleep resistanceto interference, using the A-B–A-C paradigm.A second and far less considered benefit ofthis sleep-dependent dialogue is the encod-ing capacity of the hippocampus (Fig. 5). Ifthe strengthening of cortico−cortical connec-tions takes place during sleep, albeit itera-tively, then blocking sleep after hippocampal


Figure 6. Enhancement of hippocampal declarative memory by daytime naps. (A) Correlation of recog-nition memory for recent and remote items related to individual slow-wave sleep durations. (B) Correlationbetween recognition memory activity and longer slow-wave sleep duration in the left hippocampus (Hi).Modified from Takashima et al. 2006.

learning should negate this offline transfer, pre-venting the development of independence from(or “refreshing” of) the hippocampus, and bydoing so, decrease the capacity for new hip-pocampal learning the next day. This sec-ond premise appears to accurately explain thefindings discussed in the section above on mem-ory encoding (Yoo et al. 2007b), which de-scribe a significant impairment of hippocam-pal encoding activity when sleep has not takenplace (through deprivation) being associatedwith a decreased ability to form new episodicmemories.

Two recent reports have provided further ev-idence in support of this sleep-dependent dia-logue and neural transformation of declarativememory. In the first such report, Takashimaand colleagues examined the benefit of day-time naps on episodic declarative memory con-solidation (Takashima et al. 2006). In addi-tion to a long-term evaluation of memory over3 months, there was also a short-term eval-uation of memory across the first day, whichincluded an intervening nap period (90 min)between training and testing of the originalstudied (“remote”) stimuli. Interestingly, theduration of NREM SWS during the inter-vening nap correlated positively with later-recognition memory performance (Fig. 6A),yet negatively with retrieval-related activity inthe hippocampus (Fig. 6B). Furthermore, with

increasing time following learning, there wasprogressively greater recall activity in medialprefrontal regions, and a continued dissipa-tion of retrieval-related activity within the hip-pocampus. Advancing on these findings, Ma-quet and colleagues have since demonstratedthat one night of posttraining sleep depriva-tion, even following recovery sleep, significantlyimpairs the normal modulation of hippocam-pal activity associated with episodic memoryrecollection (Gais et al. 2007). Furthermore,first-night sleep deprivation also prevented anincrease in hippocampal connectivity with themedial prefrontal cortex, a development thatwas only observed in those that slept afterlearning.

While no one study has yet demonstratedthat the neural signature of learning during theday is subsequently reactivated and driven bycharacteristics of SWS at night, and that theextent of these properties are consequently pro-portional to the degree of next-day recall andmemory reorganization, collectively, they offeran empirical foundation on which to entertainthis possibility.

Synaptic Homeostasis Hypothesis

In recent years an orthogonal theory of SWSand learning has emerged, one which postu-lates a role for sleep in regulating the synap-tic connectivity of the brain—principally the


Figure 7. Slow-wave activity and motor-skill memory. Topographical high-density EEGmaps of slow-wave activity (SWA) during NREM sleep following either (A) motor-skill learningor (B) a nonlearning control condition, and (C) the subtracted difference between SWA inthe learning versus nonlearning condition, demonstrating a local homeostatic increase abovethe learning-related central-parietal brain region. (D) the correlation between the amount ofover-night improvement on the task (measured the next day) and the extent of increase inSWA across subjects. Modified from Huber et al. 2004.

neocortex (Tononi & Cirelli 2003, 2006). Theirmodel considers NREM SWS, and specificallythe magnitude of slow-wave activity (SWA) ofSWS, as a brain-state that promotes the de-crease of synaptic connections, not an increase.Accordingly, plastic processes, such as learningand memory occurring during wakefulness, re-sult in a net increase in synaptic strength indiffuse brain circuits. The role of SWS, there-fore, and the slow oscillation in particular, is toselectively downscale or “depotentiate” synap-tic strength back to baseline levels, preventingsynaptic overpotentiation, which would resultin saturated brain plasticity. In doing so, thisrescaling would leave behind more efficient andrefined memory representations the next day,affording improved recall.

A number of human studies by Huber,Tononi, and colleagues have provided evi-dence supporting their model. For example, ithas been shown that learning of a motor-skilladaptation task during the day subsequentlytriggers locally specific increases in corticalSWA at night, the extent of which is propor-tional to both the amount of initial daytimelearning and the degree of next-day improve-ment (Fig. 7) (Huber et al. 2004). Further-more, experimentally impairing the amount ofexperience-dependent activity during the day(arm immobilization) produced the oppositeeffect—reduced amounts of SWA activity inassociated cortical regions (Huber et al. 2006).These findings substantiate the concept of lo-cal sleep-dependent neural pruning by SWS,


the goal of which may be to regulate neural ar-chitecture at a highly specific anatomical level,mapping onto corresponding locations of thememory representation.

How can the two concepts of neural reacti-vation (such as increased fMRI activity dur-ing SWS; e.g., Rasch et al. 2007) and neu-ral homeostasis (such as increased slow waveEEG activity; e.g., Huber et al. 2004) be inter-preted at a neural synaptic level? It could beargued that both these reported changes reflecteither an increased neural reactivation, or anincrease in SWA associated with homeostasisand synaptic downscaling. Furthermore, bothhypotheses, while distinctly different in a mech-anistic sense, could offer complementary ben-efits at the network level in terms of signal-to-noise ratio (SNR) and may account forovernight memory improvements (Robertson,unpublished). Specifically, homeostatic synap-tic downscaling could result in the removalof superfluous neural connections, resulting inimproved SNR. However, neural reactivationand strengthening of experience-dependentcircuits, done without removing redundantsynaptic connects, may equally improve SNR.Therefore, both mechanisms, while different,could produce a similar outcome: enhanced fi-delity of the memory representation. Presum-ably the combination of both would produceperhaps the most optimal and efficient mem-ory trace, yet a careful delineation of these pos-sibilities remains an important goal for futurestudies.

Interestingly, the homeostasis model wouldalso predict that sleep deprivation, specificallythe prevention of SWS, would also negate effec-tive next-day new learning, due to overpotenti-ation of synaptic connections. Thus, any regionthat exhibits SWA and is involved in represent-ing memory (e.g., hippocampus), would displaya corresponding inability to code further in-formation beyond a normal waking duration(∼16 hr in humans). Such a premise may of-fer an alternative explanation to the markedhippocampal encoding deficits reported underconditions of sleep loss (Yoo et al. 2007b).

A Role for Sleep Spindles?

Independent of both these models, whichconsider the role of SWS in memory pro-cessing, a number of reports have also de-scribed an association between learning andthe hallmark feature of stage-2 NREM: sleepspindles; short (∼1 s) synchronous bursts ofactivity expressed in the EEG in the 10–16 Hz frequency range (Sejnowski & Destexhe2000; Steriade 2001; Smith et al. 2004). Forexample, following learning of a motor-skillmemory task, Nishida and Walker (2007) ex-amined posttraining sleep spindles over themotor cortex, evaluating the difference in spin-dle activity in the learning hemisphere (sincesubjects perform the task with their left, non-dominant, hand), relative to the nonlearninghemisphere (Fig. 8A). Remarkably, when sleepspindle power at electrode sites above the pri-mary motor cortex of the nonlearning hemi-sphere (left) were subtracted from those in thelearning hemisphere (right), representing thewithin-subject, between-hemisphere differencein spindle activity following learning, a strongpredictive relationship to the amount of mem-ory improvement emerged (Figs. 8B & C). Simi-larly, Fogel et al. (2001) reported increased spin-dle density after intensive training on a pursuitmotor skill task, and Fogel and Smith (2006)reported increased spindle density after com-bined training on several simple proceduralmotor tasks.

Such findings indicate that the enhancementof specific memory representations is associ-ated with electrophysiological events expressedat local, anatomically discrete locations of thebrain. Contrasting with the proposed impactof SWS, the mechanistic benefit of sleep spin-dles may be related to their faster stimulatingfrequency; a range suggested to facilitate long-term potentiation (LTP; a foundational prin-cipal of synaptic strengthening in the brain)(Sejnowski & Destexhe 2000; Steriade 2001;Smith et al. 2004), and not synaptic depres-sion. This increase in spindle activity may rep-resent a local, endogenous trigger of intrinsic


Figure 8. Sleep spindles and motor-skill memory plasticity. (A) Sleep-EEG array (blue discs) superimposedon the known overnight plastic reorganization of motor memory, including the right motor cortex (red). (B)Difference in sleep spindle activity (power) following task training in the Learning (relative to Non-learning)hemisphere, which (C) accurately predicts the amount of postsleep memory improvement across subjects.Modified from Nishida & Walker 2007.

synaptic plasticity, again corresponding topo-graphically to the underlying memory repre-sentation (Nishida & Walker 2007).

Increases in posttraining spindle activity arenot limited to procedural memory tasks. Forexample, Gais et al. have shown that thereis significantly higher sleep spindle density insubjects that underwent a daytime episodiclearning session (encoding of word-pair as-sociates) compared to a control group thatdid not perform the learning session. More-over, the spindle density was associated withthe proficiency of memory recalled the nextday in the learning group (Gais et al. 2002).These findings mirror previous observations byMeier-Koll et al. (1999), who reported a sim-ilar increase in spindles following learning ofa hippocampally dependent maze task, and byClemens et al. (2005), who have since identi-fied a correlation between spindle density andovernight verbal memory retention (althoughnot memory for faces). Intriguingly, the studyby Huber et al. (2004), which implicates slowoscillatory activity associated with offline mem-ory improvement, also describes similar, albeitnear-significant, associations with activity in thesleep spindle frequency range. There may bea combinatory role for spindles in regulatingplasticity, together with SWS.

Continued evidence suggests that sleep spin-dles can be separated into two subtypes: “slow”

(11–13 Hz), which have a more anterior dis-tribution, and “fast” (13–15 Hz), which havea more posterior localization (Werth et al.1997; Zeitlhofer et al. 1997). With this di-vision has also come an interest in under-standing functional differences between eachof these spindles subtypes. The relevance ofthis separation from a memory consolidationperspective is highlighted by a recent neu-roimaging study demonstrating that fast spin-dles are associated with, among other re-gions, significantly greater activation within thehippocampal complex (Schabus et al. 2007).Investigating the role of hippocampal- andextra−hippocampal-dependent memory con-solidation in relation to spindle subtypes will bean important challenge for future research.

Reconciling Models

None of these models necessarily is wrong.Instead, aspects of each may afford com-plementary and synergistically beneficial out-comes for memory. Clues to this possibility liewithin the ordered structure of human sleep(Fig. 1), with NREM SWS dominating earlyin the night and stage-2 NREM and REMprevailing later in the night. When placedin this temporal framework a progression ofevents emerges that may be optimal for theneuroplastic modulation of memory represen-tations. From a reactivation perspective, the


predominance of hippocampal−neocorticalinteraction would take place in the early SWS-rich phase of the night, leaving cortico−corticalconnections on offer for later processing dur-ing stage-2 NREM and REM. Similarly, andeven in coincidence, SWS may downscale cor-tical (and possibly subcortical) plasticity, andit may do so in a learning-dependent man-ner, again leaving only those representations(or aspects of these representations) whichare strongest—including those strengthened byhippocampal−neocortical interplay—for pro-cessing during these latter periods of sleep,dominated by faster frequency oscillations.

This concept is analogous to the art ofsculpture. During the day, through experi-ence, substantial informational “clay” is ac-quired on the cortical pedestal; some of itrelevant, some not. Once accumulated, thebrain’s next step is carving out and select-ing the strongest and most salient memoryrepresentations (“statues”)—a mechanism thatSWS, occurring first and predominately earlyin the night, may be ideally suited for. Follow-ing such downscaling and/or dynamic selectionof memory through translocation, the remain-ing cortical representations—the rough outlineof the sculpted form—may finally be strength-ened by faster frequency oscillations, includingthose of sleep spindles (and potentially PGO-waveburst during REM; (Datta 2000; Dattaet al. 2008), more associated with the potentia-tion of synaptic connections, not their depoten-tiation. This final step is akin to polishing andimproving the detailed features of the mem-ory statue, which, in terms of computationalmodeling, would offer improved SNR qualitywithin the system. Such a cooperative mech-anism, which appreciates the temporal orderof the wake−sleep cycle (acquisition, followedby postprocessing), and, within sleep, the ultra-dian pattern of sleep-stage progression acrossa night (selection and removal, followed bystrengthening), would produce a network ofstored information that is not only more effi-cient, but for those representations remaining,more enhanced. Both these processes would

predict improved recall of remodeled individ-ual memories from the prior day and furtherafford the synaptic capacity for efficient acqui-sition of new “information clay” the next day.

Association, Integration,and Creativity

As critical as consolidation may be—an op-eration classically concerned with individualmemory items—the association and integra-tion of new experience into preexisting net-works of knowledge is equally as important, ifnot more so. The resulting creation of associa-tive webs of information offers numerous andpowerful advantages. Indeed, the final goal ofsleep-dependent memory processing may notbe the enhancement of individual memoriesin isolation, but, instead, their integration intoa common schema, and by this enhancement,facilitation of the development of universal con-cepts, a process that forms the basis of general-ized knowledge and even creativity.

Association and Integration

Perhaps the earliest demonstration that sleepmay be involved in a form of memory general-ization was by Fenn et al. (2003). Utilizing an ar-tificial grammar task, subjects were trained andlater tested on their ability to transfer phono-logical categories across different acousticpatterns. The task required forming new map-pings from complex acoustical sounds to pre-existing linguistic categories, which then gener-alized to new stimuli. As such, it involved botha declarative process of forming specific mem-ories associated with the learned stimuli, to-gether with a procedural component involvingmapping across the set of learned sounds thatsupports generalization to novel stimuli. Duringthe initial training session there was a signifi-cant improvement in recognition performanceon the task. However, when retested after a12-h waking interval, performance had de-cayed. Yet, if subjects were retested following


a night of sleep, this ability for memory gener-alization was restored. Supporting this concept,Dumay and Gaskell (2007) also demonstratedthat sleep and not equivalent time awakecan integrate related but novel phonemes intopre-existing, long-term lexical memory storesovernight.

In a related study by Gomez and colleagues(Gomez et al. 2006), infants were exposed to“phrases” from an artificial language duringa learning session—for example, phrases like“pel–wadim–jic”—until the infants became fa-miliar (as indexed by look responses). However,these three syllable units had an embedded rule,which was that the first and last unit formed arelationship of nonadjacency; in this case, pelpredicts jic. The infants were then retested sev-eral hours later, yet some infants took normallyscheduled naps, while others were scheduled ata time when they would not sleep after learn-ing. At later testing, infants again heard therecordings, along with novel phrases in whichthe predictive relationship between the first andlast word was new. Infants who did not sleeprecognized the phrases they had learned ear-lier, yet those who had slept demonstrated ageneralization of the predictive relationship tonew phrases, suggesting that the interveningprocess of sleep allowed the reinterpretation ofprior experience, and supported the abstrac-tion of commonalities—that is, the ability todetect a general pattern in new information.

Ellenbogen et al. (2007) have since testedthis sleep-dependent hypothesis of integrationby examining human relational memory—theability to generalize previously acquired associ-ations to novel situations. Participants initiallylearned five “premise pairs” (A > B, B > C,C > D, D > E, E > F). Unknown to sub-jects, the pairs contained an embedded hier-archy (A > B > C > D > E > F). Followingan offline delay of 20 min, 12 h across the day,or 12 h containing a night of sleep, knowledgeof this hierarchy was tested by examining re-lational judgments for novel “inference” pairs,either separated by one degree of associativedistance (B > D, C > E pairs) or by two degrees

of associative distance (B > E pair). Despite allgroups achieving near identical premise-pairretention after the offline delay (i.e., the build-ing block pair of the hierarchy), a striking disso-ciation was evident in the ability to make rela-tional inference judgments. Subjects that weretested soon after learning in the 20 min groupshowed no evidence of inferential ability, per-forming at chance levels (Fig. 9A). In contrast,the two 12 h groups displayed highly significantrelational memory development. Most remark-able, however, was the observation that if the12-h period contained a night of sleep, a near25% advantage in relational memory over the12 h across the day group was seen for themost distantly connected inferential judgment(the B > E pair; Fig. 9A). Together, these find-ings demonstrate that human memory integra-tion takes time to develop, requiring slow, of-fline associative processes. Furthermore, sleepappears to preferentially facilitate this integra-tion by enhancing hierarchical memory bind-ing, biasing the development of the most dis-tant/weak associative links amongst related yetseparate memory items (Fig. 9B). It is also inter-esting to note a further advantage of this sleep-dependent assimilation process. When it storesindividual premise-pairs (top row, Fig. 9B) thesize/number of items (“bits”) of informationthe brain has to code is ten (A-B, B-C, C-D, D-E, E-F). However, when it the items areformed into a hierarchy, the informational loadis compressed, reduced by nearly 50% to justsix bits (A-B-C-D-E-F). Therefore, a supple-mentary benefit of sleep-dependent memoryassociation may be the improved efficiency ofmemory storage, in addition to a more gener-alized representation.

Thus, the overnight strengthening and con-solidation of individual item memories (re-viewed above), may not be the ultimate ob-jective of sleep-dependent memory processing,especially when one considers that declarative(nonemotional) memories decay over the long-term (Wixted & Carpenter 2007). It is theninteresting to speculate whether sleep servesto facilitate two complementary objectives for


A B C D E FA B C D E F

A BA B

B CB C

C DC D

D ED E

E FE F

B CB C D ED E

A BA B C DC D E FE F

Immediate

Wake

Sleep

20 min 12 hr wake 12 hr sleep

Pe

rce

nt

Co

rre

ct

60

70

80

90

100

(chance) 50

Inference pairs

*

n.s.

**

n.s.

A

B

1 inference

1 & 2 inference

o

o o

1 2 1 2 1 2 o o o o o o

Figure 9. Sleep-dependent integration ofhuman relational memory. (A) Delayed inference(associative) memory performance (% correct) in arelational memory task following different offlinedelays. Immediate testing after just a 20 min offlinedelay, demonstrated a lack of any inferential abilityresulting in chance performance on both 1-degree(first order) and 2-degree and 2-degree (secondorder) associative judgments. Following a moreextended 12-h delay, across the day (Wake group),performance was significantly above chance acrossboth the one and two-degree inference judgments.However, following an equivalent 12-h offline delay,but containing a night of sleep (Sleep group),significantly better performance was expressed onthe more distant two-degree inference judgmentcompared with the one-degree judgment. (B) Aconceptual model of the effects of sleep on memoryintegration. Immediately after learning, the represen-tation of each premise is constituted as the choiceof one item over another (A > B, etc.), and thesepremises are isolated from one another despite hav-ing overlapping elements. After a 12-h period withno sleep, the premise representations are partiallyintegrated by their overlapping elements, sufficientto support first-order transitive inferences. How-ever, following a 12-h offline period with sleep, the

declarative memory, which span different timecourses. The first may be an initial process ofconsolidating individual item (episodic) mem-ories that are novel, which may occur in therelative short term. Over a longer time course,however, and utilizing these recently consol-idated item memories prior to their fading,sleep may begin the process of the brain’s ex-traction (of meaning) and abstraction (buildingassociational links with existing information),thereby creating more adaptive semantic net-works (McClelland et al. 1995; Squire 2007;Tse et al. 2007). Ultimately, individual itemmemories would no longer be necessary for thegoal that sleep is trying to achieve, and onlythe conceptual meaning of such experienceswould remain. Whether the subsequent loss ofitem memories is passive or whether sleep playsan active role in this process (Crick & Mitchi-son 1983) remains to be examined, but thisis a testable hypothesis, i.e., forgetting (individ-

ual items) is the price we pay for remembering (general

rules).

Creativity

One potential advantage of testing asso-ciative connections and building cross-linkedsystems of knowledge is creativity—the abilityto take existing pieces of information and com-bine them in novel ways that lead to greaterunderstanding and offer new, advantageous be-havioral repertoires. The link between creativ-ity and sleep, especially dreaming, has longbeen a topic of intense speculation. Even sci-entific examples of creativity occurring dur-ing sleep are common: from the dreams ofboth August Kekule that led to the concep-tion of a simple structure for benzene (Hubert1985) to those of Dmitry Mendeleyev that ini-tiated the creation of the periodic table of

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−premise representations are fully interleaved, support-ing both first- and second-order transitive inferences.∗P < 0.05; error bars indicate SEM. Modified fromEllenbogen et al. 2007 and Eichenbaum 2007.


elements (Strathern 2000) to the late nightdreaming of Otto Loewi that inspired theexperimental demonstration of neurochemi-cal transmission (Mazzarello 2000).Quantita-tive data have further demonstrated that solu-tion performance on tests of cognitive flexibilityusing anagram word puzzles is more than 30%better following awakenings from REM sleepcompared with NREM awakenings (Walkeret al. 2002). Similarly, a study of semantic prim-ing has demonstrated that, in contrast to the sit-uation in waking, performance following REMsleep awakenings shows a greater priming effectby weakly related words than by strong primes,while strong priming exceeds weak priming inNREM sleep (Stickgold et al. 1999), again in-dicating the highly associative properties of theREM sleep brain. Even the study of mental ac-tivity (dreams) from REM sleep indicates thatthere is not a concrete episodic replay of day-time experiences, but instead, a much more as-sociative process of semantic integration duringsleep (Fosse et al. 2003).

Yet the most striking experimental evidenceof sleep-inspired insight is arguably that re-ported by Wagner and colleagues (Wagner et al.2004). Using a mathematical “number reduc-tion task” (Thurstone & Thurstone 1941), aprocess of sleep-dependent creative insight waselegantly demonstrated. Subjects analyzed andworked through a series of 8-digit string prob-lems, using specific addition rules (Fig. 10A).Following initial training, after various peri-ods of wake or sleep, subjects returned for anadditional series of trials. When retested aftera night of sleep, subjects solved the task, us-ing this “standard” procedure, 16.5% faster.In contrast, subjects who did not sleep priorto retesting averaged less than a 6% improve-ment. However, hidden in the construction ofthe task was a much simpler way to solve theproblem. On every trial, the last three responsedigits (e.g., “4–1–9” in Fig. 10A) were the mirrorimage of the preceding three (i.e., “9–1–4”). Asa result, the second response digit always pro-vided the answer to the problem, and using such“insight,” subjects could stop after producing

WAKE(NIGHT)

WAKE

(DAY)

0%

20%

40%

60%

% S

ubje

cts

gain

ing insig

ht

9

1 1 4 4 9 4 9 4

1 4 4 1 91

SLEEP

(NIGHT)

A

B*

Figure 10. Sleep-dependent production of cre-ative insight. (A) Example of the Number ReductionTask. Subjects analyze digits in a 8-digit string of 1s,4s, and 9s, from left to right, using two rules: (1) Iftwo digits are the same, respond with that digit. Thus,starting from the left, the first two digits are both “1,”and hence the response (listed below and to the rightof the second digit) is also “1.” (2) If two digits aredifferent, respond with the remaining digit. Thus, hav-ing produced the response “1,” this response and thenext digits are analyzed. Since they differ (“1” and“4”), the next response is “the remaining digit, or“9.” This response and the next digit, “4” also differ(“9” and “4”) and so the next response is the remain-ing digit, “1.” The analysis is continued to the end,and the final response, “9” in this case, is the solu-tion to the problem. This final response is then enteredas the answer the problem. However, on every trial,the last three response digits (e.g., “4–1–9” in figureabove) are the mirror image of the preceding three(i.e., “9–1–4”). As a result, the second response digit(circled “9”) always provides the answer to the prob-lem, resulting in a “shortcut” to solving the problem,if the subject gains this hidden insight. (B) Percent-age of subjects that gained insight into this hiddenrule following an offline delay while awake acrossthe day, awake across the night of following sleepacross the night. ∗P < 0.05. Modified from Wagneret al. 2004.

the second response digit. Most dramatically,nearly 60% of the subjects who slept for a nightbetween training and retesting discovered thisshortcut the following morning (Fig. 10B). In


contrast, no more than 25% of subjects in anyof four different control groups that did notsleep had this insight. Sleeping after exposureto the problem therefore more than doubledthe likelihood of solving it (although it is inter-esting to note that this insight was not presentimmediately following sleep, but took over 100trials on average to emerge the next day).

In summary, substantial evidence now sug-gests that sleep serves a metalevel role inmemory processing that moves far beyond theconsolidation and strengthening of individualmemories and, instead, aims to intelligentlyassimilate and generalize these details offline.In doing so, sleep may offer the ability to testand build common informational schemas ofknowledge, providing for increasingly accuratestatistic predictions about the world and allow-ing for the discovery of novel, even creative,next-day solution insights.

Emotional Regulation

Despite substantial research focusing on theinteraction between sleep and cognition, espe-cially memory, the impact of sleep and sleeploss on affective and emotional regulation hasreceived more limited research attention. Thisabsence of investigation is perhaps surprisingconsidering that nearly all psychiatric and neu-rological mood disorders express co-occurringabnormalities of sleep, suggesting an intimaterelationship between sleep and emotion. Nev-ertheless, a number of recent studies evaluatingsubjective as well as objective measures of moodand affect, combined with insights from clinicaldomains, offer an emerging understanding forthe critical role of sleep in regulating emotionalbrain function.

Affective Reactivity

Together with impairments of attention andalertness, sleep deprivation is commonly asso-ciated with increased subjective reports of ir-ritability and affective volatility (Horne 1985).

Using a sleep restriction paradigm (5 h/night),Dinges et al. (1997) reported a progressive in-crease in emotional disturbance across a one-week period on the basis of questionnairemood scales. In addition, subjective descrip-tions in participants’ daily journals also indi-cated increasing complaints of emotional dif-ficulties. Zohar et al. (2005) investigated theeffects of sleep disruption on emotional re-activity to daytime work events in medicalresidents. Sleep loss was shown to amplify nega-tive emotional consequences of disruptive day-time events while blunting the positive benefitassociated with rewarding or goal-enhancingactivities.

Although these findings help to characterizethe behavioral irregularities imposed by sleeploss, evidence for the role of sleep in regulat-ing our emotional brain is surprisingly scarce.To date, only one such study has investigatedwhether a lack of sleep inappropriately mod-ulates human emotional brain reactivity (Yooet al. 2007a). Healthy young participants wereallowed to sleep normally prior to an fMRIscanning session or were sleep deprived forone night (accumulating approximately 35 hof total sleep loss). During scanning, subjectsperformed an affective stimulus viewing task in-volving the presentation of picture slides rang-ing in a gradient from emotionally neutral toincreasingly negative and aversive.

While both groups expressed significantamygdala activation in response to increas-ingly negative picture stimuli, those in thesleep-deprivation condition exhibited a re-markable +60% greater magnitude of amyg-dala reactivity, relative to the control group(Figs. 11A & B). In addition to this increasedintensity of activation, there was also a three-fold increase in the extent of amygdala vol-ume recruited in response to the aversive stim-uli in the sleep-deprivation group (Fig. 11B).Perhaps most interestingly, relative to the sleepcontrol group, there was a significant loss offunctional connectivity identified between theamygdala and the medial prefrontal cortex(mPFC) in those who were sleep deprived—a


Figure 11. The impact of sleep deprivation on emotional brain reactivity and functional connectivity.(A) Amygdala response to increasingly negative emotional stimuli in the sleep deprivation and sleep controlgroups, and (B) corresponding differences in intensity and volumetric extent of amygdala activation betweenthe two groups (average ± SEM. of left and right amygdala). (C) Depiction of associated changes in functionalconnectivity between the medial prefrontal cortex (mPFC) and the amygdala. With sleep, the prefrontal lobewas strongly connected to the amygdala, regulating and exerting and inhibitory top-down control, yet (D)Without sleep, however, mPFC connection was decreased, potentially negating top-down control and resultingin an overactive amygdala. ∗P < 0.01; error bars indicate SEM. Modified from Yoo et al. 2007b.

region known to have strong inhibitory pro-jections and hence modulatory impact on theamygdala (Sotres-Bayon et al. 2004). In con-trast, significantly greater connectivity in thedeprivation group was observed between theamygdala and the autonomic-activating cen-ters of the locus coeruleus.

Thus, without sleep, an amplified hyper-limbic reaction by the human amygdala wasobserved in response to negative emotionalstimuli. Furthermore, this altered magnitudeof limbic activity is associated with a loss offunctional connectivity with the mPFC in thesleep deprivation condition implying a failureof top-down inhibition by the prefrontal lobe

(Figs. 11C & D). It would therefore appear thata night of sleep may “reset” the correct affec-tive brain reactivity to next-day emotional chal-lenges by maintaining functional integrity ofthis mPFC−amygdala circuit and thus governappropriate behavioral repertoires (e.g., opti-mal social judgments and rational decisions).Intriguingly, a similar pattern of anatomicaldysfunction has been implicated in a numberof psychiatric mood disorders that express co-occurring sleep abnormalities (Davidson 2002;Davidson et al. 2002; New et al. 2007), directlyraising the issues of whether such factors (sleeploss and clinical mood disorders) are causallyrelated.


Emotional Information Processing

Sleep’s role in declarative memory consoli-dation, rather than being absolute, may dependon more intricate aspects of the information be-ing learned, such as novelty, meaning to extract,and also the affective salience of the material.Independent of the field of sleep and memory,there is a wealth of evidence demonstrating thatmemory processing is modulated by emotion(Cahill 2000; McGaugh 2004; Phelps 2004).Experiences which evoke emotions not only en-code more strongly, but appear to persist andeven improve over time as the delay betweenlearning and testing increases (from hours todays) (Kleinsmith & Kaplan 1963; Walker &Tarte 1963; Levonian 1972; LaBar & Phelps1998; Sharot & Phelps 2004).

Although these findings indicate a strong in-fluence of emotion on slow, time-dependentconsolidation processes, based on the coinci-dent neurophysiology that REM sleep pro-vides and the neurobiological requirements ofemotional memory processing (Cahill 2000;McGaugh 2004), work has now begun to test aselective REM-dependent hypothesis of affec-tive human memory consolidation. For exam-ple, Hu et al. (2006) compared the consolida-tion of emotionally arousing and nonarousingpicture-stimuli following a 12-h period acrossa day or following a night of sleep. A specificemotional memory benefit was observed onlyfollowing sleep and not cross an equivalent timeawake. Wagner and colleagues (Wagner et al.2001) have also shown that sleep selectively fa-vors the retention of previously learned emo-tional texts relative to neutral texts, and thatthis affective memory benefit is only presentfollowing late-night sleep (a time period rich instage-2 NREM and REM sleep). Furthermore,this emotional memory enhancement has beenshown to persist for several years (Wagner et al.2006).

Using a nap paradigm, researchers havemost recently demonstrated that sleep, andspecifically REM neurophysiology, may under-lie this consolidation benefit (Nishida et al. in

press). Subjects performed two study sessionsin which they learned emotionally negativeand neutral picture stimuli: one 4 h prior toa recognition memory test, and one 15 minprior to it. In one group, participants slept(90 min nap) after the first study session, while inthe other group, participants remained awake.Thus, items from the study sessions tested after4 h transitioned through different brain-statesin each group prior to testing—sleep in the Napgroup and no sleep in the No-Nap group—yet experienced identical brain-state conditionswhen tested after 15 min.

No change in memory for emotional (orneutral stimuli), occurred across the offlinedelay in the no-nap group. However, a sig-nificant and selective offline enhancement ofemotional memory was observed in the napgroup (Fig. 12A), the extent of which was corre-lated with the amount of REM sleep (Fig. 12B),and the speed of entry into REM (latency; notshown in figure). Furthermore, spectral analysisof the EEG demonstrated that the magnitudeof right-dominant prefrontal theta power dur-ing REM (activity in the frequency range of4.0–7.0 Hz) exhibited a significant and posi-tive relationship with the amount of emotionalmemory improvement (Figs. 12C & D).

These findings go beyond demonstratingthat affective memories are preferentially en-hanced across periods of sleep, and indicatethat the extent of emotional memory im-provement is associated with specific REMsleep characteristics—both quantity and qual-ity. Corroborating these correlations, it has pre-viously been hypothesized that REM sleep rep-resents a brain state particularly amenable toemotional memory consolidation, based on itsunique biology (Pare et al. 2002; Hu et al.2006). Neurochemically, levels of limbic andforebrain ACh are markedly elevated duringREM (Vazquez & Baghdoyan 2001), report-edly quadruple those seen during NREM anddouble those measured in quite waking (Mar-rosu et al. 1995). Considering the known im-portance of ACh in the long-term consolidationof emotional learning (McGaugh 2004), this


Figure 12. REM sleep enhancement of negative emotional memory consolidation. (A) Offline benefit(change in memory recall for 4 h versus 15 min old memories) across the day (wake, grey bar) or following a90 min nap (sleep, filled bar) (B) Correlation between the amount of offline emotional memory improvementin the nap group (i.e. the offline benefit expressed in filled bar of figure A), and the amount of REM sleepobtained within the nap, (C) Correlation strength (Pearson’s r-value) between offline benefit for emotionalmemory in the sleep group (the benefit expressed in filled bar of figure A) and the relative right versus leftprefrontal spectral-band power ([F4 – F3]) within the delta, alpha, theta, and beta spectral bands, expressedin average 0.5 Hz bins increments. Correlation strength is represented by the color range, demonstratingsignificant correlations within the theta frequency band (hot colors), and (D) exhibiting a maximum significanceat the 5.75 Hz bin. ∗P < 0.05; error bars indicate SEM. Modified from Nishida et al. unpublished.

pro-cholinergic REM state may result in a se-lective facilitation of affective memories, similarto that reported using experimental manipula-tions of ACh (Power 2004). Neurophysiologi-cally, theta oscillations have been proposed as a

carrier frequency allowing disparate brain re-gions that initially encode information to se-lectively interact offline, in a coupled relation-ship. By doing so, REM theta may afford theability to promote the strengthening of specific


memory representations across distributed net-works (Buzsaki 2002; Jones & Wilson 2005).

A REM Sleep Hypothesis of EmotionalMemory Processing

Beyond the strengthening of emotionalmemories, there may be an additional conse-quence of sleep-dependent affective modula-tion, and one that has significant implicationsfor mood disorders – that is, sleeping to for-get. Based on the emerging interaction betweensleep and emotion, below I outline a model ofaffective information processing that may of-fer brain-based explanatory insights regardingthe impact of sleep abnormalities, particularlyREM, for the initiation and/or maintenance ofmood disturbance.

While there is abundant evidence to sug-gest that emotional experiences persist in ourautobiographies over time, an equally remark-able but far less noted change is a reduction inthe affective tone associated with their re-call. The reason that affective experiences ap-pear to be encoded and consolidated morepreferentially than neutral memories is dueto autonomic neurochemical reactions elicitedat the time of the experience, creating whatwe commonly term an “emotional–memory.”However, the later recall of these experi-ences tends not to be associated with any-where near the same magnitude of autonomic(re)activation as that elicited at the moment oflearning/experience – suggesting that, over-time, the affective “blanket” previously en-veloped around the memory during encodinghas been removed, while the information con-tained within that experience (the memory) re-mains.

For example, neuroimaging studies haveshown that initial exposure and learningof emotional stimuli is associated with sub-stantially greater activation in the amygdalaand hippocampus, relative to neutral stimuli(Kilpatrick & Cahill 2003; Dolcos et al. 2004;Dolcos et al. 2005). In one of these studies(Dolcos et al. 2004), however, when partici-

pants were reexposed to these same stimuliduring recognition testing many months later,a change in the profile of activation occurred(Dolcos et al. 2005). Although the same magni-tude of differential activity between emotionaland neutral items was observed in the hip-pocampus, this was not true in the amygdala.Instead, the difference in amygdala (re)activitycompared with neutral items had dissipatedover time. This would support the idea thatthe strength of the memory (hippocampal-associated activity) remains at later recollection,yet the associated emotional reactivity to theseitems (amygdala activity) is reduced over time.

The hypothesis predicts that this decouplingpreferentially takes place overnight, such thatwe sleep to forget the emotional tone, yetsleep to remember the tagged memory of thatepisode (Fig. 13). The model further arguesthat if this process is not achieved, the mag-nitude of visceral autonomic “charge” remain-ing within autobiographical memory networkswill persist, resulting in the potential condi-tion of chronic anxiety. Based on the con-sistent relationship identified between REMand emotional processing, combined with itsunique neurobiology, the hypothesis proposesthat REM sleep provides an optimal biolog-ical state for achieving such affective “ther-apy.” Specifically, increased activity within lim-bic and paralimbic structures (including thehippocampus and amygdala) during REM mayfirst offer the ability for reactivation of previ-ously acquired affective experiences. Second,the neurophysiological signature of REM in-volving dominant theta oscillations within sub-cortical as well as cortical nodes may offerlarge-scale network cooperation at night, al-lowing the integration and, as a consequence,greater understanding of recently experiencedemotional events in the context of preexistingneocortically stored semantic memory. Third,these interactions during REM critically, andperhaps most importantly, take place within abrain that is devoid of aminergic neurochemicalconcentration (Pace-Schott & Hobson 2002),particularly noradrenergic input from the


Figure 13. Model of sleep-dependent emotional–memory processing: A sleep to forgetand sleep to remember hypothesis. When formed, a newly encoded “emotional–memory”is created in a milieu of high adrenergic tone, resulting an associated affective “blanket.”With multiple iterations of sleep, particularly REM, not only is the informational core (memory)contained within that affective experience strengthened overnight(s), resulting in improvedmemory for that event, the autonomic tone “enveloped” around the memory becomes grad-ually ameliorated (emotional forgetting). Over time, the stored information of the originalexperience ultimately becomes decoupled and freed of its autonomic “charge,” leaving justthe salient memory of that emotional experience, but without the affective tone previouslyassociated at the time of learning.

locus coeruleus; the influence of which hasbeen linked to states of high stress and anxietydisorders (Sullivan et al. 1999).

In summary, the described neuroanatomical,neurophysiological and neurochemical condi-tions of REM sleep offer a unique biologicaltheatre in which to achieve, on one hand, a bal-anced neural potentiation of the informationalcore of emotional experiences (the memory),yet may also depotentiate and ultimately ame-liorate the autonomic arousing charge orig-inally acquired at the time of learning (theemotion), negating a long-term state of anxi-ety (Fig. 13).

This model compliments previous psycho-logical theories of dreaming by Greenberg(Greenberg et al. 1972a, 1972b) and alsoCartwright (Cartwright et al. 1991, 1998,2006), which suggest that the process of REM-sleep mental activity aids in the resolution of

previous emotional conflict, resulting in im-proved next-day negative mood. Moreover,pioneering work by Cartwright et al. havedemonstrated that not only the occurrence ofdreaming but the actual content of dreamsplays an important role in the recovery fromemotional trauma, and can be predictive ofclinical remission months later (Cartwrightet al. 1991, 1998, 2006). Although the cur-rent model offers a neurobiological frameworkfor the overnight modulation and alterationof emotional memories and next-day affectivebrain reactivity, is does not discount the poten-tial contribution that the mental operation ofdreaming itself, beyond the electrophysiologi-cal underpinnings of REM, may afford to thisprocess.

The current model also asserts that if theprocess of divorcing emotion from memory isnot achieved across the first night following


an emotional event, a repeat attempt wouldoccur on the second night, since the strength ofemotional “tag” associated with the memorywould remain high. Should this process fail asecond time, the same events would continueto repeat across ensuing nights, potentially withan increasing progressive amount of REM inresponse. It is just such a cycle of REM-sleepdreaming (nightmares) that represents a diag-nostic key feature of post-traumatic stress dis-order (PTSD) (Lavie 2001). It may not be coin-cidental, therefore, that these patients continueto display hyperarousal reactions to associatedtrauma cues (Harvey et al. 2003; Pole 2007),indicating that the process of separating the af-fective tone from the emotional experience hasnot been accomplished. The reason why sucha REM mechanism may fail in PTSD remainsunknown, although the exceptional magnitudeof trauma-induced emotion at the time of learn-ing maybe so great that the system is incapableof initiating/completing one or both these pro-cesses, leaving some patients unable to depo-tentiate, integrate and hence “overcome” theexperience.

This model also makes specific experimen-tal predictions as to the fate of these twocomponents – the memory and the emotion.As partially demonstrated, the first predictionwould be that, overtime, the veracity of thememory itself would improve, and the extentto which these [negative] emotional experi-ences are strengthened would be proportionalto the amount of postexperience REM sleepobtained, as well as how quickly it is achieved(REM latency).

Secondly, using autonomic physiology mea-sures, these same predictions would hold in theinverse direction for the magnitude of emo-tional reactivity induced at the time of re-call. Together with the neuroimaging studiesof emotional memory recall over time, and thepsychological studies investigating the role ofREM sleep dreaming in mood regulation, a re-cent fMRI study offers perhaps the strongestpreliminary support of this sleep-dependentmodel of emotional-memory processing (Ster-

penich et al. 2007). The investigation demon-strated that subjects who were deprived of sleepthe first night after learning arousing emotionpicture slides not only showed reduced recalledof the information (the sleep to remember com-ponent of the hypothesis), but also showed alack of reduction in amygdala reactivity whenreexposed to these same emotional picture slideat recognition testing – as compared to a con-trol group that did sleep (the sleep to forgetcomponent of the hypothesis).

There is, however, a related study that doesnot conform to these trends. Wagner and col-leagues had participants subjectively rate andre-rate emotional picture slides after 3 h ofearly-night sleep, or 3 h of late-night sleep(Wagner et al. 2002). Valence ratings of un-pleasantness actually increased following sleep,compared to new picture slides not seen be-fore. This was true in a group that was alsoallowed to sleep the entire night. The contrastin this finding to those discussed previously re-porting a decrease in emotional reactivity re-mains unclear. It is of note that when com-pared not to new items, but to the same itemsbefore sleep (baseline ratings), no increase invalance rating was evident. This discrepancymay also be due to the dimension of valencenot assessing autonomic emotional reactivity.In fact, subjects also rated these pictures stim-uli on the basis of arousal strength, as well asvalence. There was no such amplification ofarousal reactivity following sleep, demonstrat-ing a numerical decrease overtime, relative tobaseline measures. Alternatively, it may be thatthe assessment of valence is associated withthe veracity of the memory, which is strength-ened overnight, thereby promoting the recallof perceived pleasantness (or unpleasantness).Arousal (the visceral, autonomic dimension ofemotion), in contrast, appears to be reducedover sleep, despite the beneficial strengtheningof memory recall.

The model thirdly predicts that a patho-logical increase in REM (as commonly oc-curs in depression; (Tsuno et al. 2005; Ar-mitage 2007; Gottesmann & Gottesman 2007)


may disproportionately amplify the strength ofnegative memories, so much so that, despiteconcomitant attempts at ameliorating the asso-ciated affective tone, would still create a per-ceived autobiographical history dominated bynegative memory excess (which may also fa-cilitate disadvantageous waking rumination).In contrast, the selective decrease of REM,as occurs with many antidepressants, wouldpredict a reduction of such negative memoryconsolidation and bias, although may curtailthe degree of affect decoupling that can oc-cur. Long-term, the balanced extent of accu-mulated REM should therefore correlate notonly with the persistence, in memory, of theemotional experience, it should also be associ-ated with a decreased magnitude of autonomicresponse associated with recall – all of whichare testable objectives for future research.

If such a hypothesis is correct, there wouldbe translation implications for psychiatric andmood disorders. This would require a new ap-preciation for the palliative role of sleep in treat-ment regimes, and a consideration of whetheraltering sleep architecture to regulation the bal-ance of emotion and memory of past experi-ence is a useful and viable possibility.

Conclusions

While not fully complete, we will soon havea new taxonomy of sleep-dependent memoryprocessing, and one that will supersede the po-larized all-or-none views of the past (Stickgold& Walker 2005b; Vertes & Siegel 2005). Withsuch findings, we can come to a revised appre-ciation of how both wake and sleep unite in asymbiotic alliance to coordinate the encoding,consolidation and integration of our memories,the ultimate aim of which maybe to create ageneralized catalogue of stored knowledge thatdoes not rely on the verbose retention of allpreviously learned facts.

Beyond memory and plasticity, a growingnumber of human neuroscience studies, seton a foundation of clinical insights, point to

an exciting role for sleep in regulating affec-tive brain function and emotional experience.Based on the remarkable neurobiology of sleep,and REM in particular, a unique capability forthe overnight modulation of affective networksand previously encountered emotional experi-ences may be possible, redressing and main-taining the appropriate connectivity and hencenext-day reactivity throughout limbic and as-sociated autonomic systems.

Ultimately, the timeless maternal wisdom ofmothers may have long held the answers toAllan Rechtschaffen’s original question “Whydo we sleep?” Namely, that “you should sleepon a problem” and when troubled, “get to bed,you’ll feel better in the morning.”

Acknowledgments

The author wishes to thank Edwin Robert-son, Robert Stickgold, Allison Harvey, NinadGujar, and Els van der Helm for thoughtfulinsights. This work was supported in part bygrants from the National Institutes of Health(MH069935) and the American Academy ofSleep Medicine.

Conflicts of Interest

The author declares no conflicts of interest.

References

Amzica, F., & Steriade, M. (1995). Short- and long-rangeneuronal synchronization of the slow (<1 Hz) corti-cal oscillation. J. Neurophysiol., 73, 20–38.

Armitage, R. (2007). Sleep and circadian rhythms inmood disorders. Acta Psychiatr. Scand. Suppl., 104–115.

Aserinsky, E., & Kleitman, N. (1953). Regularly occurringperiods of eye motility and concurrent phenomenaduring sleep. Science, 118, 273–274.

Aston-Jones, G., & Bloom, F. E. (1981). Activity ofnorepinephrine-containing locus coeruleus neuronsin behaving rats anticipates fluctuations in the sleep-waking cycle. J. Neurosci., 1, 876–886.

Benca, R. M., Obermeyer, W. H., Thisted, R. A., & Gillin,J. C. (1992). Sleep and psychiatric disorders. A meta-analysis. Arch. Gen. Psychiatry, 49, 651–668; discussion669–670.


Buysse, D. J. (2004). Insomnia, depression and aging. As-sessing sleep and mood interactions in older adults.Geriatrics, 59, 47–51; quiz 52.

Buzsaki, G. (1996). The hippocampo-neocortical dia-logue. Cereb. Cortex, 6 , 81–92.

Buzsaki, G. (2002). Theta oscillations in the hippocampus.Neuron, 33, 325–340.

Cahill, L. (2000). Neurobiological mechanisms of emo-tionally influenced, long-term memory. Prog. Brain

Res., 126 , 29–37.Callaway, C. W., Lydic, R., Baghdoyan, H. A., & Hobson,

J. A. (1987). Pontogeniculooccipital waves: Sponta-neous visual system activity during rapid eye move-ment sleep. Cell. Mol. Neurobiol., 7, 105–149.

Cann, A., & Ross, D. A. (1989). Olfactory stimuli as con-text cues in human memory. Am. J. Psychol., 102,91–102.

Cartwright, R., Agargun, M. Y., Kirkby, J., & Friedman,J. K. (2006). Relation of dreams to waking concerns.Psychiatry Res., 141, 261–270.

Cartwright, R., Luten, A., Young, M., Mercer, P., & Bears,M. (1998). Role of REM sleep and dream affectin overnight mood regulation: a study of normalvolunteers. Psychiatry Res., 81, 1–8.

Cartwright, R. D., Kravitz, H. M., Eastman, C. I., &Wood, E. (1991). REM latency and the recoveryfrom depression: getting over divorce. Am. J. Psychia-

try, 148, 1530–1535.Clemens, Z., Fabo, D., & Halasz, P. (2005). Overnight

verbal memory retention correlates with the numberof sleep spindles. Neuroscience, 132, 529–535.

Crick, F., & Mitchison, G. (1983). The function of dreamsleep. Nature, 304, 111–114.

Datta, S. (2000). Avoidance task training potentiates pha-sic pontine-wave density in the rat: A mechanismfor sleep-dependent plasticity. J. Neurosci., 20, 8607–8613.

Datta, S., Li, G., & Auerbach, S. (2008). Activation ofphasic pontine-wave generator in the rat: a mech-anism for expression of plasticity-related genes andproteins in the dorsal hippocampus and amygdala.Eur J. Neurosci., 27, 1876–1892.

Dave, A. S., & Margoliash, D. (2000). Song replay duringsleep and computational rules for sensorimotor vocallearning. Science, 290, 812–816.

Dave, A. S., Yu, A. C., & Margoliash, D. (1998). Behav-ioral state modulation of auditory activity in a vocalmotor system. Science, 282, 2250–2254.

Davidson, R. J. (2002). Anxiety and affective style: role ofprefrontal cortex and amygdala. Biol. Psychiatry, 51,68–80.

Davidson, R. J., Pizzagalli, D., Nitschke, J. B., & Putnam,K. (2002). Depression: perspectives from affectiveneuroscience. Annu. Rev. Psychol., 53, 545–574.

Dinges, D. F., Pack, F., Williams, K., Gillen, K. A., Powell,

J. W., et al. (1997). Cumulative sleepiness, mood dis-turbance, and psychomotor vigilance performancedecrements during a week of sleep restricted to 4–5hours per night. Sleep, 20, 267–277.

Dolcos, F., LaBar, K. S., & Cabeza, R. (2004). Interac-tion between the amygdala and the medial tempo-ral lobe memory system predicts better memory foremotional events. Neuron, 42, 855–863.

Dolcos, F., LaBar, K. S., & Cabeza, R. (2005). Remem-bering one year later: role of the amygdala and themedial temporal lobe memory system in retrievingemotional memories. Proc. Natl. Acad. Sci. USA 102,2626–2631.

Drummond, S. P., & Brown, G. G. (2001). The effects oftotal sleep deprivation on cerebral responses to cog-nitive performance. Neuropsychopharmacol. 25, S68–73.

Drummond, S. P., Brown, G. G., Gillin, J. C., Stricker, J.L., Wong, E. C., et al. (2000). Altered brain responseto verbal learning following sleep deprivation. Nature

403, 655–657.Dumay, N., & Gaskell, M. G. (2007). Sleep-associated

changes in the mental representation of spokenwords. Psychol. Sci. 18, 35–39.

Eichenbaum, H. (2004). Hippocampus: cognitive pro-cesses and neural representations that underliedeclarative memory. Neuron 44, 109–120.

Eichenbaum, H. (2007). To sleep, perchance to integrate.Proc. Natl. Acad. Sci. USA, 104, 7317–7318.

Ellenbogen, J. M., Hulbert, J. C., Stickgold, R., Dinges,D. F., & Thompson-Schill, S. L. (2006a). Interferingwith theories of sleep and memory: sleep, declarativememory, and associative interference. Curr. Biol. 16 ,1290–1294.

Ellenbogen, J. M., Payne, J. D., & Stickgold, R. (2006b).The role of sleep in declarative memory consolida-tion: passive, permissive, active or none? Curr. Opin.

Neurobiol., 16 , 716–722.Ellenbogen, J., Hu, P., Payne, J.D., Titone, D. & Walker,

M.P. (2007). Human relational memory requirestime and sleep. Proc. Natl. Acad. Sci. USA, 104, 7723–7728.

Fenn, K. M., Nusbaum, H. C., & Margoliash, D. (2003).Consolidation during sleep of perceptual learning ofspoken language. Nature, 425, 614–616.

Fogel, S., Jacob, J., & Smith, C. (2001). Increased sleepspindle activity following simple motor procedurallearning in humans. Congress Physiological Basisfor Sleep Medicine, pp. 123.

Fogel, S. M., & Smith, C. T. (2006). Learning-dependentchanges in sleep spindles and Stage 2 sleep. J. Sleep

Res., 15, 250–255.Fosse, M. J., Fosse, R., Hobson, J. A., & Stickgold, R. J.

(2003). Dreaming and episodic memory: a functionaldissociation? J. Cogn. Neurosci., 15, 1–9.


Frankland, P. W., & Bontempi, B. (2005). The organiza-tion of recent and remote memories. Nat. Rev. Neu-

rosci., 6 , 119–130.Gais, S., Albouy, G., Boly, M., Dang-Vu, T. T., Darsaud,

A., et al. (2007). Sleep transforms the cerebral traceof declarative memories. Proc. Natl. Acad. Sci. USA,104, 18778–18783.

Gais, S., & Born, J. (2004). Low acetylcholine during slow-wave sleep is critical for declarative memory consol-idation. Proc. Natl. Acad. Sci. USA, 101, 2140–2144.

Gais, S., Molle, M., Helms, K., & Born, J. (2002).Learning-dependent increases in sleep spindle den-sity. J. Neurosci., 22, 6830–6834.

Giuditta, A., Ambrosini, M. V., Montagnese, P., Mandile,P., Cotugno, M., et al. (1995). The sequential hy-pothesis of the function of sleep. Behav. Brain Res., 69,157–166.

Gomez, R. L., Bootzin, R. R., & Nadel, L. (2006). Napspromote abstraction in language-learning infants.Psychol. Sci., 17, 670–674.

Gottesmann, C., & Gottesman, I. (2007). The neurobio-logical characteristics of rapid eye movement (REM)sleep are candidate endophenotypes of depression,schizophrenia, mental retardation and dementia.Prog. Neurobiol., 81, 237–250.

Greenberg, R., Pearlman, C. A., & Gampel, D. (1972a).War neuroses and the adaptive function of REMsleep. Br. J. Med. Psychol., 45, 27–33.

Greenberg, R., Pillard, R., & Pearlman, C. (1972b). Theeffect of dream (stage REM) deprivation on adapta-tion to stress. Psychosom. Med., 34, 257–262.

Hammond, N. (2004). Fragmentary Voices: Memory and Edu-

cation at Port-Royal. Gunter Narr Verlag.Harrison, Y., & Horne, J. A. (2000). Sleep loss and tem-

poral memory. Q. J. Exp. Psychol., 53, 271–279.Hartley, D. (1801). Observations on Man, His frame, his de-

ity, and his expectations (1749/)6691. Gainesville, FL:Scholars Facsimile Reprint.

Harvey, A. G., Jones, C., & Schmidt, D. A. (2003). Sleepand posttraumatic stress disorder: a review. Clin. Psy-

chol. Rev., 23, 377–407.Hobson, J. A., McCarley, R. W., & Wyzinski, P. W. (1975).

Sleep cycle oscillation: Reciprocal discharge by twobrainstem neuronal groups. Science, 189, 55–58.

Horne, J. A. (1985). Sleep function, with particular ref-erence to sleep deprivation. Ann. Clin. Res., 17, 199–208.

Hu, P., Stylos-Allen, M., & Walker, M. P. (2006).Sleep facilitates consolidation of emotionally arous-ing declarative memory. Psychol. Sci., 17, 891–898.

Huber, R., Ghilardi, M. F., Massimini, M., Ferrarelli, F.,Riedner, B. A., et al. (2006). Arm immobilizationcauses cortical plastic changes and locally decreasessleep slow wave activity. Nat. Neurosci., 9, 1169–1176.

Huber, R., Ghilardi, M. F., Massimini, M., & Tononi, G.(2004). Local sleep and learning. Nature, 430, 78–81.

Hubert, A. (1985). “Kekule von Stradonitz, Friedrich August.”

New York: Macmillan.Jenkins, J. G., & Dallenbach, K. M. (1924). Obliviscence

during sleep and waking. Am. J. Psychol., 35, 605–12.

Ji, D., & Wilson, M. A. (2007). Coordinated memory re-play in the visual cortex and hippocampus duringsleep. Nat. Neurosci., 10, 100–107.

Jones, M. W., & Wilson, M. A. (2005). Theta rhythmscoordinate hippocampal-prefrontal interactions in aspatial memory task. PLoS Biol., 3, e402.

Kametani, H., & Kawamura, H. (1990). Alterations inacetylcholine release in the rat hippocampus duringsleep-wakefulness detected by intracerebral dialysis.Life Sci., 47, 421–426.

Kilpatrick, L., & Cahill, L. (2003). Amygdala modulationof parahippocampal and frontal regions during emo-tionally influenced memory storage. Neuroimage, 20,2091–2099.

Kleinsmith, L. J., & Kaplan, S. (1963). Paired-associatelearning as a function of arousal and interpolatedinterval. J. Exp. Psychol., 65, 190–193.

LaBar, K. S., & Phelps, E. A. (1998). Arousal-mediatedmemory consolidation: Role of the medial temporallobe in humans. Psychol. Sci., 9, 490–493.

Lavie, P. (2001). Sleep disturbances in the wake of trau-matic events. N. Engl. J. Med., 345, 1825–1832.

Levonian, E. (1972). Retention over time in relation toarousal during learning: an explanation of discrepantresults. Acta Psychol. (Amst.), 36 , 290–321.

Llinas, R., & Ribary, U. (1993). Coherent 40-Hz oscilla-tion characterizes dream state in humans. Proc. Natl.

Acad. Sci. USA, 90, 2078–2081.Louie, K., & Wilson, M. A. (2001). Temporally struc-

tured replay of awake hippocampal ensemble activityduring rapid eye movement sleep. Neuron, 29, 145–156.

Lydic, R., & Baghdoyan, H. A. (1988). Handbook Of Be-

havioral State Control: Cellular and Molecular Mechanisms.Boca Raton, FL: CRC Press.

Marr, D. (1970). A theory for cerebral neocortex. Proc. R.

Soc. Lond. B Biol. Sci., 176 , 161–234.Marrosu, F., Portas, C., Mascia, M. S., Casu, M. A., Fa,

M., et al. (1995). Microdialysis measurement of cor-tical and hippocampal acetylcholine release duringsleep-wake cycle in freely moving cats. Brain Res.,671, 329–332.

Marshall, L., & Born, J. (2007). The contribution of sleepto hippocampus-dependent memory consolidation.Trends Cogn. Sci., 11, 442–450.

Marshall, L., Helgadottir, H., Molle, M., & Born, J. (2006).Boosting slow oscillations during sleep potentiatesmemory. Nature, 444, 610–613.


McClelland, J. L., McNaughton, B. L., & O’Reilly, R.C. (1995). Why there are complementary learningsystems in the hippocampus and neocortex: insightsfrom the successes and failures of connectionist mod-els of learning and memory. Psychol. Rev., 102, 419–457.

Mazzarello, P. (2000). What dreams may come? Nature,408, 523.

McGaugh, J. L. (2004). The amygdala modulates the con-solidation of memories of emotionally arousing ex-periences. Annu. Rev. Neurosci., 27, 1–28.

Meier-Koll, A., Bussmann, B., Schmidt, C., &Neuschwander, D. (1999). Walking through a mazealters the architecture of sleep. Percept. Mot. Skills, 88,1141–1159.

Miller, G. (2007). Neuroscience. Hunting for meaningafter midnight. Science, 315, 1360–1363.

Morris, G. O., Williams, H. L., & Lubin, A. (1960). Mis-perception and disorientation during sleep. Arch. Gen.

Psychiatry, 2, 247–254.Moscovitch, M., & Nadel, L. (1998). Consolidation and

the hippocampal complex revisited: in defense of themultiple-trace model. Curr. Opin. Neurobiol., 8, 297–300.

Nadel, L., & Moscovitch, M. (1997). Memory consol-idation, retrograde amnesia and the hippocampalcomplex. Curr. Opin. Neurobiol., 7, 217–227.

New, A. S., Hazlett, E. A., Buchsbaum, M. S., Good-man, M., Mitelman, S. A., et al. (2007). Amygdala-prefrontal disconnection in borderline personalitydisorder. Neuropsychopharmacology, 32, 1629–1640.

Nishida, M., Pearsall, J., Buckner, R. L., & Walker, M.P. (In Press). Prefrontal theta during REM sleep en-hances emotional memory. Cereb. Cortex.

Nishida, M., & Walker, M. P. (2007). Daytime naps, motormemory consolidation and regionally specific sleepspindles. PloS One, 2, e341.

Nofzinger, E. A. (2005). Functional neuroimaging ofsleep. Semin. Neurol., 25, 9–18.

Pace-Schott, E. F., & Hobson, J. A. (2002). The neuro-biology of sleep: genetics, cellular physiology andsubcortical networks. Nat. Rev. Neurosci., 3, 591–605.

Paller, K. A., & Wagner, A. D. (2002). Observing thetransformation of experience into memory. Trends

Cogn. Sci., 6 , 93–102.Pare, D., Collins, D. R., & Pelletier, J. G. (2002). Amyg-

dala oscillations and the consolidation of emotionalmemories. Trends Cogn. Sci., 6 , 306–314.

Peigneux, P., Laureys, S., Fuchs, S., Collette, F., Perrin,F., et al. (2004). Are spatial memories strengthenedin the human hippocampus during slow wave sleep?Neuron, 44, 535–545.

Phelps, E. A. (2004). Human emotion and memory: inter-actions of the amygdala and hippocampal complex.Curr. Opin. Neurobiol., 14, 198–202.

Plihal, W., & Born, J. (1997). Effects of early and late noc-turnal sleep on declarative and procedural memory.J. Cogn. Neurosci., 9, 534–547.

Plihal, W., & Born, J. (1999). Memory consolidation inhuman sleep depends on inhibition of glucocorticoidrelease. Neuroreport, 10, 2741–2747.

Poe, G. R., Nitz, D. A., McNaughton, B. L., & Barnes, C.A. (2000). Experience-dependent phase-reversal ofhippocampal neuron firing during REM sleep. Brain

Res., 855, 176–180.Pole, N. (2007). The psychophysiology of posttraumatic

stress disorder: a meta-analysis. Psychol. Bull., 133,725–746.

Power, A. E. (2004). Muscarinic cholinergic contributionto memory consolidation: with attention to involve-ment of the basolateral amygdala. Curr. Med. Chem.,11, 987–996.

Rasch, B., Buchel, C., Gais, S., & Born, J. (2007).Odor cues during slow-wave sleep prompt declara-tive memory consolidation. Science, 315, 1426–1429.

Ribeiro, S., Gervasoni, D., Soares, E. S., Zhou, Y., Lin, S.C., et al. (2004). Long-lasting novelty-induced neu-ronal reverberation during slow-wave sleep in multi-ple forebrain areas induction of hippocampal long-term potentiation during waking leads to increasedextrahippocampal zif-268 expression during ensuingrapid-eye-movement sleep. PLoS Biol., 2, E24.

Rechtschaffen, A., & Kales, A. (1968). A manual standard-

ized terminology, techniques and scoring system for sleep stages

of human subjects. Bethesda, Maryland, USA: U.S. De-partment of Health.

Robertson, E. M. (2009). From creation to consolidation: anovel framework for memory processing. PLoS Biol.,

7, e19.Robertson, E. M., Pascual-Leone, A., & Miall, R. C.

(2004). Current concepts in procedural consolida-tion. Nat. Rev. Neurosci., 5, 576–82.

Schabus, M., Dang-Vu, T. T., Albouy, G., Balteau, E.,Boly, M., et al. (2007). Hemodynamic cerebral cor-relates of sleep spindles during human non-rapideye movement sleep. Proc. Natl. Acad. Sci. USA, 104,13164–13169.

Sejnowski, T. J., & Destexhe, A. (2000). Why do we sleep?Brain Res., 886 , 208–223.

Shaffery, J., Hoffmann, R., & Armitage, R. (2003). Theneurobiology of depression: perspectives from ani-mal and human sleep studies. Neuroscientist, 9, 82–98.

Sharot, T., & Phelps, E. A. (2004). How arousal modulatesmemory: disentangling the effects of attention andretention. Cogn. Affect Behav. Neurosci., 4, 294–306.

Shima, K., Nakahama, H., & Yamamoto, M. (1986). Fir-ing properties of two types of nucleus raphe dor-salis neurons during the sleep-waking cycle and theirresponses to sensory stimuli. Brain Res., 399, 317–326.


Skaggs, W. E., & McNaughton, B. L. (1996). Replay ofneuronal firing sequences in rat hippocampus duringsleep following spatial experience. Science, 271, 1870–1873.

Smith, C. (2001). Sleep states and memory processes inhumans: procedural versus declarative memory sys-tems. Sleep Med. Rev., 5, 491–506.

Smith, C. T., Aubrey, J. B., & Peters, K. R. (2004). Dif-ferent roles for REM and stage 2 sleep in motorlearning: A proposed model. Psychologica Belgica, 44,81–104.

Sotres-Bayon, F., Bush, D. E., & LeDoux, J. E. (2004).Emotional perseveration: an update on prefrontal-amygdala interactions in fear extinction. Learn. Mem.,11, 525–535.

Squire, L. R. (1992). Memory and the hippocampus: asynthesis from findings with rats, monkeys, and hu-mans. Psychol. Rev., 99, 195–231.

Squire, L. R. (2004). Memory systems of the brain: abrief history and current perspective. Neurobiol. Learn

Mem., 82, 171–177.Squire, L. R. (2007). Neuroscience. Rapid consolidation.

Science, 316 , 57–58.Squire, L. R., & Zola, S. M. (1996). Structure and function

of declarative and nondeclarative memory systems.Proc. Natl. Acad. Sci. USA, 93, 13515–13522.

Steriade, M. (2001). The Intact and Sliced Brain. Cambridge,MA: MIT Press.

Steriade, M., & Amzica, F. (1998). Coalescence of sleeprhythms and their chronology in corticothalamicnetworks. Sleep Res Online, 1, 1–10.

Steriade, M., Amzica, F., & Contreras, D. (1996). Syn-chronization of fast (30–40 Hz) spontaneous corti-cal rhythms during brain activation. J. Neurosci., 16 ,392–417.

Sterpenich, V., Albouy, G., Boly, M., Vandewalle,G., Darsaud, A., et al. (2007). Sleep-relatedhippocampo-cortical interplay during emotionalmemory recollection. PLoS Biol., 5, e282.

Stickgold, R., Scott, L., Rittenhouse, C., & Hobson, J. A.(1999). Sleep-induced changes in associative mem-ory. J. Cogn. Neurosci., 11, 182–193.

Stickgold, R., & Walker, M. P. (2005a). Memory consoli-dation and reconsolidation: What is the role of sleep?Trends Neurosci., 28, 408–415.

Stickgold, R., & Walker, M. P. (2005b). Sleep andmemory: the ongoing debate. Sleep, 28, 1225–1227.

Strathern, P. (2000). Mendeleyev’s Dream: The Quest for the

Elements. London: Hamish Hamilton.Sullivan, G. M., Coplan, J. D., Kent, J. M., & Gorman,

J. M. (1999). The noradrenergic system in patho-logical anxiety: a focus on panic with relevance togeneralized anxiety and phobias. Biol. Psychiatry, 46 ,1205–1218.

Takashima, A., Petersson, K. M., Rutters, F., Tendolkar,I., Jensen, O., et al. (2006). Declarative memory con-solidation in humans: A prospective functional mag-netic resonance imaging study. Proc. Natl. Acad. Sci.

USA, 103, 756–761.Thurstone, L. L., & Thurstone, T. G. (1941). Factorial

studies of intelligence. Psychometric Monogr., 2, 94.Tononi, G., & Cirelli, C. (2003). Sleep and synaptic

homeostasis: a hypothesis. Brain Res. Bull., 62, 143–150.

Tononi, G., & Cirelli, C. (2006). Sleep function and synap-tic homeostasis. Sleep Med. Rev., 10, 49–62.

Tse, D., Langston, R. F., Kakeyama, M., Bethus, I.,Spooner, P. A., et al. (2007). Schemas and memoryconsolidation. Science, 316 , 76–82.

Tsuno, N., Besset, A., & Ritchie, K. (2005). Sleep anddepression. J. Clin. Psychiatry, 66 , 1254–1269.

Vazquez, J., & Baghdoyan, H. A. (2001). Basal forebrainacetylcholine release during REM sleep is signifi-cantly greater than during waking. Am. J. Physiol.

Regul. Integr. Comp. Physiol., 280, R598–601.Vertes, R. P., & Siegel, J. M. (2005). Time for the sleep

community to take a critical look at the purportedrole of sleep in memory processing. Sleep, 28, 1228–1229; discussion 1230–1233.

Wagner, U., Fischer, S., & Born, J. (2002). Changes inemotional responses to aversive pictures across peri-ods rich in slow-wave sleep versus rapid eye move-ment sleep. Psychosom. Med., 64, 627–634.

Wagner, U., Gais, S., & Born, J. (2001). Emotional mem-ory formation is enhanced across sleep intervals withhigh amounts of rapid eye movement sleep. Learn.

Mem., 8, 112–119.Wagner, U., Gais, S., Haider, H., Verleger, R., & Born, J.

(2004). Sleep inspires insight. Nature, 427, 352–355.Wagner, U., Hallschmid, M., Rasch, B., & Born, J. (2006).

Brief sleep after learning keeps emotional memoriesalive for years. Biol. Psychiatry, 60, 788–790.

Walker, E. L., & Tarte, R. D. (1963). Memory storage asa function of arousal and time with homogeneousand heterogeneous lists. J. Verbal Learn. Verbal Behav.,2, 113–119.

Walker, M. P. (2005). A refined model of sleep and thetime course of memory formation. Behav. Brain Sci.,28, 51–64.

Walker, M. P., Liston, C., Hobson, J. A., & Stickgold, R.(2002). Cognitive flexibility across the sleep-wake cy-cle: REM-sleep enhancement of anagram problemsolving. Brain Res. Cogn. Brain Res., 14, 317–324.

Walker, M. P., & Stickgold, R. (2006). Sleep, memory andplasticity. Annu. Rev. Psychol., 10, 139–166.

Werth, E., Achermann, P., Dijk, D. J., & Borbely, A. A.(1997). Spindle frequency activity in the sleep EEG:individual differences and topographic distribution.Electroencephalogr. Clin. Neurophysiol., 103, 535–542.


Wilson, M. A., & McNaughton, B. L. (1994). Reactivationof hippocampal ensemble memories during sleep.Science, 265, 676–679.

Wixted, J. T., & Carpenter, S. K. (2007). The Wickelgrenpower law and the Ebbinghaus savings function. Psy-

chol. Sci., 18, 133–134.Yoo, S. S., Gujar, N., Hu, P., Jolesz, F. A., & Walker,

M. P. (2007a). The human emotional brain withoutsleep – a prefrontal amygdala disconnect. Curr. Biol.,17, R877–R878.

Yoo, S. S., Hu, P. T., Gujar, N., Jolesz, F. A., & Walker, M.

P. (2007b). A deficit in the ability to form new hu-man memories without sleep. Nat. Neurosci., 10, 385–392.

Zeitlhofer, J., Gruber, G., Anderer, P., Asenbaum, S.,Schimicek, P., et al. (1997). Topographic distribu-tion of sleep spindles in young healthy subjects. J.

Sleep Res., 6 , 149–155.Zohar, D., Tzischinsky, O., Epstein, R., & Lavie, P. (2005).

The effects of sleep loss on medical residents’ emo-tional reactions to work events: a cognitive-energymodel. Sleep, 28, 47–54.

the role of sleep in cognition and emotion · the year in cognitive neuroscience 2009 the role of...

Documents