Sie sind auf Seite 1von 30


The Role of Sleep in Cognition and Emotion

Matthew P. Walker
Sleep and Neuroimaging Laboratory, Department of Psychology & Helen Wills
Neuroscience 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 specific
sleep-stage physiologies in the modulation, regulation, and even preparation of cog-
nitive and emotional brain processes. First, evidence for the role of sleep in memory
processing will be discussed, principally focusing on declarative memory. Second, at a
neural level several mechanistic models of sleep-dependent plasticity underlying these
effects will be reviewed, with a synthesis of these features offered that may explain the
ordered structure of sleep, and the orderly evolution of memory stages. Third, 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 individual
memory items, but, instead, their abstracted assimilation into a schema of generalized
knowledge. Fourth, the newly emerging benefit of sleep in regulating emotional brain
reactivity 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 vast amount of time this state takes from
the biggest mistake the evolutionary process has ever made.” our lives, we still lack any consensus function
Allan Rechtschaffen for sleep. In part, this is perhaps because sleep,
University of Chicago Sleep Laboratory like its counterpart wakefulness, may serve not
Smithsonian, November 1978 one but many functions, for brain and body
Centrally, sleep is a brain phenomenon, and
Introduction over the past 20 years, an exciting revival has
taken place within the neurosciences, one that
A perplexing question continues to elude focuses on the question of why we sleep, and
scientific judgment: “Why do we sleep?” In specifically targeting the role of sleep in a num-
accepting the utility of evolution, as candidly ber of cognitive and emotional processes. This
stated by pioneering sleep researcher Allan review aims to provide a synthesis of these re-
Rechtschaffen, sleep is likely to support a fun- cent findings in humans, with the goal of ex-
damental need of the organism. Yet, despite tracting consistent themes across domains of
brain function that appear to be regulated by
sleep. Providing a mechanistic foundation on
Address for correspondence: Matthew P. Walker, Sleep and which to consider these findings, the first sec-
Neuroimaging Laboratory, Department of Psychology & Helen tion of this chapter briefly summarizes the brain
Wills Neuroscience Institute, University of California, Berkeley,
California 94720-1650, USA. Voice: 510-642-5292; fax: 510-642-5293.
substrates 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.

Walker: Sleep in Cognition and Emotion 169


Sleep Stage

NREM-4 stage-2 N R E M

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

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

section explores the role of sleep in memory this organizing principal (deep NREM early in
and brain plasticity and also examines compet- the night, stage 2 NREM and REM late in
ing models of sleep-dependent learning. The the night) remains unknown–another perplex-
third section addresses the role of sleep beyond ing mystery of sleep.
memory consolidation, in processes of associa- As NREM sleep progresses, electroen-
tion, integration, and creativity. The final sec- cephalographic (EEG) activity begins to slow in
tion discusses the more recent and emerging frequency. Throughout stage-2 NREM there is
role for sleep in emotional and affective brain the presence of phasic electrical events includ-
regulation. ing K-complexes (large electrical sharp waves
in the EEG) and sleep spindles (short syn-
chronized 10–16 Hz) EEG oscillations (Steri-
Sleep Neurobiology ade & Amzica 1998). The deepest stages of
NREM, stages 3 and 4, are often grouped
The sleep of mammalian species has been together under the term “slow wave sleep”
broadly classified into two distinct types: non- (SWS), reflecting the occurrence of low fre-
rapid eye movement (NREM) sleep and rapid quency waves (1–4 Hz and <1 Hz), which have
eye movement (REM) sleep, with NREM sleep themselves been termed “slow-wave activity”
being further divided in primates and cats into (SWA), representing an expression of underly-
four substages (1–4) corresponding, in that or- ing mass cortical synchrony (Amzica & Steri-
der, to increasing depth of sleep (Rechtschaffen ade 1995). During REM sleep, however, EEG
& Kales 1968). In humans, NREM and REM waveforms once again change in their com-
sleep alternate or “cycle” across the night in position, associated with oscillatory activity in
an ultradian pattern every 90 min (Fig. 1). the theta band range (4–7 Hz), together with
Although this NREM−REM cycle length re- higher frequency synchronous activity in the
mains largely stable across the night, the ratio 30–80 Hz (“gamma”) range (Llinas & Ribary
of NREM to REM within each 90-min cycle 1993; Steriade et al. 1996). Periodic bursts of
changes, so that early in the night stages 3 and REM also take place, a defining characteristic
4 of NREM dominate, while stage 2 NREM of REM sleep, associated with the occurrence
and REM sleep prevail in the latter half of the of phasic endogenous waveforms expressed
night. Interestingly, the functional reasons for in, among other regions, the pons (P), lateral
170 Annals of the New York Academy of Sciences

geniculate nuclei of the thalamus (G), and the the modulation, regulation, and preparation of
occipital cortex (O), and as such, have been numerous brain functions.
termed “PGO waves” (Callaway et al. 1987).
As the brain passes through these sleep
stages, it also undergoes dramatic alterations Memory Processing
in neurochemistry. In NREM sleep subcorti- and Brain Plasticity
cal cholinergic systems in the brain stem and
forebrain become markedly less active (Hobson In considering the role of sleep in memory
et al. 1975; Lydic & Baghdoyan 1988), while processing, it is pertinent one appreciate that
firing rates of serotonergic raphé neurons and memories evolve (Walker & Stickgold 2006).
noradrenergic locus coeruleus neurons are also Specifically, memories pass through discrete
reduced relative to waking levels (Aston-Jones stages in their “life span.” The conception of
& Bloom 1981; Shima et al. 1986). During a memory begins with the process of encoding,
REM sleep both these aminergic populations resulting in a stored representation of an expe-
are strongly inhibited, while cholinergic sys- rience within the brain (Paller & Wagner 2002).
tems become as/more active compared to what However, it is now understood that a vast num-
they are during wake (Kametani & Kawamura ber of postencoding memory processes can take
1990; Marrosu et al. 1995), resulting in a brain place (Stickgold & Walker 2005a). For mem-
state largely devoid of aminergic modulation ories to persist over the longer time course of
and dominated by acetylcholine. minutes to years, an offline, nonconscious oper-
At a whole-brain systems level, neuroimag- ation of event consolidation appears to be nec-
ing techniques have revealed complex and essary, affording memories greater resistance
dramatically different patterns of functional to decay (a process of stabilization), or even im-
anatomy associated with NREM and REM proved recollection (a process of enhancement)
sleep (for review, see Nofzinger 2005). During (Robertson et al. 2004; Walker 2005). Sleep has
NREM SWS, rostral brain-stem regions, tha- been implicated in both the encoding and con-
lamic nuclei, basal ganglia, hypothalamus, pre- solidation of memory.
frontal cortex, cingulate corticies, and medial
regions of the temporal lobe all appear to un- Sleep and Memory Encoding
dergo reduced activity. However, during REM
sleep significant elevations in activity have been One of the earliest human studies to report
reported in the pontine tegmentum, thalamic the effects of sleep and sleep deprivation on
nuclei, occipital cortex, mediobasal prefrontal declarative memory encoding was by Morris
lobes, and associated limbic groups, including et al. (1960), indicating that “temporal mem-
the amygdala, hippocampus, and anterior cin- ory” (memory involving when events occur)
gulate cortex. In contrast, the dorso−lateral was significantly disrupted by a night of pre-
prefrontal cortex, posterior cingulate, and pari- training sleep loss. These findings have been
etal cortex appear least active in REM sleep. revisited in a more rigorous study by Harri-
Although this summary only begins to de- son and Horne (2000), again using the tem-
scribe the range of neural processes that are af- poral memory paradigm. Significant impair-
fected by the brain’s daily transit through sleep ments in retention were evident in a group of
states, it clearly demonstrates that sleep itself subjects deprived of sleep for 36 h, the sub-
cannot be treated as a homogeneous entity, one jects scoring significantly lower than controls,
which may or may not alter cognitive and emo- even in a subgroup that received caffeine to
tional processes. Instead, this constellation of overcome nonspecific effects of lower alert-
sleep stages offers a range of distinct neurobio- ness. Furthermore, the sleep-deprived subjects
logical mechanisms that can potentially support displayed significantly worse insight into their
Walker: Sleep in Cognition and Emotion 171

1.2 * ** Sleep 1.6
n.s. Sleep Deprived
Encoding Efficiency (d')


0.6 0.8


0.0 0.0
Figure 2. Sleep deprivation and encoding of emotional and nonemotional declarative memory. Effects
of 38 h of total sleep deprivation on encoding of human declarative memory (A) when combined across all
emotional 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 represent

memory-encoding performance, resulting in for emotional material. We have investigated

lower predictive ability of performance. the impact of sleep deprivation on the encod-
Pioneering work by Drummond and col- ing of emotionally negative, positive, and neu-
leagues examined the neural basis of sim- tral words (Walker, unpublished results). When
ilar memory impairments using functional combined across all stimulus types, subjects in
MRI (fMRI), investigating the effects of 35 h the sleep-deprived condition exhibited a strik-
of total sleep deprivation on verbal learning ing 40% reduction in the ability to form new
(Drummond et al. 2000). In those who were human memories under conditions of sleep
sleep deprived, regions of the medial temporal deprivation (Fig. 2A). However, when these
lobe were significantly less active during learn- data were separated into the three affective
ing, relative to a control group that had slept, categories (negative, positive, or neutral), the
while the prefrontal cortex actually expressed magnitude of encoding impairment differed
greater activation. Most interesting, the pari- (Fig. 2B). In those that had slept, both posi-
etal lobes, which were not activated in the con- tive and negative stimuli were associated with
trol group during learning, were significantly superior retention levels relative to the neutral
active in the deprivation group. Such findings condition, consistent with the notion that emo-
suggest that inadequate sleep prior to learn- tion facilitates memory encoding (Phelps 2004).
ing (at least following one night) produces bidi- However, there was severe disruption of en-
rectional changes in episodic encoding activity, coding and hence later retention for neutral
involving the inability of the medial temporal and especially positive emotional memory in
lobe to engage normally during learning, com- the sleep-deprived group. In contrast, a relative
bined with potential compensation attempts by resistance of negative emotional memory was
prefrontal regions, which in turn may facilitate observed in the deprivation group. These data
recruitment of parietal lobe function (Drum- suggest that, while the effects of sleep depriva-
mond & Brown 2001). tion are directionally consistent across memory
The impact of sleep deprivation on mem- subcategories, the most profound impact is on
ory formation may be especially pronounced the encoding of positive emotional stimuli, and
172 Annals of the New York Academy of Sciences

to a lesser degree, emotionally neutral stimuli. gyrus (IFG) displayed a significant positive, po-
In contrast, the encoding of negative memory tentially compensatory, relationship with mem-
appears to be more resistant to the effects of ory performance in those who were sleep de-
prior sleep loss, at least following one night. prived (Figs. 3B & C).
Intriguingly, these data may offer novel in- Taken together, this collection of findings
sights into affective mood disorders that express indicate the critical need for sleep-before-
co-occurring sleep abnormalities (Benca et al. learning in the preparation of key neural
1992; Buysse 2004). Indeed, if one compares structures for efficient next-day learning. With-
the profiles of memory encoding in Fig. 2B, out adequate sleep, hippocampal function
it is clear that those who slept encoded and becomes markedly disrupted, resulting in a de-
retained a balanced mix of both positive and creased ability for recording new experiences,
negative memories. In contrast, those who did the extent of which appears to be further gov-
not sleep displayed a skewed relative distribu- erned by alterations in prefrontal encoding
tion of encoding, resulting in an overriding dynamics.
dominance of negative memories, combined
with a retention deficit of positive and neutral
memories. This selective alteration in mem- Sleep and Memory Consolidation
ory encoding may provide an experimental Using a variety of behavioral paradigms, ev-
explanation for the higher incidence of depres- idence for the role of sleep in memory consoli-
sion in populations that suffer sleep disruption dation has now been reported across a diverse
(Shaffery et al. 2003; Buysse 2004), which, due range of phylogeny. Perhaps the earliest refer-
to these specific deficits, may impose a neg- ence to the beneficial impact of sleep on mem-
ative remembering bias, despite the fact that ory is by the Roman rhetorician Quintilian,
these subjects experienced equally positive- and who stated:
negative-reinforcing event histories.
[it] is a curious fact, of which the reason is not obvious, that
The impact of sleep deprivation on the neu- the interval of a single night will greatly increase the strength
ral dynamics associated with declarative mem- of the memory. . . . Whatever the cause, things which could
ory encoding has recently been examined using not be recalled on the spot are easily coordinated the next day,
event-related fMRI (Yoo et al. 2007a). In addi- and time itself, which is generally accounted one of the causes
tion to performance impairments under con- of forgetfulness, actually serves to strengthen the memory.
(Hammond 2004).
dition of sleep deprivation, and relative to a
control group that slept, a highly significant and In the early eighteenth and twentieth cen-
selective deficit was identified in bilateral re- turies respectively, David Hartley (Hartley
gions of the hippocampus—a structure known 1801) and Jenkins and Dallenback (Jenkins &
to be critical for learning new episodic infor- Dallenbach 1924) indicated that the strength of
mation (Eichenbaum 2004) (Fig. 3A). While a memory may be better preserved by periods
these findings indicated that, at a group level, of sleep than it is by equivalent periods of time
sleep deprivation markedly impairs hippocam- awake. Following the discovery of discrete sleep
pal memory function, when examined within stages (Aserinsky & Kleitman 1953), research
each group separately, the success of encod- investigating the influence of sleep on memory
ing, from low to high, was further associated has become gradually more complex at both
with activity in different regions of the pre- a behavioral and mechanistic level. A robust
frontal lobe. In those that slept prior to learn- and consistent literature has demonstrated the
ing, the right dorsal/middle lateral prefrontal need for sleep after learning in the subsequent
cortex showed a strong positive relationship consolidation and enhancement of procedural
with the proficiency of memory encoding. In memories; the evidence for which has recently
contrast, a region in the right inferior frontal been reviewed elsewhere (Walker & Stickgold
Walker: Sleep in Cognition and Emotion 173

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 group
in bilateral posterior hippocampus, together with a histogram of parameter estimates (effect
size) of averaged hippocampal activity in each group. Effects are significant at P < 0.001;
>5 contiguous voxels. (B) correlation analysis with memory performance showing regions
of 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 Yoo
et al. 2007a.

2006). Early work focusing on the role for sleep effect of sleep on the consolidation of declar-
in declarative memory processing was some- ative memory—our focus here (Smith 2001;
what less consistent, but more recent findings Ellenbogen et al. 2006b; Walker & Stickgold
have now begun to reveal a robust beneficial 2006; Marshall & Born 2007).
174 Annals of the New York Academy of Sciences

Several reports by Born and his colleagues *

showed offline improvement on a word-pair
association task following sleep, an improve- 100

ment attributed to early night sleep, rich in **

SWS (Plihal & Born 1997, 1999; Gais & Born
2004). More recently, the same group demon-
strated that, in addition to classically defined

Percent Recall
slow delta waves (0.5–4 Hz), the very slow corti- 60
cal oscillation (<1 Hz) appears to be important
for the consolidation of declarative memories.
Following subject learning of a word-pair list, 40
a technique called “direct current stimulation”
was used to induce slow oscillation-like field
potentials in the prefrontal cortex (in this case, 20
at 0.75 Hz) during early night SWS (Marshall
et al. 2006). Direct current stimulation not only
increased the amount of slow oscillations dur- 0
Wake Sleep Wake-I Sleep-I
ing the simulation period (and for some time
(no interference) (interference)
after), but also enhanced next-day word-pair
retention, suggesting a critical role for SWS
neurophysiology in the offline consolidation of Figure 4. Impact of sleep on the consolidation
and stabilization of declarative memory. Percent cor-
episodic facts. rect recall for B words from the original A-B pair
Rather than simply testing memory recall, after a 12-h retention interval of either wake or
Ellenbogen and colleagues have since revealed sleep following no interference or interference learn-
the extent of sleep’s ability to protect declar- ing (list A-C). † P < 0.10, ∗ P < 0.05, ∗∗ P < 0.001;
ative memories using experimentally induced error bars indicate SEM. Modified from Ellenbogen
learning disruption (Ellenbogen et al. 2006a).
Taking advantage of a classic interference tech-
nique called the A-B–A-C paradigm, subjects cantly more resistant to interference, whereas,
first learned unrelated word-paired associates, across a waking day, memories were far more
designated as list A-B (e.g., leaf-wheel, etc.). Af- susceptible to this antagonistic learning chal-
ter sleep at night, or wakefulness during the lenge. Yet it was only by using an interfering
day, half of the subjects in each group learned a challenge, that of the A-C list, that the true
new, interfering list containing a new associate benefit of sleep’s protection of memory was re-
paired with the first word, designated as list A-C vealed, a benefit that would not necessarily have
(e.g., leaf-nail, etc.), before being tested on the been evident in a standard study−test memory
original A-B list (e.g., leaf-wheel, etc.). In the paradigm.
groups that did not experience the interfering One mechanism proposed as underlying
challenge—that is, those who were simply be- these effects on hippocampal-dependent learn-
ing trained and then tested on list A-B—sleep ing tasks (see next section, also) is the reactiva-
provided a modest benefit to memory recollec- tion of memory representations at night. A con-
tion (Fig. 4A). However, when testing the groups siderable number of reports have investigated
that were exposed to interfering list learning (list the firing patterns of large networks of indi-
A-C) prior to recalling the original list (list A- vidual neurons across the wake−sleep cycle in
B), a large and significant protective benefit was animals. The signature firing patterns of these
seen in those that slept (Fig. 4B). Thus, mem- hippocampal and cortical networks, expressed
ories tested after a night of sleep were signifi- during waking performance of spatial tasks and
Walker: Sleep in Cognition and Emotion 175

novel experiences, appear to be “replayed” dur- ual declarative memories, and may indicate an
ing subsequent SWS (and in some studies, also active reprocessing of hippocampal-bound in-
REM) (Wilson & McNaughton 1994; Skaggs formation during SWS.
& McNaughton 1996; Dave et al. 1998; Dave
& Margoliash 2000; Poe et al. 2000; Louie & Models of Sleep-Dependent
Wilson 2001; Ribeiro et al. 2004; Jones & Wil- Memory Processing
son 2005; Ji & Wilson 2007). Homologous ev-
idence has been reported in the human brain Elucidating the neural mechanisms that
using a virtual maze task in combination with control and promote sleep-dependent human
positron emission tomography (PET) scanning memory consolidation remains an active topic
(Peigneux et al. 2004). Daytime learning was of research, and debate (Miller 2007). It is per-
initially associated with hippocampal activity. haps unlikely that multiple different memory
Then, during posttraining sleep, there was a systems, involving diverse cortical and/or sub-
reemergence of hippocampal activation, specif- cortical networks, require the same underlying
ically during SWS. Most compelling, however, neural mechanisms for their modulation. Even
was that the amount of SWS reactivation in the if they do, it is not clear that this process would
hippocampus was proportional to the amount rely on just one type of sleep-stage physiology
of next-day task improvement, suggesting that (Giuditta et al. 1995). At present, two intriguing
this reactivation is associated with offline mem- models of sleep-dependent plasticity, relevant
ory improvement. to declarative memory, have been offered to
Building on the framework that memories, account for the overnight facilitation of recall,
particularly those involving the hippocampus, which build on different aspects of neural activ-
are reactivated at night during sleep, Rasch ity during sleep: (1) hippocampal−neocortical
et al. have taken advantage of the classical dialogue, (2) synaptic homeostasis hypothesis.
psychology effect of cue-dependent recall, and
translated it into a sleep-dependent consolida- Hippocampal−Neocortical Dialogue
tion paradigm (Rasch et al. 2007). It is well There is considerable agreement that struc-
known that memory can be strongly modu- tures within the medial temporal lobe (MTL),
lated by smell (Cann & Ross 1989); most of us most notably the hippocampal complex, are
have associated the smell of a certain perfume crucial for the formation and retrieval of new
or cologne with a particular person, and when declarative memories. These structures are be-
we encounter that same perfume again, it often lieved to guide the reinstatement of recently
results in the powerful cued recall of memories formed memories by binding together pat-
of that particular person. In this study, however, terns of cortical activation that were present
following learning of a spatial memory task that at the time of initial learning. A classical model
was paired with the smell of rose, the odor was of declarative memory consolidation suggests
not re-presented at retrieval, but instead dur- that information initially requires MTL bind-
ing subsequent SWS that night—a time when ing, but over time, and by way of slow of-
consolidation was presumed to be occurring. fline processes, it is eventually integrated into
Relative to a control condition where the odor neocortical circuits (Fig. 5). Neocortical struc-
was not presented again during SWS, the re- tures thus become the eventual storage site for
perfusion of the rose scent at night resulted in consolidated episodic memories through cross-
significantly improved recall the following day. cortical connections, and, as a consequence,
Moreover, the re-presentation of the odor re- the MTL is not necessary for these memo-
sulted in greater (re)activation of the hippocam- ries’ retrieval. Therefore, the classical model
pus during SWS. These findings support the of memory consolidation holds that neocorti-
role of SWS in the consolidation of individ- cal structures become increasing important for
176 Annals of the New York Academy of Sciences

Figure 5. Model of sleep-dependent hippocampal-neocortical memory consolidation. At

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

the retention and retrieval of successfully con- sites, thereby reinforcing them (Fig. 5). Eventu-
solidated episodic memories, while the corre- ally, this strengthening would allow the original
sponding contribution of the hippocampus pro- information to be activated in the cortex, in-
gressively decreases (Squire 1992; McClelland dependent of the hippocampus. Buzsaki (1996)
et al. 1995; Squire & Zola 1996; Squire 2004). has since advanced on these ideas, proposing a
It should be noted, however, that controversy model of consolidation that involves two stages
remains about the role of these MTL structures or states of hippocampal activity, the first in-
in the retrieval of declarative memories after the volving a mode of “recording” during wake,
passage of time. This has led to the emergence which shifts to a second stage, involving “play-
of alternate consolidation models, most notably back” mode during NREM SWS, specifically
Nadel and Moscovitch’s “multiple trace the- during bursts of neural activity called “sharp-
ory” (Nadel & Moscovitch 1997; Moscovitch waves.”
& Nadel 1998), which posits that hippocampal Interestingly, these models make two pre-
involvement is always critical for the retrieval dictions about the impact of sleep on declar-
of episodic (but not semantic) memories, and ative memory. The first is that declarative
that these memories remain permanently de- memories from the day prior should be more
pendent on hippocampal−neocortical connec- resistant to interference the next day, due
tions (for discussion beyond the scope of this to the increased cortico−cortical connections
review see Frankl & Bontempi 2005). formed during overnight consolidation (Fig. 5).
In addition to its role in binding distributed It is precisely this behavioral effect that was
cortical memory components, Marr and, also, reported in the study by Ellenbogen et al.
McClelland et al. suggested that the hippocam- (2006a) showing greater post-sleep resistance
pus plays a critical role in reactivating these to interference, using the A-B–A-C paradigm.
networks, specifically during sleep (Marr 1970; A second and far less considered benefit of
McClelland et al. 1995). This process of re- this sleep-dependent dialogue is the encod-
activation, over multiple sleep cycles across a ing capacity of the hippocampus (Fig. 5). If
night and/or multiple occurrences of sleep over the strengthening of cortico−cortical connec-
many nights, would gradually strengthen the tions takes place during sleep, albeit itera-
initially weak connections between neocortical tively, then blocking sleep after hippocampal
Walker: Sleep in Cognition and Emotion 177

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) Correlation
between 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- increasing time following learning, there was
venting the development of independence from progressively greater recall activity in medial
(or “refreshing” of) the hippocampus, and by prefrontal regions, and a continued dissipa-
doing so, decrease the capacity for new hip- tion of retrieval-related activity within the hip-
pocampal learning the next day. This sec- pocampus. Advancing on these findings, Ma-
ond premise appears to accurately explain the quet and colleagues have since demonstrated
findings discussed in the section above on mem- that one night of posttraining sleep depriva-
ory encoding (Yoo et al. 2007b), which de- tion, even following recovery sleep, significantly
scribe a significant impairment of hippocam- impairs the normal modulation of hippocam-
pal encoding activity when sleep has not taken pal activity associated with episodic memory
place (through deprivation) being associated recollection (Gais et al. 2007). Furthermore,
with a decreased ability to form new episodic first-night sleep deprivation also prevented an
memories. increase in hippocampal connectivity with the
Two recent reports have provided further ev- medial prefrontal cortex, a development that
idence in support of this sleep-dependent dia- was only observed in those that slept after
logue and neural transformation of declarative learning.
memory. In the first such report, Takashima While no one study has yet demonstrated
and colleagues examined the benefit of day- that the neural signature of learning during the
time naps on episodic declarative memory con- day is subsequently reactivated and driven by
solidation (Takashima et al. 2006). In addi- characteristics of SWS at night, and that the
tion to a long-term evaluation of memory over extent of these properties are consequently pro-
3 months, there was also a short-term eval- portional to the degree of next-day recall and
uation of memory across the first day, which memory reorganization, collectively, they offer
included an intervening nap period (90 min) an empirical foundation on which to entertain
between training and testing of the original this possibility.
studied (“remote”) stimuli. Interestingly, the
duration of NREM SWS during the inter- Synaptic Homeostasis Hypothesis
vening nap correlated positively with later- In recent years an orthogonal theory of SWS
recognition memory performance (Fig. 6A), and learning has emerged, one which postu-
yet negatively with retrieval-related activity in lates a role for sleep in regulating the synap-
the hippocampus (Fig. 6B). Furthermore, with tic connectivity of the brain—principally the
178 Annals of the New York Academy of Sciences

Figure 7. Slow-wave activity and motor-skill memory. Topographical high-density EEG

maps of slow-wave activity (SWA) during NREM sleep following either (A) motor-skill learning
or (B) a nonlearning control condition, and (C) the subtracted difference between SWA in
the learning versus nonlearning condition, demonstrating a local homeostatic increase above
the learning-related central-parietal brain region. (D) the correlation between the amount of
over-night improvement on the task (measured the next day) and the extent of increase in
SWA across subjects. Modified from Huber et al. 2004.

neocortex (Tononi & Cirelli 2003, 2006). Their A number of human studies by Huber,
model considers NREM SWS, and specifically Tononi, and colleagues have provided evi-
the magnitude of slow-wave activity (SWA) of dence supporting their model. For example, it
SWS, as a brain-state that promotes the de- has been shown that learning of a motor-skill
crease of synaptic connections, not an increase. adaptation task during the day subsequently
Accordingly, plastic processes, such as learning triggers locally specific increases in cortical
and memory occurring during wakefulness, re- SWA at night, the extent of which is propor-
sult in a net increase in synaptic strength in tional to both the amount of initial daytime
diffuse brain circuits. The role of SWS, there- learning and the degree of next-day improve-
fore, and the slow oscillation in particular, is to ment (Fig. 7) (Huber et al. 2004). Further-
selectively downscale or “depotentiate” synap- more, experimentally impairing the amount of
tic strength back to baseline levels, preventing experience-dependent activity during the day
synaptic overpotentiation, which would result (arm immobilization) produced the opposite
in saturated brain plasticity. In doing so, this effect—reduced amounts of SWA activity in
rescaling would leave behind more efficient and associated cortical regions (Huber et al. 2006).
refined memory representations the next day, These findings substantiate the concept of lo-
affording improved recall. cal sleep-dependent neural pruning by SWS,
Walker: Sleep in Cognition and Emotion 179

the goal of which may be to regulate neural ar- A Role for Sleep Spindles?
chitecture at a highly specific anatomical level,
mapping onto corresponding locations of the Independent of both these models, which
memory representation. consider the role of SWS in memory pro-
How can the two concepts of neural reacti- cessing, a number of reports have also de-
vation (such as increased fMRI activity dur- scribed an association between learning and
ing SWS; e.g., Rasch et al. 2007) and neu- the hallmark feature of stage-2 NREM: sleep
ral homeostasis (such as increased slow wave spindles; short (∼1 s) synchronous bursts of
EEG activity; e.g., Huber et al. 2004) be inter- activity expressed in the EEG in the 10–
preted at a neural synaptic level? It could be 16 Hz frequency range (Sejnowski & Destexhe
argued that both these reported changes reflect 2000; Steriade 2001; Smith et al. 2004). For
either an increased neural reactivation, or an example, following learning of a motor-skill
increase in SWA associated with homeostasis memory task, Nishida and Walker (2007) ex-
and synaptic downscaling. Furthermore, both amined posttraining sleep spindles over the
hypotheses, while distinctly different in a mech- motor cortex, evaluating the difference in spin-
anistic sense, could offer complementary ben- dle activity in the learning hemisphere (since
efits at the network level in terms of signal- subjects perform the task with their left, non-
to-noise ratio (SNR) and may account for dominant, hand), relative to the nonlearning
overnight memory improvements (Robertson, hemisphere (Fig. 8A). Remarkably, when sleep
unpublished). Specifically, homeostatic synap- spindle power at electrode sites above the pri-
tic downscaling could result in the removal mary motor cortex of the nonlearning hemi-
of superfluous neural connections, resulting in sphere (left) were subtracted from those in the
improved SNR. However, neural reactivation learning hemisphere (right), representing the
and strengthening of experience-dependent within-subject, between-hemisphere difference
circuits, done without removing redundant in spindle activity following learning, a strong
synaptic connects, may equally improve SNR. predictive relationship to the amount of mem-
Therefore, both mechanisms, while different, ory improvement emerged (Figs. 8B & C). Simi-
could produce a similar outcome: enhanced fi- larly, Fogel et al. (2001) reported increased spin-
delity of the memory representation. Presum- dle density after intensive training on a pursuit
ably the combination of both would produce motor skill task, and Fogel and Smith (2006)
perhaps the most optimal and efficient mem- reported increased spindle density after com-
ory trace, yet a careful delineation of these pos- bined training on several simple procedural
sibilities remains an important goal for future motor tasks.
studies. Such findings indicate that the enhancement
Interestingly, the homeostasis model would of specific memory representations is associ-
also predict that sleep deprivation, specifically ated with electrophysiological events expressed
the prevention of SWS, would also negate effec- at local, anatomically discrete locations of the
tive next-day new learning, due to overpotenti- brain. Contrasting with the proposed impact
ation of synaptic connections. Thus, any region of SWS, the mechanistic benefit of sleep spin-
that exhibits SWA and is involved in represent- dles may be related to their faster stimulating
ing memory (e.g., hippocampus), would display frequency; a range suggested to facilitate long-
a corresponding inability to code further in- term potentiation (LTP; a foundational prin-
formation beyond a normal waking duration cipal of synaptic strengthening in the brain)
(∼16 hr in humans). Such a premise may of- (Sejnowski & Destexhe 2000; Steriade 2001;
fer an alternative explanation to the marked Smith et al. 2004), and not synaptic depres-
hippocampal encoding deficits reported under sion. This increase in spindle activity may rep-
conditions of sleep loss (Yoo et al. 2007b). resent a local, endogenous trigger of intrinsic
180 Annals of the New York Academy of Sciences

Figure 8. Sleep spindles and motor-skill memory plasticity. (A) Sleep-EEG array (blue discs) superimposed
on 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- (11–13 Hz), which have a more anterior dis-
graphically to the underlying memory repre- tribution, and “fast” (13–15 Hz), which have
sentation (Nishida & Walker 2007). a more posterior localization (Werth et al.
Increases in posttraining spindle activity are 1997; Zeitlhofer et al. 1997). With this di-
not limited to procedural memory tasks. For vision has also come an interest in under-
example, Gais et al. have shown that there standing functional differences between each
is significantly higher sleep spindle density in of these spindles subtypes. The relevance of
subjects that underwent a daytime episodic this separation from a memory consolidation
learning session (encoding of word-pair as- perspective is highlighted by a recent neu-
sociates) compared to a control group that roimaging study demonstrating that fast spin-
did not perform the learning session. More- dles are associated with, among other re-
over, the spindle density was associated with gions, significantly greater activation within the
the proficiency of memory recalled the next hippocampal complex (Schabus et al. 2007).
day in the learning group (Gais et al. 2002). Investigating the role of hippocampal- and
These findings mirror previous observations by extra−hippocampal-dependent memory con-
Meier-Koll et al. (1999), who reported a sim- solidation in relation to spindle subtypes will be
ilar increase in spindles following learning of an important challenge for future research.
a hippocampally dependent maze task, and by
Clemens et al. (2005), who have since identi- Reconciling Models
fied a correlation between spindle density and None of these models necessarily is wrong.
overnight verbal memory retention (although Instead, aspects of each may afford com-
not memory for faces). Intriguingly, the study plementary and synergistically beneficial out-
by Huber et al. (2004), which implicates slow comes for memory. Clues to this possibility lie
oscillatory activity associated with offline mem- within the ordered structure of human sleep
ory improvement, also describes similar, albeit (Fig. 1), with NREM SWS dominating early
near-significant, associations with activity in the in the night and stage-2 NREM and REM
sleep spindle frequency range. There may be prevailing later in the night. When placed
a combinatory role for spindles in regulating in this temporal framework a progression of
plasticity, together with SWS. events emerges that may be optimal for the
Continued evidence suggests that sleep spin- neuroplastic modulation of memory represen-
dles can be separated into two subtypes: “slow” tations. From a reactivation perspective, the
Walker: Sleep in Cognition and Emotion 181

predominance of hippocampal−neocortical predict improved recall of remodeled individ-

interaction would take place in the early SWS- ual memories from the prior day and further
rich phase of the night, leaving cortico−cortical afford the synaptic capacity for efficient acqui-
connections on offer for later processing dur- sition of new “information clay” the next day.
ing stage-2 NREM and REM. Similarly, and
even in coincidence, SWS may downscale cor-
tical (and possibly subcortical) plasticity, and Association, Integration,
it may do so in a learning-dependent man- and Creativity
ner, again leaving only those representations
(or aspects of these representations) which As critical as consolidation may be—an op-
are strongest—including those strengthened by eration classically concerned with individual
hippocampal−neocortical interplay—for pro- memory items—the association and integra-
cessing during these latter periods of sleep, tion of new experience into preexisting net-
dominated by faster frequency oscillations. works of knowledge is equally as important, if
This concept is analogous to the art of not more so. The resulting creation of associa-
sculpture. During the day, through experi- tive webs of information offers numerous and
ence, substantial informational “clay” is ac- powerful advantages. Indeed, the final goal of
quired on the cortical pedestal; some of it sleep-dependent memory processing may not
relevant, some not. Once accumulated, the be the enhancement of individual memories
brain’s next step is carving out and select- in isolation, but, instead, their integration into
ing the strongest and most salient memory a common schema, and by this enhancement,
representations (“statues”)—a mechanism that facilitation of the development of universal con-
SWS, occurring first and predominately early cepts, a process that forms the basis of general-
in the night, may be ideally suited for. Follow- ized knowledge and even creativity.
ing such downscaling and/or dynamic selection
of memory through translocation, the remain- Association and Integration
ing cortical representations—the rough outline
of the sculpted form—may finally be strength- Perhaps the earliest demonstration that sleep
ened by faster frequency oscillations, including may be involved in a form of memory general-
those of sleep spindles (and potentially PGO- ization was by Fenn et al. (2003). Utilizing an ar-
waveburst during REM; (Datta 2000; Datta tificial grammar task, subjects were trained and
et al. 2008), more associated with the potentia- later tested on their ability to transfer phono-
tion of synaptic connections, not their depoten- logical categories across different acoustic
tiation. This final step is akin to polishing and patterns. The task required forming new map-
improving the detailed features of the mem- pings from complex acoustical sounds to pre-
ory statue, which, in terms of computational existing linguistic categories, which then gener-
modeling, would offer improved SNR quality alized to new stimuli. As such, it involved both
within the system. Such a cooperative mech- a declarative process of forming specific mem-
anism, which appreciates the temporal order ories associated with the learned stimuli, to-
of the wake−sleep cycle (acquisition, followed gether with a procedural component involving
by postprocessing), and, within sleep, the ultra- mapping across the set of learned sounds that
dian pattern of sleep-stage progression across supports generalization to novel stimuli. During
a night (selection and removal, followed by the initial training session there was a signifi-
strengthening), would produce a network of cant improvement in recognition performance
stored information that is not only more effi- on the task. However, when retested after a
cient, but for those representations remaining, 12-h waking interval, performance had de-
more enhanced. Both these processes would cayed. Yet, if subjects were retested following
182 Annals of the New York Academy of Sciences

a night of sleep, this ability for memory gener- of associative distance (B > E pair). Despite all
alization was restored. Supporting this concept, groups achieving near identical premise-pair
Dumay and Gaskell (2007) also demonstrated retention after the offline delay (i.e., the build-
that sleep and not equivalent time awake ing block pair of the hierarchy), a striking disso-
can integrate related but novel phonemes into ciation was evident in the ability to make rela-
pre-existing, long-term lexical memory stores tional inference judgments. Subjects that were
overnight. tested soon after learning in the 20 min group
In a related study by Gomez and colleagues showed no evidence of inferential ability, per-
(Gomez et al. 2006), infants were exposed to forming at chance levels (Fig. 9A). In contrast,
“phrases” from an artificial language during the two 12 h groups displayed highly significant
a learning session—for example, phrases like relational memory development. Most remark-
“pel–wadim–jic”—until the infants became fa- able, however, was the observation that if the
miliar (as indexed by look responses). However, 12-h period contained a night of sleep, a near
these three syllable units had an embedded rule, 25% advantage in relational memory over the
which was that the first and last unit formed a 12 h across the day group was seen for the
relationship of nonadjacency; in this case, pel most distantly connected inferential judgment
predicts jic. The infants were then retested sev- (the B > E pair; Fig. 9A). Together, these find-
eral hours later, yet some infants took normally ings demonstrate that human memory integra-
scheduled naps, while others were scheduled at tion takes time to develop, requiring slow, of-
a time when they would not sleep after learn- fline associative processes. Furthermore, sleep
ing. At later testing, infants again heard the appears to preferentially facilitate this integra-
recordings, along with novel phrases in which tion by enhancing hierarchical memory bind-
the predictive relationship between the first and ing, biasing the development of the most dis-
last word was new. Infants who did not sleep tant/weak associative links amongst related yet
recognized the phrases they had learned ear- separate memory items (Fig. 9B). It is also inter-
lier, yet those who had slept demonstrated a esting to note a further advantage of this sleep-
generalization of the predictive relationship to dependent assimilation process. When it stores
new phrases, suggesting that the intervening individual premise-pairs (top row, Fig. 9B) the
process of sleep allowed the reinterpretation of size/number of items (“bits”) of information
prior experience, and supported the abstrac- the brain has to code is ten (A-B, B-C, C-
tion of commonalities—that is, the ability to D, D-E, E-F). However, when it the items are
detect a general pattern in new information. formed into a hierarchy, the informational load
Ellenbogen et al. (2007) have since tested is compressed, reduced by nearly 50% to just
this sleep-dependent hypothesis of integration six bits (A-B-C-D-E-F). Therefore, a supple-
by examining human relational memory—the mentary benefit of sleep-dependent memory
ability to generalize previously acquired associ- association may be the improved efficiency of
ations to novel situations. Participants initially memory storage, in addition to a more gener-
learned five “premise pairs” (A > B, B > C, alized representation.
C > D, D > E, E > F). Unknown to sub- Thus, the overnight strengthening and con-
jects, the pairs contained an embedded hier- solidation of individual item memories (re-
archy (A > B > C > D > E > F). Following viewed above), may not be the ultimate ob-
an offline delay of 20 min, 12 h across the day, jective of sleep-dependent memory processing,
or 12 h containing a night of sleep, knowledge especially when one considers that declarative
of this hierarchy was tested by examining re- (nonemotional) memories decay over the long-
lational judgments for novel “inference” pairs, term (Wixted & Carpenter 2007). It is then
either separated by one degree of associative interesting to speculate whether sleep serves
distance (B > D, C > E pairs) or by two degrees to facilitate two complementary objectives for
Walker: Sleep in Cognition and Emotion 183

* declarative memory, which span different time

A * courses. The first may be an initial process of
consolidating individual item (episodic) mem-
90 ories that are novel, which may occur in the
n.s. relative short term. Over a longer time course,
Percent Correct

80 however, and utilizing these recently consol-

idated item memories prior to their fading,
70 sleep may begin the process of the brain’s ex-
traction (of meaning) and abstraction (building
60 associational links with existing information),
thereby creating more adaptive semantic net-
(chance) 50
o o o o o o
works (McClelland et al. 1995; Squire 2007;
1 2 1 2 1 2
Inference pairs
Tse et al. 2007). Ultimately, individual item
20 min 12 hr wake 12 hr sleep
memories would no longer be necessary for the
goal that sleep is trying to achieve, and only
the conceptual meaning of such experiences
Immediate A B C D E F
B C D E would remain. Whether the subsequent loss of
item memories is passive or whether sleep plays
Wake C D E F
an active role in this process (Crick & Mitchi-
B C D E son 1983) remains to be examined, but this
1o inference is a testable hypothesis, i.e., forgetting (individ-
ual items) is the price we pay for remembering (general
Sleep A B C D E F
1o & 2 o inference

Figure 9. Sleep-dependent integration of Creativity

human relational memory. (A) Delayed inference
(associative) memory performance (% correct) in a One potential advantage of testing asso-
relational memory task following different offline ciative connections and building cross-linked
delays. Immediate testing after just a 20 min offline
systems of knowledge is creativity—the ability
delay, demonstrated a lack of any inferential ability
resulting in chance performance on both 1-degree to take existing pieces of information and com-
(first order) and 2-degree and 2-degree (second bine them in novel ways that lead to greater
order) associative judgments. Following a more understanding and offer new, advantageous be-
extended 12-h delay, across the day (Wake group), havioral repertoires. The link between creativ-
performance was significantly above chance across ity and sleep, especially dreaming, has long
both the one and two-degree inference judgments.
However, following an equivalent 12-h offline delay,
been a topic of intense speculation. Even sci-
but containing a night of sleep (Sleep group), entific examples of creativity occurring dur-
significantly better performance was expressed on ing sleep are common: from the dreams of
the more distant two-degree inference judgment both August Kekulé that led to the concep-
compared with the one-degree judgment. (B) A tion of a simple structure for benzene (Hubert
conceptual model of the effects of sleep on memory
1985) to those of Dmitry Mendeleyev that ini-
integration. Immediately after learning, the represen-
tation of each premise is constituted as the choice tiated the creation of the periodic table of
of one item over another (A > B, etc.), and these
premises are isolated from one another despite hav-
ing overlapping elements. After a 12-h period with ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
no sleep, the premise representations are partially premise representations are fully interleaved, support-
integrated by their overlapping elements, sufficient ing both first- and second-order transitive inferences.

to support first-order transitive inferences. How- P < 0.05; error bars indicate SEM. Modified from
ever, following a 12-h offline period with sleep, the Ellenbogen et al. 2007 and Eichenbaum 2007.
184 Annals of the New York Academy of Sciences

elements (Strathern 2000) to the late night A 1 1 4 4 9 4 9 4

dreaming of Otto Loewi that inspired the
experimental demonstration of neurochemi- 1 9 1 4 4 1 9
cal transmission (Mazzarello 2000).Quantita-
tive data have further demonstrated that solu- *
B 60%
tion performance on tests of cognitive flexibility

% Subjects gaining insight

using anagram word puzzles is more than 30%
better following awakenings from REM sleep
compared with NREM awakenings (Walker 40%
et al. 2002). Similarly, a study of semantic prim-
ing has demonstrated that, in contrast to the sit-
uation in waking, performance following REM 20%
sleep awakenings shows a greater priming effect
by weakly related words than by strong primes,
while strong priming exceeds weak priming in
NREM sleep (Stickgold et al. 1999), again in- WAKE WAKE SLEEP
dicating the highly associative properties of the (DAY) (NIGHT) (NIGHT)
REM sleep brain. Even the study of mental ac- Figure 10. Sleep-dependent production of cre-
tivity (dreams) from REM sleep indicates that ative insight. (A) Example of the Number Reduction
there is not a concrete episodic replay of day- Task. Subjects analyze digits in a 8-digit string of 1s,
time experiences, but instead, a much more as- 4s, and 9s, from left to right, using two rules: (1) If
sociative process of semantic integration during two digits are the same, respond with that digit. Thus,
starting from the left, the first two digits are both “1,”
sleep (Fosse et al. 2003). and hence the response (listed below and to the right
Yet the most striking experimental evidence of the second digit) is also “1.” (2) If two digits are
of sleep-inspired insight is arguably that re- different, respond with the remaining digit. Thus, hav-
ported by Wagner and colleagues (Wagner et al. ing produced the response “1,” this response and the
2004). Using a mathematical “number reduc- next digits are analyzed. Since they differ (“1” and
“4”), the next response is “the remaining digit, or
tion task” (Thurstone & Thurstone 1941), a
“9.” This response and the next digit, “4” also differ
process of sleep-dependent creative insight was (“9” and “4”) and so the next response is the remain-
elegantly demonstrated. Subjects analyzed and ing digit, “1.” The analysis is continued to the end,
worked through a series of 8-digit string prob- and the final response, “9” in this case, is the solu-
lems, using specific addition rules (Fig. 10A). tion to the problem. This final response is then entered
Following initial training, after various peri- as the answer the problem. However, on every trial,
the last three response digits (e.g., “4–1–9” in figure
ods of wake or sleep, subjects returned for an above) are the mirror image of the preceding three
additional series of trials. When retested after (i.e., “9–1–4”). As a result, the second response digit
a night of sleep, subjects solved the task, us- (circled “9”) always provides the answer to the prob-
ing this “standard” procedure, 16.5% faster. lem, resulting in a “shortcut” to solving the problem,
In contrast, subjects who did not sleep prior if the subject gains this hidden insight. (B) Percent-
age of subjects that gained insight into this hidden
to retesting averaged less than a 6% improve-
rule following an offline delay while awake across
ment. However, hidden in the construction of the day, awake across the night of following sleep
the task was a much simpler way to solve the across the night. ∗ P < 0.05. Modified from Wagner
problem. On every trial, the last three response et al. 2004.
digits (e.g., “4–1–9” in Fig. 10A) were the mirror
image of the preceding three (i.e., “9–1–4”). As the second response digit. Most dramatically,
a result, the second response digit always pro- nearly 60% of the subjects who slept for a night
vided the answer to the problem, and using such between training and retesting discovered this
“insight,” subjects could stop after producing shortcut the following morning (Fig. 10B). In
Walker: Sleep in Cognition and Emotion 185

contrast, no more than 25% of subjects in any Using a sleep restriction paradigm (5 h/night),
of four different control groups that did not Dinges et al. (1997) reported a progressive in-
sleep had this insight. Sleeping after exposure crease in emotional disturbance across a one-
to the problem therefore more than doubled week period on the basis of questionnaire
the likelihood of solving it (although it is inter- mood scales. In addition, subjective descrip-
esting to note that this insight was not present tions in participants’ daily journals also indi-
immediately following sleep, but took over 100 cated increasing complaints of emotional dif-
trials on average to emerge the next day). ficulties. Zohar et al. (2005) investigated the
In summary, substantial evidence now sug- effects of sleep disruption on emotional re-
gests that sleep serves a metalevel role in activity to daytime work events in medical
memory processing that moves far beyond the residents. Sleep loss was shown to amplify nega-
consolidation and strengthening of individual tive emotional consequences of disruptive day-
memories and, instead, aims to intelligently time events while blunting the positive benefit
assimilate and generalize these details offline. associated with rewarding or goal-enhancing
In doing so, sleep may offer the ability to test activities.
and build common informational schemas of Although these findings help to characterize
knowledge, providing for increasingly accurate the behavioral irregularities imposed by sleep
statistic predictions about the world and allow- loss, evidence for the role of sleep in regulat-
ing for the discovery of novel, even creative, ing our emotional brain is surprisingly scarce.
next-day solution insights. To date, only one such study has investigated
whether a lack of sleep inappropriately mod-
ulates human emotional brain reactivity (Yoo
Emotional Regulation et al. 2007a). Healthy young participants were
allowed to sleep normally prior to an fMRI
Despite substantial research focusing on the scanning session or were sleep deprived for
interaction between sleep and cognition, espe- one night (accumulating approximately 35 h
cially memory, the impact of sleep and sleep of total sleep loss). During scanning, subjects
loss on affective and emotional regulation has performed an affective stimulus viewing task in-
received more limited research attention. This volving the presentation of picture slides rang-
absence of investigation is perhaps surprising ing in a gradient from emotionally neutral to
considering that nearly all psychiatric and neu- increasingly negative and aversive.
rological mood disorders express co-occurring While both groups expressed significant
abnormalities of sleep, suggesting an intimate amygdala activation in response to increas-
relationship between sleep and emotion. Nev- ingly negative picture stimuli, those in the
ertheless, a number of recent studies evaluating sleep-deprivation condition exhibited a re-
subjective as well as objective measures of mood markable +60% greater magnitude of amyg-
and affect, combined with insights from clinical dala reactivity, relative to the control group
domains, offer an emerging understanding for (Figs. 11A & B). In addition to this increased
the critical role of sleep in regulating emotional intensity of activation, there was also a three-
brain function. fold increase in the extent of amygdala vol-
ume recruited in response to the aversive stim-
Affective Reactivity uli in the sleep-deprivation group (Fig. 11B).
Perhaps most interestingly, relative to the sleep
Together with impairments of attention and control group, there was a significant loss of
alertness, sleep deprivation is commonly asso- functional connectivity identified between the
ciated with increased subjective reports of ir- amygdala and the medial prefrontal cortex
ritability and affective volatility (Horne 1985). (mPFC) in those who were sleep deprived—a
186 Annals of the New York Academy of Sciences

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 control
groups, and (B) corresponding differences in intensity and volumetric extent of amygdala activation between
the two groups (average ± SEM. of left and right amygdala). (C) Depiction of associated changes in functional
connectivity between the medial prefrontal cortex (mPFC) and the amygdala. With sleep, the prefrontal lobe
was 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 resulting
in an overactive amygdala. ∗ P < 0.01; error bars indicate SEM. Modified from Yoo et al. 2007b.

region known to have strong inhibitory pro- (Figs. 11C & D). It would therefore appear that
jections and hence modulatory impact on the a night of sleep may “reset” the correct affec-
amygdala (Sotres-Bayon et al. 2004). In con- tive brain reactivity to next-day emotional chal-
trast, significantly greater connectivity in the lenges by maintaining functional integrity of
deprivation group was observed between the this mPFC−amygdala circuit and thus govern
amygdala and the autonomic-activating cen- appropriate behavioral repertoires (e.g., opti-
ters of the locus coeruleus. mal social judgments and rational decisions).
Thus, without sleep, an amplified hyper- Intriguingly, a similar pattern of anatomical
limbic reaction by the human amygdala was dysfunction has been implicated in a number
observed in response to negative emotional of psychiatric mood disorders that express co-
stimuli. Furthermore, this altered magnitude occurring sleep abnormalities (Davidson 2002;
of limbic activity is associated with a loss of Davidson et al. 2002; New et al. 2007), directly
functional connectivity with the mPFC in the raising the issues of whether such factors (sleep
sleep deprivation condition implying a failure loss and clinical mood disorders) are causally
of top-down inhibition by the prefrontal lobe related.
Walker: Sleep in Cognition and Emotion 187

Emotional Information Processing press). Subjects performed two study sessions

in which they learned emotionally negative
Sleep’s role in declarative memory consoli- and neutral picture stimuli: one 4 h prior to
dation, rather than being absolute, may depend a recognition memory test, and one 15 min
on more intricate aspects of the information be- prior to it. In one group, participants slept
ing learned, such as novelty, meaning to extract, (90 min nap) after the first study session, while in
and also the affective salience of the material. the other group, participants remained awake.
Independent of the field of sleep and memory, Thus, items from the study sessions tested after
there is a wealth of evidence demonstrating that 4 h transitioned through different brain-states
memory processing is modulated by emotion in each group prior to testing—sleep in the Nap
(Cahill 2000; McGaugh 2004; Phelps 2004). group and no sleep in the No-Nap group—
Experiences which evoke emotions not only en- yet experienced identical brain-state conditions
code more strongly, but appear to persist and when tested after 15 min.
even improve over time as the delay between No change in memory for emotional (or
learning and testing increases (from hours to neutral stimuli), occurred across the offline
days) (Kleinsmith & Kaplan 1963; Walker & delay in the no-nap group. However, a sig-
Tarte 1963; Levonian 1972; LaBar & Phelps nificant and selective offline enhancement of
1998; Sharot & Phelps 2004). emotional memory was observed in the nap
Although these findings indicate a strong in- group (Fig. 12A), the extent of which was corre-
fluence of emotion on slow, time-dependent lated with the amount of REM sleep (Fig. 12B),
consolidation processes, based on the coinci- and the speed of entry into REM (latency; not
dent neurophysiology that REM sleep pro- shown in figure). Furthermore, spectral analysis
vides and the neurobiological requirements of of the EEG demonstrated that the magnitude
emotional memory processing (Cahill 2000; of right-dominant prefrontal theta power dur-
McGaugh 2004), work has now begun to test a ing REM (activity in the frequency range of
selective REM-dependent hypothesis of affec- 4.0–7.0 Hz) exhibited a significant and posi-
tive human memory consolidation. For exam- tive relationship with the amount of emotional
ple, Hu et al. (2006) compared the consolida- memory improvement (Figs. 12C & D).
tion of emotionally arousing and nonarousing These findings go beyond demonstrating
picture-stimuli following a 12-h period across that affective memories are preferentially en-
a day or following a night of sleep. A specific hanced across periods of sleep, and indicate
emotional memory benefit was observed only that the extent of emotional memory im-
following sleep and not cross an equivalent time provement is associated with specific REM
awake. Wagner and colleagues (Wagner et al. sleep characteristics—both quantity and qual-
2001) have also shown that sleep selectively fa- ity. Corroborating these correlations, it has pre-
vors the retention of previously learned emo- viously been hypothesized that REM sleep rep-
tional texts relative to neutral texts, and that resents a brain state particularly amenable to
this affective memory benefit is only present emotional memory consolidation, based on its
following late-night sleep (a time period rich in unique biology (Pare et al. 2002; Hu et al.
stage-2 NREM and REM sleep). Furthermore, 2006). Neurochemically, levels of limbic and
this emotional memory enhancement has been forebrain ACh are markedly elevated during
shown to persist for several years (Wagner et al. REM (Vazquez & Baghdoyan 2001), report-
2006). edly quadruple those seen during NREM and
Using a nap paradigm, researchers have double those measured in quite waking (Mar-
most recently demonstrated that sleep, and rosu et al. 1995). Considering the known im-
specifically REM neurophysiology, may under- portance of ACh in the long-term consolidation
lie this consolidation benefit (Nishida et al. in of emotional learning (McGaugh 2004), this
188 Annals of the New York Academy of Sciences

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 a
90 min nap (sleep, filled bar) (B) Correlation between the amount of offline emotional memory improvement
in the nap group (i.e. the offline benefit expressed in filled bar of figure A), and the amount of REM sleep
obtained within the nap, (C) Correlation strength (Pearson’s r -value) between offline benefit for emotional
memory in the sleep group (the benefit expressed in filled bar of figure A) and the relative right versus left
prefrontal spectral-band power ([F4 – F3]) within the delta, alpha, theta, and beta spectral bands, expressed
in average 0.5 Hz bins increments. Correlation strength is represented by the color range, demonstrating
significant correlations within the theta frequency band (hot colors), and (D) exhibiting a maximum significance
at 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- carrier frequency allowing disparate brain re-
lective facilitation of affective memories, similar gions that initially encode information to se-
to that reported using experimental manipula- lectively interact offline, in a coupled relation-
tions of ACh (Power 2004). Neurophysiologi- ship. By doing so, REM theta may afford the
cally, theta oscillations have been proposed as a ability to promote the strengthening of specific
Walker: Sleep in Cognition and Emotion 189

memory representations across distributed net- pants were reexposed to these same stimuli
works (Buzsaki 2002; Jones & Wilson 2005). during recognition testing many months later,
a change in the profile of activation occurred
A REM Sleep Hypothesis of Emotional (Dolcos et al. 2005). Although the same magni-
Memory Processing tude of differential activity between emotional
and neutral items was observed in the hip-
Beyond the strengthening of emotional pocampus, this was not true in the amygdala.
memories, there may be an additional conse- Instead, the difference in amygdala (re)activity
quence of sleep-dependent affective modula- compared with neutral items had dissipated
tion, and one that has significant implications over time. This would support the idea that
for mood disorders – that is, sleeping to for- the strength of the memory (hippocampal-
get. Based on the emerging interaction between associated activity) remains at later recollection,
sleep and emotion, below I outline a model of yet the associated emotional reactivity to these
affective information processing that may of- items (amygdala activity) is reduced over time.
fer brain-based explanatory insights regarding The hypothesis predicts that this decoupling
the impact of sleep abnormalities, particularly preferentially takes place overnight, such that
REM, for the initiation and/or maintenance of we sleep to forget the emotional tone, yet
mood disturbance. sleep to remember the tagged memory of that
While there is abundant evidence to sug- episode (Fig. 13). The model further argues
gest that emotional experiences persist in our that if this process is not achieved, the mag-
autobiographies over time, an equally remark- nitude of visceral autonomic “charge” remain-
able but far less noted change is a reduction in ing within autobiographical memory networks
the affective tone associated with their re- will persist, resulting in the potential condi-
call. The reason that affective experiences ap- tion of chronic anxiety. Based on the con-
pear to be encoded and consolidated more sistent relationship identified between REM
preferentially than neutral memories is due and emotional processing, combined with its
to autonomic neurochemical reactions elicited unique neurobiology, the hypothesis proposes
at the time of the experience, creating what that REM sleep provides an optimal biolog-
we commonly term an “emotional–memory.” ical state for achieving such affective “ther-
However, the later recall of these experi- apy.” Specifically, increased activity within lim-
ences tends not to be associated with any- bic and paralimbic structures (including the
where near the same magnitude of autonomic hippocampus and amygdala) during REM may
(re)activation as that elicited at the moment of first offer the ability for reactivation of previ-
learning/experience – suggesting that, over- ously acquired affective experiences. Second,
time, the affective “blanket” previously en- the neurophysiological signature of REM in-
veloped around the memory during encoding volving dominant theta oscillations within sub-
has been removed, while the information con- cortical as well as cortical nodes may offer
tained within that experience (the memory) re- large-scale network cooperation at night, al-
mains. lowing the integration and, as a consequence,
For example, neuroimaging studies have greater understanding of recently experienced
shown that initial exposure and learning emotional events in the context of preexisting
of emotional stimuli is associated with sub- neocortically stored semantic memory. Third,
stantially greater activation in the amygdala these interactions during REM critically, and
and hippocampus, relative to neutral stimuli perhaps most importantly, take place within a
(Kilpatrick & Cahill 2003; Dolcos et al. 2004; brain that is devoid of aminergic neurochemical
Dolcos et al. 2005). In one of these studies concentration (Pace-Schott & Hobson 2002),
(Dolcos et al. 2004), however, when partici- particularly noradrenergic input from the
190 Annals of the New York Academy of Sciences

Figure 13. Model of sleep-dependent emotional–memory processing: A sleep to forget

and 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 improved
memory for that event, the autonomic tone “enveloped” around the memory becomes grad-
ually ameliorated (emotional forgetting). Over time, the stored information of the original
experience ultimately becomes decoupled and freed of its autonomic “charge,” leaving just
the salient memory of that emotional experience, but without the affective tone previously
associated at the time of learning.

locus coeruleus; the influence of which has previous emotional conflict, resulting in im-
been linked to states of high stress and anxiety proved next-day negative mood. Moreover,
disorders (Sullivan et al. 1999). pioneering work by Cartwright et al. have
In summary, the described neuroanatomical, demonstrated that not only the occurrence of
neurophysiological and neurochemical condi- dreaming but the actual content of dreams
tions of REM sleep offer a unique biological plays an important role in the recovery from
theatre in which to achieve, on one hand, a bal- emotional trauma, and can be predictive of
anced neural potentiation of the informational clinical remission months later (Cartwright
core of emotional experiences (the memory), et al. 1991, 1998, 2006). Although the cur-
yet may also depotentiate and ultimately ame- rent model offers a neurobiological framework
liorate the autonomic arousing charge orig- for the overnight modulation and alteration
inally acquired at the time of learning (the of emotional memories and next-day affective
emotion), negating a long-term state of anxi- brain reactivity, is does not discount the poten-
ety (Fig. 13). tial contribution that the mental operation of
This model compliments previous psycho- dreaming itself, beyond the electrophysiologi-
logical theories of dreaming by Greenberg cal underpinnings of REM, may afford to this
(Greenberg et al. 1972a, 1972b) and also process.
Cartwright (Cartwright et al. 1991, 1998, The current model also asserts that if the
2006), which suggest that the process of REM- process of divorcing emotion from memory is
sleep mental activity aids in the resolution of not achieved across the first night following
Walker: Sleep in Cognition and Emotion 191

an emotional event, a repeat attempt would penich et al. 2007). The investigation demon-
occur on the second night, since the strength of strated that subjects who were deprived of sleep
emotional “tag” associated with the memory the first night after learning arousing emotion
would remain high. Should this process fail a picture slides not only showed reduced recalled
second time, the same events would continue of the information (the sleep to remember com-
to repeat across ensuing nights, potentially with ponent of the hypothesis), but also showed a
an increasing progressive amount of REM in lack of reduction in amygdala reactivity when
response. It is just such a cycle of REM-sleep reexposed to these same emotional picture slide
dreaming (nightmares) that represents a diag- at recognition testing – as compared to a con-
nostic key feature of post-traumatic stress dis- trol group that did sleep (the sleep to forget
order (PTSD) (Lavie 2001). It may not be coin- component of the hypothesis).
cidental, therefore, that these patients continue There is, however, a related study that does
to display hyperarousal reactions to associated not conform to these trends. Wagner and col-
trauma cues (Harvey et al. 2003; Pole 2007), leagues had participants subjectively rate and
indicating that the process of separating the af- re-rate emotional picture slides after 3 h of
fective tone from the emotional experience has early-night sleep, or 3 h of late-night sleep
not been accomplished. The reason why such (Wagner et al. 2002). Valence ratings of un-
a REM mechanism may fail in PTSD remains pleasantness actually increased following sleep,
unknown, although the exceptional magnitude compared to new picture slides not seen be-
of trauma-induced emotion at the time of learn- fore. This was true in a group that was also
ing maybe so great that the system is incapable allowed to sleep the entire night. The contrast
of initiating/completing one or both these pro- in this finding to those discussed previously re-
cesses, leaving some patients unable to depo- porting a decrease in emotional reactivity re-
tentiate, integrate and hence “overcome” the mains unclear. It is of note that when com-
experience. pared not to new items, but to the same items
This model also makes specific experimen- before sleep (baseline ratings), no increase in
tal predictions as to the fate of these two valance rating was evident. This discrepancy
components – the memory and the emotion. may also be due to the dimension of valence
As partially demonstrated, the first prediction not assessing autonomic emotional reactivity.
would be that, overtime, the veracity of the In fact, subjects also rated these pictures stim-
memory itself would improve, and the extent uli on the basis of arousal strength, as well as
to which these [negative] emotional experi- valence. There was no such amplification of
ences are strengthened would be proportional arousal reactivity following sleep, demonstrat-
to the amount of postexperience REM sleep ing a numerical decrease overtime, relative to
obtained, as well as how quickly it is achieved baseline measures. Alternatively, it may be that
(REM latency). the assessment of valence is associated with
Secondly, using autonomic physiology mea- the veracity of the memory, which is strength-
sures, these same predictions would hold in the ened overnight, thereby promoting the recall
inverse direction for the magnitude of emo- of perceived pleasantness (or unpleasantness).
tional reactivity induced at the time of re- Arousal (the visceral, autonomic dimension of
call. Together with the neuroimaging studies emotion), in contrast, appears to be reduced
of emotional memory recall over time, and the over sleep, despite the beneficial strengthening
psychological studies investigating the role of of memory recall.
REM sleep dreaming in mood regulation, a re- The model thirdly predicts that a patho-
cent fMRI study offers perhaps the strongest logical increase in REM (as commonly oc-
preliminary support of this sleep-dependent curs in depression; (Tsuno et al. 2005; Ar-
model of emotional-memory processing (Ster- mitage 2007; Gottesmann & Gottesman 2007)
192 Annals of the New York Academy of Sciences

may disproportionately amplify the strength of an exciting role for sleep in regulating affec-
negative memories, so much so that, despite tive brain function and emotional experience.
concomitant attempts at ameliorating the asso- Based on the remarkable neurobiology of sleep,
ciated affective tone, would still create a per- and REM in particular, a unique capability for
ceived autobiographical history dominated by the overnight modulation of affective networks
negative memory excess (which may also fa- and previously encountered emotional experi-
cilitate disadvantageous waking rumination). ences may be possible, redressing and main-
In contrast, the selective decrease of REM, taining the appropriate connectivity and hence
as occurs with many antidepressants, would next-day reactivity throughout limbic and as-
predict a reduction of such negative memory sociated autonomic systems.
consolidation and bias, although may curtail Ultimately, the timeless maternal wisdom of
the degree of affect decoupling that can oc- mothers may have long held the answers to
cur. Long-term, the balanced extent of accu- Allan Rechtschaffen’s original question “Why
mulated REM should therefore correlate not do we sleep?” Namely, that “you should sleep
only with the persistence, in memory, of the on a problem” and when troubled, “get to bed,
emotional experience, it should also be associ- you’ll feel better in the morning.”
ated with a decreased magnitude of autonomic
response associated with recall – all of which Acknowledgments
are testable objectives for future research.
If such a hypothesis is correct, there would The author wishes to thank Edwin Robert-
be translation implications for psychiatric and son, Robert Stickgold, Allison Harvey, Ninad
mood disorders. This would require a new ap- Gujar, and Els van der Helm for thoughtful
preciation for the palliative role of sleep in treat- insights. This work was supported in part by
ment regimes, and a consideration of whether grants from the National Institutes of Health
altering sleep architecture to regulation the bal- (MH069935) and the American Academy of
ance of emotion and memory of past experi- Sleep Medicine.
ence is a useful and viable possibility.
Conflicts of Interest

Conclusions The author declares no conflicts of interest.

While not fully complete, we will soon have References

a new taxonomy of sleep-dependent memory
processing, and one that will supersede the po- Amzica, F., & Steriade, M. (1995). Short- and long-range
larized all-or-none views of the past (Stickgold neuronal synchronization of the slow (<1 Hz) corti-
& Walker 2005b; Vertes & Siegel 2005). With cal oscillation. J. Neurophysiol., 73, 20–38.
such findings, we can come to a revised appre- Armitage, R. (2007). Sleep and circadian rhythms in
ciation of how both wake and sleep unite in a mood disorders. Acta Psychiatr. Scand. Suppl., 104–115.
Aserinsky, E., & Kleitman, N. (1953). Regularly occurring
symbiotic alliance to coordinate the encoding, periods of eye motility and concurrent phenomena
consolidation and integration of our memories, during sleep. Science, 118, 273–274.
the ultimate aim of which maybe to create a Aston-Jones, G., & Bloom, F. E. (1981). Activity of
generalized catalogue of stored knowledge that norepinephrine-containing locus coeruleus neurons
does not rely on the verbose retention of all in behaving rats anticipates fluctuations in the sleep-
waking cycle. J. Neurosci., 1, 876–886.
previously learned facts.
Benca, R. M., Obermeyer, W. H., Thisted, R. A., & Gillin,
Beyond memory and plasticity, a growing J. C. (1992). Sleep and psychiatric disorders. A meta-
number of human neuroscience studies, set analysis. Arch. Gen. Psychiatry, 49, 651–668; discussion
on a foundation of clinical insights, point to 669–670.
Walker: Sleep in Cognition and Emotion 193

Buysse, D. J. (2004). Insomnia, depression and aging. As- J. W., et al. (1997). Cumulative sleepiness, mood dis-
sessing sleep and mood interactions in older adults. turbance, and psychomotor vigilance performance
Geriatrics, 59, 47–51; quiz 52. decrements during a week of sleep restricted to 4–5
Buzsaki, G. (1996). The hippocampo-neocortical dia- hours per night. Sleep, 20, 267–277.
logue. Cereb. Cortex, 6, 81–92. Dolcos, F., LaBar, K. S., & Cabeza, R. (2004). Interac-
Buzsaki, G. (2002). Theta oscillations in the hippocampus. tion between the amygdala and the medial tempo-
Neuron, 33, 325–340. ral lobe memory system predicts better memory for
Cahill, L. (2000). Neurobiological mechanisms of emo- emotional events. Neuron, 42, 855–863.
tionally influenced, long-term memory. Prog. Brain Dolcos, F., LaBar, K. S., & Cabeza, R. (2005). Remem-
Res., 126, 29–37. bering one year later: role of the amygdala and the
Callaway, C. W., Lydic, R., Baghdoyan, H. A., & Hobson, medial temporal lobe memory system in retrieving
J. A. (1987). Pontogeniculooccipital waves: Sponta- emotional memories. Proc. Natl. Acad. Sci. USA 102,
neous visual system activity during rapid eye move- 2626–2631.
ment sleep. Cell. Mol. Neurobiol., 7, 105–149. Drummond, S. P., & Brown, G. G. (2001). The effects of
Cann, A., & Ross, D. A. (1989). Olfactory stimuli as con- total sleep deprivation on cerebral responses to cog-
text cues in human memory. Am. J. Psychol., 102, nitive performance. Neuropsychopharmacol. 25, S68–
91–102. 73.
Cartwright, R., Agargun, M. Y., Kirkby, J., & Friedman, Drummond, S. P., Brown, G. G., Gillin, J. C., Stricker, J.
J. K. (2006). Relation of dreams to waking concerns. L., Wong, E. C., et al. (2000). Altered brain response
Psychiatry Res., 141, 261–270. to verbal learning following sleep deprivation. Nature
Cartwright, R., Luten, A., Young, M., Mercer, P., & Bears, 403, 655–657.
M. (1998). Role of REM sleep and dream affect Dumay, N., & Gaskell, M. G. (2007). Sleep-associated
in overnight mood regulation: a study of normal changes in the mental representation of spoken
volunteers. Psychiatry Res., 81, 1–8. words. Psychol. Sci. 18, 35–39.
Cartwright, R. D., Kravitz, H. M., Eastman, C. I., & Eichenbaum, H. (2004). Hippocampus: cognitive pro-
Wood, E. (1991). REM latency and the recovery cesses and neural representations that underlie
from depression: getting over divorce. Am. J. Psychia- declarative memory. Neuron 44, 109–120.
try, 148, 1530–1535. Eichenbaum, H. (2007). To sleep, perchance to integrate.
Clemens, Z., Fabo, D., & Halasz, P. (2005). Overnight Proc. Natl. Acad. Sci. USA, 104, 7317–7318.
verbal memory retention correlates with the number Ellenbogen, J. M., Hulbert, J. C., Stickgold, R., Dinges,
of sleep spindles. Neuroscience, 132, 529–535. D. F., & Thompson-Schill, S. L. (2006a). Interfering
Crick, F., & Mitchison, G. (1983). The function of dream with theories of sleep and memory: sleep, declarative
sleep. Nature, 304, 111–114. memory, and associative interference. Curr. Biol. 16,
Datta, S. (2000). Avoidance task training potentiates pha- 1290–1294.
sic pontine-wave density in the rat: A mechanism Ellenbogen, J. M., Payne, J. D., & Stickgold, R. (2006b).
for sleep-dependent plasticity. J. Neurosci., 20, 8607– The role of sleep in declarative memory consolida-
8613. tion: passive, permissive, active or none? Curr. Opin.
Datta, S., Li, G., & Auerbach, S. (2008). Activation of Neurobiol., 16, 716–722.
phasic pontine-wave generator in the rat: a mech- Ellenbogen, J., Hu, P., Payne, J.D., Titone, D. & Walker,
anism for expression of plasticity-related genes and M.P. (2007). Human relational memory requires
proteins in the dorsal hippocampus and amygdala. time and sleep. Proc. Natl. Acad. Sci. USA, 104, 7723–
Eur J. Neurosci., 27, 1876–1892. 7728.
Dave, A. S., & Margoliash, D. (2000). Song replay during Fenn, K. M., Nusbaum, H. C., & Margoliash, D. (2003).
sleep and computational rules for sensorimotor vocal Consolidation during sleep of perceptual learning of
learning. Science, 290, 812–816. spoken language. Nature, 425, 614–616.
Dave, A. S., Yu, A. C., & Margoliash, D. (1998). Behav- Fogel, S., Jacob, J., & Smith, C. (2001). Increased sleep
ioral state modulation of auditory activity in a vocal spindle activity following simple motor procedural
motor system. Science, 282, 2250–2254. learning in humans. Congress Physiological Basis
Davidson, R. J. (2002). Anxiety and affective style: role of for Sleep Medicine, pp. 123.
prefrontal cortex and amygdala. Biol. Psychiatry, 51, Fogel, S. M., & Smith, C. T. (2006). Learning-dependent
68–80. changes in sleep spindles and Stage 2 sleep. J. Sleep
Davidson, R. J., Pizzagalli, D., Nitschke, J. B., & Putnam, Res., 15, 250–255.
K. (2002). Depression: perspectives from affective Fosse, M. J., Fosse, R., Hobson, J. A., & Stickgold, R. J.
neuroscience. Annu. Rev. Psychol., 53, 545–574. (2003). Dreaming and episodic memory: a functional
Dinges, D. F., Pack, F., Williams, K., Gillen, K. A., Powell, dissociation? J. Cogn. Neurosci., 15, 1–9.
194 Annals of the New York Academy of Sciences

Frankland, P. W., & Bontempi, B. (2005). The organiza- Huber, R., Ghilardi, M. F., Massimini, M., & Tononi, G.
tion of recent and remote memories. Nat. Rev. Neu- (2004). Local sleep and learning. Nature, 430, 78–81.
rosci., 6, 119–130. Hubert, A. (1985). “Kekulé von Stradonitz, Friedrich August.”
Gais, S., Albouy, G., Boly, M., Dang-Vu, T. T., Darsaud, New York: Macmillan.
A., et al. (2007). Sleep transforms the cerebral trace Jenkins, J. G., & Dallenbach, K. M. (1924). Obliviscence
of declarative memories. Proc. Natl. Acad. Sci. USA, during sleep and waking. Am. J. Psychol., 35, 605–
104, 18778–18783. 12.
Gais, S., & Born, J. (2004). Low acetylcholine during slow- Ji, D., & Wilson, M. A. (2007). Coordinated memory re-
wave sleep is critical for declarative memory consol- play in the visual cortex and hippocampus during
idation. Proc. Natl. Acad. Sci. USA, 101, 2140–2144. sleep. Nat. Neurosci., 10, 100–107.
Gais, S., Molle, M., Helms, K., & Born, J. (2002). Jones, M. W., & Wilson, M. A. (2005). Theta rhythms
Learning-dependent increases in sleep spindle den- coordinate hippocampal-prefrontal interactions in a
sity. J. Neurosci., 22, 6830–6834. spatial memory task. PLoS Biol., 3, e402.
Giuditta, A., Ambrosini, M. V., Montagnese, P., Mandile, Kametani, H., & Kawamura, H. (1990). Alterations in
P., Cotugno, M., et al. (1995). The sequential hy- acetylcholine release in the rat hippocampus during
pothesis of the function of sleep. Behav. Brain Res., 69, sleep-wakefulness detected by intracerebral dialysis.
157–166. Life Sci., 47, 421–426.
Gomez, R. L., Bootzin, R. R., & Nadel, L. (2006). Naps Kilpatrick, L., & Cahill, L. (2003). Amygdala modulation
promote abstraction in language-learning infants. of parahippocampal and frontal regions during emo-
Psychol. Sci., 17, 670–674. tionally influenced memory storage. Neuroimage, 20,
Gottesmann, C., & Gottesman, I. (2007). The neurobio- 2091–2099.
logical characteristics of rapid eye movement (REM) Kleinsmith, L. J., & Kaplan, S. (1963). Paired-associate
sleep are candidate endophenotypes of depression, learning as a function of arousal and interpolated
schizophrenia, mental retardation and dementia. interval. J. Exp. Psychol., 65, 190–193.
Prog. Neurobiol., 81, 237–250. LaBar, K. S., & Phelps, E. A. (1998). Arousal-mediated
Greenberg, R., Pearlman, C. A., & Gampel, D. (1972a). memory consolidation: Role of the medial temporal
War neuroses and the adaptive function of REM lobe in humans. Psychol. Sci., 9, 490–493.
sleep. Br. J. Med. Psychol., 45, 27–33. Lavie, P. (2001). Sleep disturbances in the wake of trau-
Greenberg, R., Pillard, R., & Pearlman, C. (1972b). The matic events. N. Engl. J. Med., 345, 1825–1832.
effect of dream (stage REM) deprivation on adapta- Levonian, E. (1972). Retention over time in relation to
tion to stress. Psychosom. Med., 34, 257–262. arousal during learning: an explanation of discrepant
Hammond, N. (2004). Fragmentary Voices: Memory and Edu- results. Acta Psychol. (Amst.), 36, 290–321.
cation at Port-Royal. Gunter Narr Verlag. Llinas, R., & Ribary, U. (1993). Coherent 40-Hz oscilla-
Harrison, Y., & Horne, J. A. (2000). Sleep loss and tem- tion characterizes dream state in humans. Proc. Natl.
poral memory. Q. J. Exp. Psychol., 53, 271–279. Acad. Sci. USA, 90, 2078–2081.
Hartley, D. (1801). Observations on Man, His frame, his de- Louie, K., & Wilson, M. A. (2001). Temporally struc-
ity, and his expectations (1749/)6691. Gainesville, FL: tured replay of awake hippocampal ensemble activity
Scholars Facsimile Reprint. during rapid eye movement sleep. Neuron, 29, 145–
Harvey, A. G., Jones, C., & Schmidt, D. A. (2003). Sleep 156.
and posttraumatic stress disorder: a review. Clin. Psy- Lydic, R., & Baghdoyan, H. A. (1988). Handbook Of Be-
chol. Rev., 23, 377–407. havioral State Control: Cellular and Molecular Mechanisms.
Hobson, J. A., McCarley, R. W., & Wyzinski, P. W. (1975). Boca Raton, FL: CRC Press.
Sleep cycle oscillation: Reciprocal discharge by two Marr, D. (1970). A theory for cerebral neocortex. Proc. R.
brainstem neuronal groups. Science, 189, 55–58. Soc. Lond. B Biol. Sci., 176, 161–234.
Horne, J. A. (1985). Sleep function, with particular ref- Marrosu, F., Portas, C., Mascia, M. S., Casu, M. A., Fa,
erence to sleep deprivation. Ann. Clin. Res., 17, 199– M., et al. (1995). Microdialysis measurement of cor-
208. tical and hippocampal acetylcholine release during
Hu, P., Stylos-Allen, M., & Walker, M. P. (2006). sleep-wake cycle in freely moving cats. Brain Res.,
Sleep facilitates consolidation of emotionally arous- 671, 329–332.
ing declarative memory. Psychol. Sci., 17, 891–898. Marshall, L., & Born, J. (2007). The contribution of sleep
Huber, R., Ghilardi, M. F., Massimini, M., Ferrarelli, F., to hippocampus-dependent memory consolidation.
Riedner, B. A., et al. (2006). Arm immobilization Trends Cogn. Sci., 11, 442–450.
causes cortical plastic changes and locally decreases Marshall, L., Helgadottir, H., Molle, M., & Born, J. (2006).
sleep slow wave activity. Nat. Neurosci., 9, 1169– Boosting slow oscillations during sleep potentiates
1176. memory. Nature, 444, 610–613.
Walker: Sleep in Cognition and Emotion 195

McClelland, J. L., McNaughton, B. L., & O’Reilly, R. Plihal, W., & Born, J. (1997). Effects of early and late noc-
C. (1995). Why there are complementary learning turnal sleep on declarative and procedural memory.
systems in the hippocampus and neocortex: insights J. Cogn. Neurosci., 9, 534–547.
from the successes and failures of connectionist mod- Plihal, W., & Born, J. (1999). Memory consolidation in
els of learning and memory. Psychol. Rev., 102, 419– human sleep depends on inhibition of glucocorticoid
457. release. Neuroreport, 10, 2741–2747.
Mazzarello, P. (2000). What dreams may come? Nature, Poe, G. R., Nitz, D. A., McNaughton, B. L., & Barnes, C.
408, 523. A. (2000). Experience-dependent phase-reversal of
McGaugh, J. L. (2004). The amygdala modulates the con- hippocampal neuron firing during REM sleep. Brain
solidation of memories of emotionally arousing ex- Res., 855, 176–180.
periences. Annu. Rev. Neurosci., 27, 1–28. Pole, N. (2007). The psychophysiology of posttraumatic
Meier-Koll, A., Bussmann, B., Schmidt, C., & stress disorder: a meta-analysis. Psychol. Bull., 133,
Neuschwander, D. (1999). Walking through a maze 725–746.
alters the architecture of sleep. Percept. Mot. Skills, 88, Power, A. E. (2004). Muscarinic cholinergic contribution
1141–1159. to memory consolidation: with attention to involve-
Miller, G. (2007). Neuroscience. Hunting for meaning ment of the basolateral amygdala. Curr. Med. Chem.,
after midnight. Science, 315, 1360–1363. 11, 987–996.
Morris, G. O., Williams, H. L., & Lubin, A. (1960). Mis- Rasch, B., Buchel, C., Gais, S., & Born, J. (2007).
perception and disorientation during sleep. Arch. Gen. Odor cues during slow-wave sleep prompt declara-
Psychiatry, 2, 247–254. tive memory consolidation. Science, 315, 1426–1429.
Moscovitch, M., & Nadel, L. (1998). Consolidation and Ribeiro, S., Gervasoni, D., Soares, E. S., Zhou, Y., Lin, S.
the hippocampal complex revisited: in defense of the C., et al. (2004). Long-lasting novelty-induced neu-
multiple-trace model. Curr. Opin. Neurobiol., 8, 297– ronal reverberation during slow-wave sleep in multi-
300. ple forebrain areas induction of hippocampal long-
Nadel, L., & Moscovitch, M. (1997). Memory consol- term potentiation during waking leads to increased
idation, retrograde amnesia and the hippocampal extrahippocampal zif-268 expression during ensuing
complex. Curr. Opin. Neurobiol., 7, 217–227. rapid-eye-movement sleep. PLoS Biol., 2, E24.
New, A. S., Hazlett, E. A., Buchsbaum, M. S., Good- Rechtschaffen, A., & Kales, A. (1968). A manual standard-
man, M., Mitelman, S. A., et al. (2007). Amygdala- ized terminology, techniques and scoring system for sleep stages
prefrontal disconnection in borderline personality of human subjects. Bethesda, Maryland, USA: U.S. De-
disorder. Neuropsychopharmacology, 32, 1629–1640. partment of Health.
Nishida, M., Pearsall, J., Buckner, R. L., & Walker, M. Robertson, E. M. (2009). From creation to consolidation: a
P. (In Press). Prefrontal theta during REM sleep en- novel framework for memory processing. PLoS Biol.,
hances emotional memory. Cereb. Cortex. 7, e19.
Nishida, M., & Walker, M. P. (2007). Daytime naps, motor Robertson, E. M., Pascual-Leone, A., & Miall, R. C.
memory consolidation and regionally specific sleep (2004). Current concepts in procedural consolida-
spindles. PloS One, 2, e341. tion. Nat. Rev. Neurosci., 5, 576–82.
Nofzinger, E. A. (2005). Functional neuroimaging of Schabus, M., Dang-Vu, T. T., Albouy, G., Balteau, E.,
sleep. Semin. Neurol., 25, 9–18. Boly, M., et al. (2007). Hemodynamic cerebral cor-
Pace-Schott, E. F., & Hobson, J. A. (2002). The neuro- relates of sleep spindles during human non-rapid
biology of sleep: genetics, cellular physiology and eye movement sleep. Proc. Natl. Acad. Sci. USA, 104,
subcortical networks. Nat. Rev. Neurosci., 3, 591–605. 13164–13169.
Paller, K. A., & Wagner, A. D. (2002). Observing the Sejnowski, T. J., & Destexhe, A. (2000). Why do we sleep?
transformation of experience into memory. Trends Brain Res., 886, 208–223.
Cogn. Sci., 6, 93–102. Shaffery, J., Hoffmann, R., & Armitage, R. (2003). The
Pare, D., Collins, D. R., & Pelletier, J. G. (2002). Amyg- neurobiology of depression: perspectives from ani-
dala oscillations and the consolidation of emotional mal and human sleep studies. Neuroscientist, 9, 82–98.
memories. Trends Cogn. Sci., 6, 306–314. Sharot, T., & Phelps, E. A. (2004). How arousal modulates
Peigneux, P., Laureys, S., Fuchs, S., Collette, F., Perrin, memory: disentangling the effects of attention and
F., et al. (2004). Are spatial memories strengthened retention. Cogn. Affect Behav. Neurosci., 4, 294–306.
in the human hippocampus during slow wave sleep? Shima, K., Nakahama, H., & Yamamoto, M. (1986). Fir-
Neuron, 44, 535–545. ing properties of two types of nucleus raphe dor-
Phelps, E. A. (2004). Human emotion and memory: inter- salis neurons during the sleep-waking cycle and their
actions of the amygdala and hippocampal complex. responses to sensory stimuli. Brain Res., 399, 317–
Curr. Opin. Neurobiol., 14, 198–202. 326.
196 Annals of the New York Academy of Sciences

Skaggs, W. E., & McNaughton, B. L. (1996). Replay of Takashima, A., Petersson, K. M., Rutters, F., Tendolkar,
neuronal firing sequences in rat hippocampus during I., Jensen, O., et al. (2006). Declarative memory con-
sleep following spatial experience. Science, 271, 1870– solidation in humans: A prospective functional mag-
1873. netic resonance imaging study. Proc. Natl. Acad. Sci.
Smith, C. (2001). Sleep states and memory processes in USA, 103, 756–761.
humans: procedural versus declarative memory sys- Thurstone, L. L., & Thurstone, T. G. (1941). Factorial
tems. Sleep Med. Rev., 5, 491–506. studies of intelligence. Psychometric Monogr., 2, 94.
Smith, C. T., Aubrey, J. B., & Peters, K. R. (2004). Dif- Tononi, G., & Cirelli, C. (2003). Sleep and synaptic
ferent roles for REM and stage 2 sleep in motor homeostasis: a hypothesis. Brain Res. Bull., 62, 143–
learning: A proposed model. Psychologica Belgica, 44, 150.
81–104. Tononi, G., & Cirelli, C. (2006). Sleep function and synap-
Sotres-Bayon, F., Bush, D. E., & LeDoux, J. E. (2004). tic homeostasis. Sleep Med. Rev., 10, 49–62.
Emotional perseveration: an update on prefrontal- Tse, D., Langston, R. F., Kakeyama, M., Bethus, I.,
amygdala interactions in fear extinction. Learn. Mem., Spooner, P. A., et al. (2007). Schemas and memory
11, 525–535. consolidation. Science, 316, 76–82.
Squire, L. R. (1992). Memory and the hippocampus: a Tsuno, N., Besset, A., & Ritchie, K. (2005). Sleep and
synthesis from findings with rats, monkeys, and hu- depression. J. Clin. Psychiatry, 66, 1254–1269.
mans. Psychol. Rev., 99, 195–231. Vazquez, J., & Baghdoyan, H. A. (2001). Basal forebrain
Squire, L. R. (2004). Memory systems of the brain: a acetylcholine release during REM sleep is signifi-
brief history and current perspective. Neurobiol. Learn cantly greater than during waking. Am. J. Physiol.
Mem., 82, 171–177. Regul. Integr. Comp. Physiol., 280, R598–601.
Squire, L. R. (2007). Neuroscience. Rapid consolidation. Vertes, R. P., & Siegel, J. M. (2005). Time for the sleep
Science, 316, 57–58. community to take a critical look at the purported
Squire, L. R., & Zola, S. M. (1996). Structure and function role of sleep in memory processing. Sleep, 28, 1228–
of declarative and nondeclarative memory systems. 1229; discussion 1230–1233.
Proc. Natl. Acad. Sci. USA, 93, 13515–13522. Wagner, U., Fischer, S., & Born, J. (2002). Changes in
Steriade, M. (2001). The Intact and Sliced Brain. Cambridge, emotional responses to aversive pictures across peri-
MA: MIT Press. ods rich in slow-wave sleep versus rapid eye move-
Steriade, M., & Amzica, F. (1998). Coalescence of sleep ment sleep. Psychosom. Med., 64, 627–634.
rhythms and their chronology in corticothalamic Wagner, U., Gais, S., & Born, J. (2001). Emotional mem-
networks. Sleep Res Online, 1, 1–10. ory formation is enhanced across sleep intervals with
Steriade, M., Amzica, F., & Contreras, D. (1996). Syn- high amounts of rapid eye movement sleep. Learn.
chronization of fast (30–40 Hz) spontaneous corti- Mem., 8, 112–119.
cal rhythms during brain activation. J. Neurosci., 16, Wagner, U., Gais, S., Haider, H., Verleger, R., & Born, J.
392–417. (2004). Sleep inspires insight. Nature, 427, 352–355.
Sterpenich, V., Albouy, G., Boly, M., Vandewalle, Wagner, U., Hallschmid, M., Rasch, B., & Born, J. (2006).
G., Darsaud, A., et al. (2007). Sleep-related Brief sleep after learning keeps emotional memories
hippocampo-cortical interplay during emotional alive for years. Biol. Psychiatry, 60, 788–790.
memory recollection. PLoS Biol., 5, e282. Walker, E. L., & Tarte, R. D. (1963). Memory storage as
Stickgold, R., Scott, L., Rittenhouse, C., & Hobson, J. A. a function of arousal and time with homogeneous
(1999). Sleep-induced changes in associative mem- and heterogeneous lists. J. Verbal Learn. Verbal Behav.,
ory. J. Cogn. Neurosci., 11, 182–193. 2, 113–119.
Stickgold, R., & Walker, M. P. (2005a). Memory consoli- Walker, M. P. (2005). A refined model of sleep and the
dation and reconsolidation: What is the role of sleep? time course of memory formation. Behav. Brain Sci.,
Trends Neurosci., 28, 408–415. 28, 51–64.
Stickgold, R., & Walker, M. P. (2005b). Sleep and Walker, M. P., Liston, C., Hobson, J. A., & Stickgold, R.
memory: the ongoing debate. Sleep, 28, 1225– (2002). Cognitive flexibility across the sleep-wake cy-
1227. cle: REM-sleep enhancement of anagram problem
Strathern, P. (2000). Mendeleyev’s Dream: The Quest for the solving. Brain Res. Cogn. Brain Res., 14, 317–324.
Elements. London: Hamish Hamilton. Walker, M. P., & Stickgold, R. (2006). Sleep, memory and
Sullivan, G. M., Coplan, J. D., Kent, J. M., & Gorman, plasticity. Annu. Rev. Psychol., 10, 139–166.
J. M. (1999). The noradrenergic system in patho- Werth, E., Achermann, P., Dijk, D. J., & Borbely, A. A.
logical anxiety: a focus on panic with relevance to (1997). Spindle frequency activity in the sleep EEG:
generalized anxiety and phobias. Biol. Psychiatry, 46, individual differences and topographic distribution.
1205–1218. Electroencephalogr. Clin. Neurophysiol., 103, 535–542.
Walker: Sleep in Cognition and Emotion 197

Wilson, M. A., & McNaughton, B. L. (1994). Reactivation P. (2007b). A deficit in the ability to form new hu-
of hippocampal ensemble memories during sleep. man memories without sleep. Nat. Neurosci., 10, 385–
Science, 265, 676–679. 392.
Wixted, J. T., & Carpenter, S. K. (2007). The Wickelgren Zeitlhofer, J., Gruber, G., Anderer, P., Asenbaum, S.,
power law and the Ebbinghaus savings function. Psy- Schimicek, P., et al. (1997). Topographic distribu-
chol. Sci., 18, 133–134. tion of sleep spindles in young healthy subjects. J.
Yoo, S. S., Gujar, N., Hu, P., Jolesz, F. A., & Walker, Sleep Res., 6, 149–155.
M. P. (2007a). The human emotional brain without Zohar, D., Tzischinsky, O., Epstein, R., & Lavie, P. (2005).
sleep – a prefrontal amygdala disconnect. Curr. Biol., The effects of sleep loss on medical residents’ emo-
17, R877–R878. tional reactions to work events: a cognitive-energy
Yoo, S. S., Hu, P. T., Gujar, N., Jolesz, F. A., & Walker, M. model. Sleep, 28, 47–54.